qemu-e2k/target/ppc/mmu-radix64.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

778 lines
26 KiB
C
Raw Normal View History

target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
/*
* PowerPC Radix MMU mulation helpers for QEMU.
*
* Copyright (c) 2016 Suraj Jitindar Singh, IBM Corporation
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
#include "qemu/osdep.h"
#include "cpu.h"
#include "exec/exec-all.h"
#include "qemu/error-report.h"
#include "sysemu/kvm.h"
#include "kvm_ppc.h"
#include "exec/log.h"
#include "internal.h"
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
#include "mmu-radix64.h"
#include "mmu-book3s-v3.h"
static bool ppc_radix64_get_fully_qualified_addr(const CPUPPCState *env,
vaddr eaddr,
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
uint64_t *lpid, uint64_t *pid)
{
/* When EA(2:11) are nonzero, raise a segment interrupt */
if (eaddr & ~R_EADDR_VALID_MASK) {
return false;
}
if (FIELD_EX64(env->msr, MSR, HV)) { /* MSR[HV] -> Hypervisor/bare metal */
switch (eaddr & R_EADDR_QUADRANT) {
case R_EADDR_QUADRANT0:
*lpid = 0;
*pid = env->spr[SPR_BOOKS_PID];
break;
case R_EADDR_QUADRANT1:
*lpid = env->spr[SPR_LPIDR];
*pid = env->spr[SPR_BOOKS_PID];
break;
case R_EADDR_QUADRANT2:
*lpid = env->spr[SPR_LPIDR];
*pid = 0;
break;
case R_EADDR_QUADRANT3:
*lpid = 0;
*pid = 0;
break;
default:
g_assert_not_reached();
}
} else { /* !MSR[HV] -> Guest */
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
switch (eaddr & R_EADDR_QUADRANT) {
case R_EADDR_QUADRANT0: /* Guest application */
*lpid = env->spr[SPR_LPIDR];
*pid = env->spr[SPR_BOOKS_PID];
break;
case R_EADDR_QUADRANT1: /* Illegal */
case R_EADDR_QUADRANT2:
return false;
case R_EADDR_QUADRANT3: /* Guest OS */
*lpid = env->spr[SPR_LPIDR];
*pid = 0; /* pid set to 0 -> addresses guest operating system */
break;
default:
g_assert_not_reached();
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
}
}
return true;
}
static void ppc_radix64_raise_segi(PowerPCCPU *cpu, MMUAccessType access_type,
vaddr eaddr)
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
{
CPUState *cs = CPU(cpu);
CPUPPCState *env = &cpu->env;
switch (access_type) {
case MMU_INST_FETCH:
/* Instruction Segment Interrupt */
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
cs->exception_index = POWERPC_EXCP_ISEG;
break;
case MMU_DATA_STORE:
case MMU_DATA_LOAD:
/* Data Segment Interrupt */
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
cs->exception_index = POWERPC_EXCP_DSEG;
env->spr[SPR_DAR] = eaddr;
break;
default:
g_assert_not_reached();
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
}
env->error_code = 0;
}
static inline const char *access_str(MMUAccessType access_type)
{
return access_type == MMU_DATA_LOAD ? "reading" :
(access_type == MMU_DATA_STORE ? "writing" : "execute");
}
static void ppc_radix64_raise_si(PowerPCCPU *cpu, MMUAccessType access_type,
vaddr eaddr, uint32_t cause)
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
{
CPUState *cs = CPU(cpu);
CPUPPCState *env = &cpu->env;
qemu_log_mask(CPU_LOG_MMU, "%s for %s @0x%"VADDR_PRIx" cause %08x\n",
__func__, access_str(access_type),
eaddr, cause);
switch (access_type) {
case MMU_INST_FETCH:
/* Instruction Storage Interrupt */
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
cs->exception_index = POWERPC_EXCP_ISI;
env->error_code = cause;
break;
case MMU_DATA_STORE:
cause |= DSISR_ISSTORE;
/* fall through */
case MMU_DATA_LOAD:
/* Data Storage Interrupt */
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
cs->exception_index = POWERPC_EXCP_DSI;
env->spr[SPR_DSISR] = cause;
env->spr[SPR_DAR] = eaddr;
env->error_code = 0;
break;
default:
g_assert_not_reached();
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
}
}
static void ppc_radix64_raise_hsi(PowerPCCPU *cpu, MMUAccessType access_type,
vaddr eaddr, hwaddr g_raddr, uint32_t cause)
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
{
CPUState *cs = CPU(cpu);
CPUPPCState *env = &cpu->env;
env->error_code = 0;
if (cause & DSISR_PRTABLE_FAULT) {
/* HDSI PRTABLE_FAULT gets the originating access type in error_code */
env->error_code = access_type;
access_type = MMU_DATA_LOAD;
}
qemu_log_mask(CPU_LOG_MMU, "%s for %s @0x%"VADDR_PRIx" 0x%"
HWADDR_PRIx" cause %08x\n",
__func__, access_str(access_type),
eaddr, g_raddr, cause);
switch (access_type) {
case MMU_INST_FETCH:
/* H Instruction Storage Interrupt */
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
cs->exception_index = POWERPC_EXCP_HISI;
env->spr[SPR_ASDR] = g_raddr;
env->error_code = cause;
break;
case MMU_DATA_STORE:
cause |= DSISR_ISSTORE;
/* fall through */
case MMU_DATA_LOAD:
/* H Data Storage Interrupt */
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
cs->exception_index = POWERPC_EXCP_HDSI;
env->spr[SPR_HDSISR] = cause;
env->spr[SPR_HDAR] = eaddr;
env->spr[SPR_ASDR] = g_raddr;
break;
default:
g_assert_not_reached();
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
}
}
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
static bool ppc_radix64_check_prot(PowerPCCPU *cpu, MMUAccessType access_type,
uint64_t pte, int *fault_cause, int *prot,
target/ppc: fix address translation bug for radix mmus This commit attempts to fix a technical hiccup first mentioned by Richard Henderson in https://lists.nongnu.org/archive/html/qemu-devel/2021-05/msg06247.html To sumarize the hiccup here, when radix-style mmus are translating an address, they might need to call a second level of translation, with hypervisor privileges. However, the way it was being done up until this point meant that the second level translation had the same privileges as the first level. It could lead to a bug in address translation when running KVM inside a TCG guest, but this bug was never experienced by users, so this isn't as much a bug fix as it is a correctness cleanup. This patch attempts that cleanup by making radix64_*_xlate functions receive the mmu_idx, and passing one with the correct permission for the second level translation. The mmuidx macros added by this patch are only correct for non-bookE mmus, because BookE style set the IS and DS bits inverted and there might be other subtle differences. However, there doesn't seem to be BookE cpus that have radix-style mmus, so we left a comment there to document the issue, in case a machine does have that and was missed. As part of this cleanup, we now need to send the correct mmmu_idx when calling get_phys_page_debug, otherwise we might not be able to see the memory that the CPU could Suggested-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Cédric Le Goater <clg@kaod.org> Tested-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20210628133610.1143-2-bruno.larsen@eldorado.org.br> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2021-06-28 15:36:08 +02:00
int mmu_idx, bool partition_scoped)
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
{
CPUPPCState *env = &cpu->env;
int need_prot;
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
/* Check Page Attributes (pte58:59) */
if ((pte & R_PTE_ATT) == R_PTE_ATT_NI_IO && access_type == MMU_INST_FETCH) {
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
/*
* Radix PTE entries with the non-idempotent I/O attribute are treated
* as guarded storage
*/
*fault_cause |= SRR1_NOEXEC_GUARD;
return true;
}
/* Determine permissions allowed by Encoded Access Authority */
if (!partition_scoped && (pte & R_PTE_EAA_PRIV) &&
FIELD_EX64(env->msr, MSR, PR)) {
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
*prot = 0;
target/ppc: fix address translation bug for radix mmus This commit attempts to fix a technical hiccup first mentioned by Richard Henderson in https://lists.nongnu.org/archive/html/qemu-devel/2021-05/msg06247.html To sumarize the hiccup here, when radix-style mmus are translating an address, they might need to call a second level of translation, with hypervisor privileges. However, the way it was being done up until this point meant that the second level translation had the same privileges as the first level. It could lead to a bug in address translation when running KVM inside a TCG guest, but this bug was never experienced by users, so this isn't as much a bug fix as it is a correctness cleanup. This patch attempts that cleanup by making radix64_*_xlate functions receive the mmu_idx, and passing one with the correct permission for the second level translation. The mmuidx macros added by this patch are only correct for non-bookE mmus, because BookE style set the IS and DS bits inverted and there might be other subtle differences. However, there doesn't seem to be BookE cpus that have radix-style mmus, so we left a comment there to document the issue, in case a machine does have that and was missed. As part of this cleanup, we now need to send the correct mmmu_idx when calling get_phys_page_debug, otherwise we might not be able to see the memory that the CPU could Suggested-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Cédric Le Goater <clg@kaod.org> Tested-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20210628133610.1143-2-bruno.larsen@eldorado.org.br> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2021-06-28 15:36:08 +02:00
} else if (mmuidx_pr(mmu_idx) || (pte & R_PTE_EAA_PRIV) ||
partition_scoped) {
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
*prot = ppc_radix64_get_prot_eaa(pte);
} else { /* !MSR_PR && !(pte & R_PTE_EAA_PRIV) && !partition_scoped */
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
*prot = ppc_radix64_get_prot_eaa(pte);
*prot &= ppc_radix64_get_prot_amr(cpu); /* Least combined permissions */
}
/* Check if requested access type is allowed */
need_prot = prot_for_access_type(access_type);
if (need_prot & ~*prot) { /* Page Protected for that Access */
*fault_cause |= access_type == MMU_INST_FETCH ? SRR1_NOEXEC_GUARD :
DSISR_PROTFAULT;
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
return true;
}
return false;
}
static int ppc_radix64_check_rc(MMUAccessType access_type, uint64_t pte)
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
{
switch (access_type) {
case MMU_DATA_STORE:
if (!(pte & R_PTE_C)) {
break;
}
/* fall through */
case MMU_INST_FETCH:
case MMU_DATA_LOAD:
if (!(pte & R_PTE_R)) {
break;
}
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
/* R/C bits are already set appropriately for this access */
return 0;
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
}
return 1;
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
}
static bool ppc_radix64_is_valid_level(int level, int psize, uint64_t nls)
{
bool ret;
/*
* Check if this is a valid level, according to POWER9 and POWER10
* Processor User's Manuals, sections 4.10.4.1 and 5.10.6.1, respectively:
* Supported Radix Tree Configurations and Resulting Page Sizes.
*
* Note: these checks are specific to POWER9 and POWER10 CPUs. Any future
* CPUs that supports a different Radix MMU configuration will need their
* own implementation.
*/
switch (level) {
case 0: /* Root Page Dir */
ret = psize == 52 && nls == 13;
break;
case 1:
case 2:
ret = nls == 9;
break;
case 3:
ret = nls == 9 || nls == 5;
break;
default:
ret = false;
}
if (unlikely(!ret)) {
qemu_log_mask(LOG_GUEST_ERROR, "invalid radix configuration: "
"level %d size %d nls %"PRIu64"\n",
level, psize, nls);
}
return ret;
}
static int ppc_radix64_next_level(AddressSpace *as, vaddr eaddr,
uint64_t *pte_addr, uint64_t *nls,
int *psize, uint64_t *pte, int *fault_cause)
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
{
uint64_t index, mask, nlb, pde;
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
/* Read page <directory/table> entry from guest address space */
pde = ldq_phys(as, *pte_addr);
if (!(pde & R_PTE_VALID)) { /* Invalid Entry */
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
*fault_cause |= DSISR_NOPTE;
return 1;
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
}
*pte = pde;
*psize -= *nls;
if (!(pde & R_PTE_LEAF)) { /* Prepare for next iteration */
*nls = pde & R_PDE_NLS;
index = eaddr >> (*psize - *nls); /* Shift */
index &= ((1UL << *nls) - 1); /* Mask */
nlb = pde & R_PDE_NLB;
mask = MAKE_64BIT_MASK(0, *nls + 3);
if (nlb & mask) {
qemu_log_mask(LOG_GUEST_ERROR,
"%s: misaligned page dir/table base: 0x"TARGET_FMT_lx
" page dir size: 0x"TARGET_FMT_lx"\n",
__func__, nlb, mask + 1);
nlb &= ~mask;
}
*pte_addr = nlb + index * sizeof(pde);
}
return 0;
}
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
static int ppc_radix64_walk_tree(AddressSpace *as, vaddr eaddr,
uint64_t base_addr, uint64_t nls,
hwaddr *raddr, int *psize, uint64_t *pte,
int *fault_cause, hwaddr *pte_addr)
{
uint64_t index, pde, rpn, mask;
int level = 0;
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
index = eaddr >> (*psize - nls); /* Shift */
index &= ((1UL << nls) - 1); /* Mask */
mask = MAKE_64BIT_MASK(0, nls + 3);
if (base_addr & mask) {
qemu_log_mask(LOG_GUEST_ERROR,
"%s: misaligned page dir base: 0x"TARGET_FMT_lx
" page dir size: 0x"TARGET_FMT_lx"\n",
__func__, base_addr, mask + 1);
base_addr &= ~mask;
}
*pte_addr = base_addr + index * sizeof(pde);
do {
int ret;
if (!ppc_radix64_is_valid_level(level++, *psize, nls)) {
*fault_cause |= DSISR_R_BADCONFIG;
return 1;
}
ret = ppc_radix64_next_level(as, eaddr, pte_addr, &nls, psize, &pde,
fault_cause);
if (ret) {
return ret;
}
} while (!(pde & R_PTE_LEAF));
*pte = pde;
rpn = pde & R_PTE_RPN;
mask = (1UL << *psize) - 1;
/* Or high bits of rpn and low bits to ea to form whole real addr */
*raddr = (rpn & ~mask) | (eaddr & mask);
return 0;
target/ppc: Implement ISA V3.00 radix page fault handler ISA V3.00 introduced a new radix mmu model. Implement the page fault handler for this so we can run a tcg guest in radix mode and perform address translation correctly. In real mode (mmu turned off) addresses are masked to remove the top 4 bits and then are subject to partition scoped translation, since we only support pseries at this stage it is only necessary to perform the masking and then we're done. In virtual mode (mmu turned on) address translation if performed as follows: 1. Use the quadrant to determine the fully qualified address. The fully qualified address is defined as the combination of the effective address, the effective logical partition id (LPID) and the effective process id (PID). Based on the quadrant (EA63:62) we set the pid and lpid like so: quadrant 0: lpid = LPIDR, pid = PIDR quadrant 1: HV only (not allowed in pseries) quadrant 2: HV only (not allowed in pseries) quadrant 3: lpid = LPIDR, pid = 0 If we can't get the fully qualified address we raise a segment interrupt. 2. Find the guest radix tree We ask the virtual hypervisor for the partition table which was registered with H_REGISTER_PROC_TBL which points us to the process table in guest memory. We then index this table by pid to get the process table entry which points us to the appropriate radix tree to translate the address. If the process table isn't big enough to contain an entry for the current pid then we raise a storage interrupt. 3. Walk the radix tree Next we walk the radix tree where each level is a table of page directory entries indexed by some number of bits from the effective address, where the number of bits is determined by the table size. We continue to walk the tree (while entries are valid and the table is of minimum size) until we reach a table of page table entries, indicated by having the leaf bit set. The appropriate pte is then checked for sufficient access permissions, the reference and change bits are updated and the real address is calculated from the real page number bits of the pte and the low bits of the effective address. If we can't find an entry or can't access the entry bacause of permissions then we raise a storage interrupt. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> [dwg: Add missing parentheses to macro] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2017-05-02 08:37:17 +02:00
}
static bool validate_pate(PowerPCCPU *cpu, uint64_t lpid, ppc_v3_pate_t *pate)
{
CPUPPCState *env = &cpu->env;
if (!(pate->dw0 & PATE0_HR)) {
return false;
}
if (lpid == 0 && !FIELD_EX64(env->msr, MSR, HV)) {
return false;
}
if ((pate->dw0 & PATE1_R_PRTS) < 5) {
return false;
}
/* More checks ... */
return true;
}
static int ppc_radix64_partition_scoped_xlate(PowerPCCPU *cpu,
MMUAccessType orig_access_type,
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
vaddr eaddr, hwaddr g_raddr,
ppc_v3_pate_t pate,
hwaddr *h_raddr, int *h_prot,
int *h_page_size, bool pde_addr,
int mmu_idx, uint64_t lpid,
bool guest_visible)
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
{
MMUAccessType access_type = orig_access_type;
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
int fault_cause = 0;
hwaddr pte_addr;
uint64_t pte;
if (pde_addr) {
/*
* Translation of process-scoped tables/directories is performed as
* a read-access.
*/
access_type = MMU_DATA_LOAD;
}
qemu_log_mask(CPU_LOG_MMU, "%s for %s @0x%"VADDR_PRIx
" mmu_idx %u 0x%"HWADDR_PRIx"\n",
__func__, access_str(access_type),
eaddr, mmu_idx, g_raddr);
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
*h_page_size = PRTBE_R_GET_RTS(pate.dw0);
/* No valid pte or access denied due to protection */
if (ppc_radix64_walk_tree(CPU(cpu)->as, g_raddr, pate.dw0 & PRTBE_R_RPDB,
pate.dw0 & PRTBE_R_RPDS, h_raddr, h_page_size,
&pte, &fault_cause, &pte_addr) ||
target/ppc: fix address translation bug for radix mmus This commit attempts to fix a technical hiccup first mentioned by Richard Henderson in https://lists.nongnu.org/archive/html/qemu-devel/2021-05/msg06247.html To sumarize the hiccup here, when radix-style mmus are translating an address, they might need to call a second level of translation, with hypervisor privileges. However, the way it was being done up until this point meant that the second level translation had the same privileges as the first level. It could lead to a bug in address translation when running KVM inside a TCG guest, but this bug was never experienced by users, so this isn't as much a bug fix as it is a correctness cleanup. This patch attempts that cleanup by making radix64_*_xlate functions receive the mmu_idx, and passing one with the correct permission for the second level translation. The mmuidx macros added by this patch are only correct for non-bookE mmus, because BookE style set the IS and DS bits inverted and there might be other subtle differences. However, there doesn't seem to be BookE cpus that have radix-style mmus, so we left a comment there to document the issue, in case a machine does have that and was missed. As part of this cleanup, we now need to send the correct mmmu_idx when calling get_phys_page_debug, otherwise we might not be able to see the memory that the CPU could Suggested-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Cédric Le Goater <clg@kaod.org> Tested-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20210628133610.1143-2-bruno.larsen@eldorado.org.br> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2021-06-28 15:36:08 +02:00
ppc_radix64_check_prot(cpu, access_type, pte,
&fault_cause, h_prot, mmu_idx, true)) {
if (pde_addr) { /* address being translated was that of a guest pde */
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
fault_cause |= DSISR_PRTABLE_FAULT;
}
if (guest_visible) {
ppc_radix64_raise_hsi(cpu, orig_access_type,
eaddr, g_raddr, fault_cause);
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
}
return 1;
}
if (guest_visible) {
if (ppc_radix64_check_rc(access_type, pte)) {
/*
* Per ISA 3.1 Book III, 7.5.3 and 7.5.5, failure to set R/C during
* partition-scoped translation when effLPID = 0 results in normal
* (non-Hypervisor) Data and Instruction Storage Interrupts
* respectively.
*
* ISA 3.0 is ambiguous about this, but tests on POWER9 hardware
* seem to exhibit the same behavior.
*/
if (lpid > 0) {
ppc_radix64_raise_hsi(cpu, access_type, eaddr, g_raddr,
DSISR_ATOMIC_RC);
} else {
ppc_radix64_raise_si(cpu, access_type, eaddr, DSISR_ATOMIC_RC);
}
return 1;
}
}
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
return 0;
}
/*
* The spapr vhc has a flat partition scope provided by qemu memory when
* not nested.
*
* When running a nested guest, the addressing is 2-level radix on top of the
* vhc memory, so it works practically identically to the bare metal 2-level
* radix. So that code is selected directly. A cleaner and more flexible nested
* hypervisor implementation would allow the vhc to provide a ->nested_xlate()
* function but that is not required for the moment.
*/
static bool vhyp_flat_addressing(PowerPCCPU *cpu)
{
if (cpu->vhyp) {
return !vhyp_cpu_in_nested(cpu);
}
return false;
}
static int ppc_radix64_process_scoped_xlate(PowerPCCPU *cpu,
MMUAccessType access_type,
vaddr eaddr, uint64_t pid,
ppc_v3_pate_t pate, hwaddr *g_raddr,
int *g_prot, int *g_page_size,
int mmu_idx, uint64_t lpid,
bool guest_visible)
{
CPUState *cs = CPU(cpu);
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
CPUPPCState *env = &cpu->env;
uint64_t offset, size, prtb, prtbe_addr, prtbe0, base_addr, nls, index, pte;
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
int fault_cause = 0, h_page_size, h_prot;
hwaddr h_raddr, pte_addr;
int ret;
qemu_log_mask(CPU_LOG_MMU, "%s for %s @0x%"VADDR_PRIx
" mmu_idx %u pid %"PRIu64"\n",
__func__, access_str(access_type),
eaddr, mmu_idx, pid);
prtb = (pate.dw1 & PATE1_R_PRTB);
size = 1ULL << ((pate.dw1 & PATE1_R_PRTS) + 12);
if (prtb & (size - 1)) {
/* Process Table not properly aligned */
if (guest_visible) {
ppc_radix64_raise_si(cpu, access_type, eaddr, DSISR_R_BADCONFIG);
}
return 1;
}
/* Index Process Table by PID to Find Corresponding Process Table Entry */
offset = pid * sizeof(struct prtb_entry);
if (offset >= size) {
/* offset exceeds size of the process table */
if (guest_visible) {
ppc_radix64_raise_si(cpu, access_type, eaddr, DSISR_NOPTE);
}
return 1;
}
prtbe_addr = prtb + offset;
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
if (vhyp_flat_addressing(cpu)) {
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
prtbe0 = ldq_phys(cs->as, prtbe_addr);
} else {
/*
* Process table addresses are subject to partition-scoped
* translation
*
* On a Radix host, the partition-scoped page table for LPID=0
* is only used to translate the effective addresses of the
* process table entries.
*/
/* mmu_idx is 5 because we're translating from hypervisor scope */
ret = ppc_radix64_partition_scoped_xlate(cpu, access_type, eaddr,
prtbe_addr, pate, &h_raddr,
&h_prot, &h_page_size, true,
5, lpid, guest_visible);
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
if (ret) {
return ret;
}
prtbe0 = ldq_phys(cs->as, h_raddr);
}
/* Walk Radix Tree from Process Table Entry to Convert EA to RA */
*g_page_size = PRTBE_R_GET_RTS(prtbe0);
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
base_addr = prtbe0 & PRTBE_R_RPDB;
nls = prtbe0 & PRTBE_R_RPDS;
if (FIELD_EX64(env->msr, MSR, HV) || vhyp_flat_addressing(cpu)) {
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
/*
* Can treat process table addresses as real addresses
*/
ret = ppc_radix64_walk_tree(cs->as, eaddr & R_EADDR_MASK, base_addr,
nls, g_raddr, g_page_size, &pte,
&fault_cause, &pte_addr);
if (ret) {
/* No valid PTE */
if (guest_visible) {
ppc_radix64_raise_si(cpu, access_type, eaddr, fault_cause);
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
}
return ret;
}
} else {
uint64_t rpn, mask;
int level = 0;
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
index = (eaddr & R_EADDR_MASK) >> (*g_page_size - nls); /* Shift */
index &= ((1UL << nls) - 1); /* Mask */
pte_addr = base_addr + (index * sizeof(pte));
/*
* Each process table address is subject to a partition-scoped
* translation
*/
do {
target/ppc: fix address translation bug for radix mmus This commit attempts to fix a technical hiccup first mentioned by Richard Henderson in https://lists.nongnu.org/archive/html/qemu-devel/2021-05/msg06247.html To sumarize the hiccup here, when radix-style mmus are translating an address, they might need to call a second level of translation, with hypervisor privileges. However, the way it was being done up until this point meant that the second level translation had the same privileges as the first level. It could lead to a bug in address translation when running KVM inside a TCG guest, but this bug was never experienced by users, so this isn't as much a bug fix as it is a correctness cleanup. This patch attempts that cleanup by making radix64_*_xlate functions receive the mmu_idx, and passing one with the correct permission for the second level translation. The mmuidx macros added by this patch are only correct for non-bookE mmus, because BookE style set the IS and DS bits inverted and there might be other subtle differences. However, there doesn't seem to be BookE cpus that have radix-style mmus, so we left a comment there to document the issue, in case a machine does have that and was missed. As part of this cleanup, we now need to send the correct mmmu_idx when calling get_phys_page_debug, otherwise we might not be able to see the memory that the CPU could Suggested-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Cédric Le Goater <clg@kaod.org> Tested-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20210628133610.1143-2-bruno.larsen@eldorado.org.br> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2021-06-28 15:36:08 +02:00
/* mmu_idx is 5 because we're translating from hypervisor scope */
ret = ppc_radix64_partition_scoped_xlate(cpu, access_type, eaddr,
pte_addr, pate, &h_raddr,
&h_prot, &h_page_size,
true, 5, lpid,
guest_visible);
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
if (ret) {
return ret;
}
if (!ppc_radix64_is_valid_level(level++, *g_page_size, nls)) {
fault_cause |= DSISR_R_BADCONFIG;
ret = 1;
} else {
ret = ppc_radix64_next_level(cs->as, eaddr & R_EADDR_MASK,
&h_raddr, &nls, g_page_size,
&pte, &fault_cause);
}
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
if (ret) {
/* No valid pte */
if (guest_visible) {
ppc_radix64_raise_si(cpu, access_type, eaddr, fault_cause);
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
}
return ret;
}
pte_addr = h_raddr;
} while (!(pte & R_PTE_LEAF));
rpn = pte & R_PTE_RPN;
mask = (1UL << *g_page_size) - 1;
/* Or high bits of rpn and low bits to ea to form whole real addr */
*g_raddr = (rpn & ~mask) | (eaddr & mask);
}
target/ppc: fix address translation bug for radix mmus This commit attempts to fix a technical hiccup first mentioned by Richard Henderson in https://lists.nongnu.org/archive/html/qemu-devel/2021-05/msg06247.html To sumarize the hiccup here, when radix-style mmus are translating an address, they might need to call a second level of translation, with hypervisor privileges. However, the way it was being done up until this point meant that the second level translation had the same privileges as the first level. It could lead to a bug in address translation when running KVM inside a TCG guest, but this bug was never experienced by users, so this isn't as much a bug fix as it is a correctness cleanup. This patch attempts that cleanup by making radix64_*_xlate functions receive the mmu_idx, and passing one with the correct permission for the second level translation. The mmuidx macros added by this patch are only correct for non-bookE mmus, because BookE style set the IS and DS bits inverted and there might be other subtle differences. However, there doesn't seem to be BookE cpus that have radix-style mmus, so we left a comment there to document the issue, in case a machine does have that and was missed. As part of this cleanup, we now need to send the correct mmmu_idx when calling get_phys_page_debug, otherwise we might not be able to see the memory that the CPU could Suggested-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Cédric Le Goater <clg@kaod.org> Tested-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20210628133610.1143-2-bruno.larsen@eldorado.org.br> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2021-06-28 15:36:08 +02:00
if (ppc_radix64_check_prot(cpu, access_type, pte, &fault_cause,
g_prot, mmu_idx, false)) {
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
/* Access denied due to protection */
if (guest_visible) {
ppc_radix64_raise_si(cpu, access_type, eaddr, fault_cause);
}
return 1;
}
if (guest_visible) {
/* R/C bits not appropriately set for access */
if (ppc_radix64_check_rc(access_type, pte)) {
ppc_radix64_raise_si(cpu, access_type, eaddr, DSISR_ATOMIC_RC);
return 1;
}
}
return 0;
}
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
/*
* Radix tree translation is a 2 steps translation process:
*
* 1. Process-scoped translation: Guest Eff Addr -> Guest Real Addr
* 2. Partition-scoped translation: Guest Real Addr -> Host Real Addr
*
* MSR[HV]
* +-------------+----------------+---------------+
* | | HV = 0 | HV = 1 |
* +-------------+----------------+---------------+
* | Relocation | Partition | No |
* | = Off | Scoped | Translation |
* Relocation +-------------+----------------+---------------+
* | Relocation | Partition & | Process |
* | = On | Process Scoped | Scoped |
* +-------------+----------------+---------------+
*/
static bool ppc_radix64_xlate_impl(PowerPCCPU *cpu, vaddr eaddr,
MMUAccessType access_type, hwaddr *raddr,
int *psizep, int *protp, int mmu_idx,
bool guest_visible)
{
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
CPUPPCState *env = &cpu->env;
uint64_t lpid, pid;
ppc_v3_pate_t pate;
int psize, prot;
hwaddr g_raddr;
bool relocation;
target/ppc: fix address translation bug for radix mmus This commit attempts to fix a technical hiccup first mentioned by Richard Henderson in https://lists.nongnu.org/archive/html/qemu-devel/2021-05/msg06247.html To sumarize the hiccup here, when radix-style mmus are translating an address, they might need to call a second level of translation, with hypervisor privileges. However, the way it was being done up until this point meant that the second level translation had the same privileges as the first level. It could lead to a bug in address translation when running KVM inside a TCG guest, but this bug was never experienced by users, so this isn't as much a bug fix as it is a correctness cleanup. This patch attempts that cleanup by making radix64_*_xlate functions receive the mmu_idx, and passing one with the correct permission for the second level translation. The mmuidx macros added by this patch are only correct for non-bookE mmus, because BookE style set the IS and DS bits inverted and there might be other subtle differences. However, there doesn't seem to be BookE cpus that have radix-style mmus, so we left a comment there to document the issue, in case a machine does have that and was missed. As part of this cleanup, we now need to send the correct mmmu_idx when calling get_phys_page_debug, otherwise we might not be able to see the memory that the CPU could Suggested-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Cédric Le Goater <clg@kaod.org> Tested-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20210628133610.1143-2-bruno.larsen@eldorado.org.br> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2021-06-28 15:36:08 +02:00
assert(!(mmuidx_hv(mmu_idx) && cpu->vhyp));
target/ppc: fix address translation bug for radix mmus This commit attempts to fix a technical hiccup first mentioned by Richard Henderson in https://lists.nongnu.org/archive/html/qemu-devel/2021-05/msg06247.html To sumarize the hiccup here, when radix-style mmus are translating an address, they might need to call a second level of translation, with hypervisor privileges. However, the way it was being done up until this point meant that the second level translation had the same privileges as the first level. It could lead to a bug in address translation when running KVM inside a TCG guest, but this bug was never experienced by users, so this isn't as much a bug fix as it is a correctness cleanup. This patch attempts that cleanup by making radix64_*_xlate functions receive the mmu_idx, and passing one with the correct permission for the second level translation. The mmuidx macros added by this patch are only correct for non-bookE mmus, because BookE style set the IS and DS bits inverted and there might be other subtle differences. However, there doesn't seem to be BookE cpus that have radix-style mmus, so we left a comment there to document the issue, in case a machine does have that and was missed. As part of this cleanup, we now need to send the correct mmmu_idx when calling get_phys_page_debug, otherwise we might not be able to see the memory that the CPU could Suggested-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Cédric Le Goater <clg@kaod.org> Tested-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20210628133610.1143-2-bruno.larsen@eldorado.org.br> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2021-06-28 15:36:08 +02:00
relocation = !mmuidx_real(mmu_idx);
/* HV or virtual hypervisor Real Mode Access */
if (!relocation && (mmuidx_hv(mmu_idx) || vhyp_flat_addressing(cpu))) {
/* In real mode top 4 effective addr bits (mostly) ignored */
*raddr = eaddr & 0x0FFFFFFFFFFFFFFFULL;
/* In HV mode, add HRMOR if top EA bit is clear */
target/ppc: fix address translation bug for radix mmus This commit attempts to fix a technical hiccup first mentioned by Richard Henderson in https://lists.nongnu.org/archive/html/qemu-devel/2021-05/msg06247.html To sumarize the hiccup here, when radix-style mmus are translating an address, they might need to call a second level of translation, with hypervisor privileges. However, the way it was being done up until this point meant that the second level translation had the same privileges as the first level. It could lead to a bug in address translation when running KVM inside a TCG guest, but this bug was never experienced by users, so this isn't as much a bug fix as it is a correctness cleanup. This patch attempts that cleanup by making radix64_*_xlate functions receive the mmu_idx, and passing one with the correct permission for the second level translation. The mmuidx macros added by this patch are only correct for non-bookE mmus, because BookE style set the IS and DS bits inverted and there might be other subtle differences. However, there doesn't seem to be BookE cpus that have radix-style mmus, so we left a comment there to document the issue, in case a machine does have that and was missed. As part of this cleanup, we now need to send the correct mmmu_idx when calling get_phys_page_debug, otherwise we might not be able to see the memory that the CPU could Suggested-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Cédric Le Goater <clg@kaod.org> Tested-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20210628133610.1143-2-bruno.larsen@eldorado.org.br> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2021-06-28 15:36:08 +02:00
if (mmuidx_hv(mmu_idx) || !env->has_hv_mode) {
if (!(eaddr >> 63)) {
*raddr |= env->spr[SPR_HRMOR];
}
}
*protp = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
*psizep = TARGET_PAGE_BITS;
return true;
}
/*
* Check UPRT (we avoid the check in real mode to deal with
* transitional states during kexec.
*/
if (guest_visible && !ppc64_use_proc_tbl(cpu)) {
qemu_log_mask(LOG_GUEST_ERROR,
"LPCR:UPRT not set in radix mode ! LPCR="
TARGET_FMT_lx "\n", env->spr[SPR_LPCR]);
}
/* Virtual Mode Access - get the fully qualified address */
if (!ppc_radix64_get_fully_qualified_addr(&cpu->env, eaddr, &lpid, &pid)) {
if (guest_visible) {
ppc_radix64_raise_segi(cpu, access_type, eaddr);
}
return false;
}
/* Get Partition Table */
if (cpu->vhyp) {
PPCVirtualHypervisorClass *vhc;
vhc = PPC_VIRTUAL_HYPERVISOR_GET_CLASS(cpu->vhyp);
if (!vhc->get_pate(cpu->vhyp, cpu, lpid, &pate)) {
if (guest_visible) {
ppc_radix64_raise_hsi(cpu, access_type, eaddr, eaddr,
DSISR_R_BADCONFIG);
}
return false;
}
} else {
if (!ppc64_v3_get_pate(cpu, lpid, &pate)) {
if (guest_visible) {
ppc_radix64_raise_hsi(cpu, access_type, eaddr, eaddr,
DSISR_R_BADCONFIG);
}
return false;
}
if (!validate_pate(cpu, lpid, &pate)) {
if (guest_visible) {
ppc_radix64_raise_hsi(cpu, access_type, eaddr, eaddr,
DSISR_R_BADCONFIG);
}
return false;
}
}
*psizep = INT_MAX;
*protp = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
/*
* Perform process-scoped translation if relocation enabled.
*
* - Translates an effective address to a host real address in
* quadrants 0 and 3 when HV=1.
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
*
* - Translates an effective address to a guest real address.
*/
if (relocation) {
int ret = ppc_radix64_process_scoped_xlate(cpu, access_type, eaddr, pid,
pate, &g_raddr, &prot,
&psize, mmu_idx, lpid,
guest_visible);
if (ret) {
return false;
}
*psizep = MIN(*psizep, psize);
*protp &= prot;
} else {
g_raddr = eaddr & R_EADDR_MASK;
}
if (vhyp_flat_addressing(cpu)) {
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
*raddr = g_raddr;
} else {
/*
* Perform partition-scoped translation if !HV or HV access to
* quadrants 1 or 2. Translates a guest real address to a host
* real address.
*/
target/ppc: fix address translation bug for radix mmus This commit attempts to fix a technical hiccup first mentioned by Richard Henderson in https://lists.nongnu.org/archive/html/qemu-devel/2021-05/msg06247.html To sumarize the hiccup here, when radix-style mmus are translating an address, they might need to call a second level of translation, with hypervisor privileges. However, the way it was being done up until this point meant that the second level translation had the same privileges as the first level. It could lead to a bug in address translation when running KVM inside a TCG guest, but this bug was never experienced by users, so this isn't as much a bug fix as it is a correctness cleanup. This patch attempts that cleanup by making radix64_*_xlate functions receive the mmu_idx, and passing one with the correct permission for the second level translation. The mmuidx macros added by this patch are only correct for non-bookE mmus, because BookE style set the IS and DS bits inverted and there might be other subtle differences. However, there doesn't seem to be BookE cpus that have radix-style mmus, so we left a comment there to document the issue, in case a machine does have that and was missed. As part of this cleanup, we now need to send the correct mmmu_idx when calling get_phys_page_debug, otherwise we might not be able to see the memory that the CPU could Suggested-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Cédric Le Goater <clg@kaod.org> Tested-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20210628133610.1143-2-bruno.larsen@eldorado.org.br> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2021-06-28 15:36:08 +02:00
if (lpid || !mmuidx_hv(mmu_idx)) {
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
int ret;
ret = ppc_radix64_partition_scoped_xlate(cpu, access_type, eaddr,
g_raddr, pate, raddr,
&prot, &psize, false,
mmu_idx, lpid,
guest_visible);
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
if (ret) {
return false;
target/ppc: Add support for Radix partition-scoped translation The Radix tree translation model currently supports process-scoped translation for the PowerNV machine (Hypervisor mode) and for the pSeries machine (Guest mode). Guests running under an emulated Hypervisor (PowerNV machine) require a new type of Radix translation, called partition-scoped, which is missing today. The Radix tree translation is a 2 steps process. The first step, process-scoped translation, converts an effective Address to a guest real address, and the second step, partition-scoped translation, converts a guest real address to a host real address. There are difference cases to covers : * Hypervisor real mode access: no Radix translation. * Hypervisor or host application access (quadrant 0 and 3) with relocation on: process-scoped translation. * Guest OS real mode access: only partition-scoped translation. * Guest OS real or guest application access (quadrant 0 and 3) with relocation on: both process-scoped translation and partition-scoped translations. * Hypervisor access in quadrant 1 and 2 with relocation on: both process-scoped translation and partition-scoped translations. The radix tree partition-scoped translation is performed using tables pointed to by the first double-word of the Partition Table Entries and process-scoped translation uses tables pointed to by the Process Table Entries (second double-word of the Partition Table Entries). Both partition-scoped and process-scoped translations process are identical and thus the radix tree traversing code is largely reused. However, errors in partition-scoped translations generate hypervisor exceptions. Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by: Greg Kurz <groug@kaod.org> Signed-off-by: Cédric Le Goater <clg@kaod.org> Message-Id: <20200403140056.59465-5-clg@kaod.org> [dwg: Fixup from Greg Kurz folded in] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-04-03 16:00:56 +02:00
}
*psizep = MIN(*psizep, psize);
*protp &= prot;
} else {
*raddr = g_raddr;
}
}
return true;
}
bool ppc_radix64_xlate(PowerPCCPU *cpu, vaddr eaddr, MMUAccessType access_type,
hwaddr *raddrp, int *psizep, int *protp, int mmu_idx,
bool guest_visible)
{
bool ret = ppc_radix64_xlate_impl(cpu, eaddr, access_type, raddrp,
psizep, protp, mmu_idx, guest_visible);
qemu_log_mask(CPU_LOG_MMU, "%s for %s @0x%"VADDR_PRIx
" mmu_idx %u (prot %c%c%c) -> 0x%"HWADDR_PRIx"\n",
__func__, access_str(access_type),
eaddr, mmu_idx,
*protp & PAGE_READ ? 'r' : '-',
*protp & PAGE_WRITE ? 'w' : '-',
*protp & PAGE_EXEC ? 'x' : '-',
*raddrp);
return ret;
}