From 192b6a780598976feb7321ff007754f8511a4129 Mon Sep 17 00:00:00 2001 From: "Aneesh Kumar K.V" Date: Sun, 12 Jul 2020 18:50:47 +0530 Subject: [PATCH 1/4] powerpc/book3s64/pkeys: Fix pkey_access_permitted() for execute disable pkey Even if the IAMR value denies execute access, the current code returns true from pkey_access_permitted() for an execute permission check, if the AMR read pkey bit is cleared. This results in repeated page fault loop with a test like below: #define _GNU_SOURCE #include #include #include #include #include #include #include #include #include #include #ifdef SYS_pkey_mprotect #undef SYS_pkey_mprotect #endif #ifdef SYS_pkey_alloc #undef SYS_pkey_alloc #endif #ifdef SYS_pkey_free #undef SYS_pkey_free #endif #undef PKEY_DISABLE_EXECUTE #define PKEY_DISABLE_EXECUTE 0x4 #define SYS_pkey_mprotect 386 #define SYS_pkey_alloc 384 #define SYS_pkey_free 385 #define PPC_INST_NOP 0x60000000 #define PPC_INST_BLR 0x4e800020 #define PROT_RWX (PROT_READ | PROT_WRITE | PROT_EXEC) static int sys_pkey_mprotect(void *addr, size_t len, int prot, int pkey) { return syscall(SYS_pkey_mprotect, addr, len, prot, pkey); } static int sys_pkey_alloc(unsigned long flags, unsigned long access_rights) { return syscall(SYS_pkey_alloc, flags, access_rights); } static int sys_pkey_free(int pkey) { return syscall(SYS_pkey_free, pkey); } static void do_execute(void *region) { /* jump to region */ asm volatile( "mtctr %0;" "bctrl" : : "r"(region) : "ctr", "lr"); } static void do_protect(void *region) { size_t pgsize; int i, pkey; pgsize = getpagesize(); pkey = sys_pkey_alloc(0, PKEY_DISABLE_EXECUTE); assert (pkey > 0); /* perform mprotect */ assert(!sys_pkey_mprotect(region, pgsize, PROT_RWX, pkey)); do_execute(region); /* free pkey */ assert(!sys_pkey_free(pkey)); } int main(int argc, char **argv) { size_t pgsize, numinsns; unsigned int *region; int i; /* allocate memory region to protect */ pgsize = getpagesize(); region = memalign(pgsize, pgsize); assert(region != NULL); assert(!mprotect(region, pgsize, PROT_RWX)); /* fill page with NOPs with a BLR at the end */ numinsns = pgsize / sizeof(region[0]); for (i = 0; i < numinsns - 1; i++) region[i] = PPC_INST_NOP; region[i] = PPC_INST_BLR; do_protect(region); return EXIT_SUCCESS; } The fix is to only check the IAMR for an execute check, the AMR value is not relevant. Fixes: f2407ef3ba22 ("powerpc: helper to validate key-access permissions of a pte") Cc: stable@vger.kernel.org # v4.16+ Reported-by: Sandipan Das Signed-off-by: Aneesh Kumar K.V [mpe: Add detail to change log, tweak wording & formatting] Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20200712132047.1038594-1-aneesh.kumar@linux.ibm.com --- arch/powerpc/mm/book3s64/pkeys.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/arch/powerpc/mm/book3s64/pkeys.c b/arch/powerpc/mm/book3s64/pkeys.c index ca5fcb4bff32..d174106bab67 100644 --- a/arch/powerpc/mm/book3s64/pkeys.c +++ b/arch/powerpc/mm/book3s64/pkeys.c @@ -354,12 +354,14 @@ static bool pkey_access_permitted(int pkey, bool write, bool execute) u64 amr; pkey_shift = pkeyshift(pkey); - if (execute && !(read_iamr() & (IAMR_EX_BIT << pkey_shift))) - return true; + if (execute) + return !(read_iamr() & (IAMR_EX_BIT << pkey_shift)); - amr = read_amr(); /* Delay reading amr until absolutely needed */ - return ((!write && !(amr & (AMR_RD_BIT << pkey_shift))) || - (write && !(amr & (AMR_WR_BIT << pkey_shift)))); + amr = read_amr(); + if (write) + return !(amr & (AMR_WR_BIT << pkey_shift)); + + return !(amr & (AMR_RD_BIT << pkey_shift)); } bool arch_pte_access_permitted(u64 pte, bool write, bool execute) From b710d27bf72068b15b2f0305d825988183e2ff28 Mon Sep 17 00:00:00 2001 From: Satheesh Rajendran Date: Fri, 19 Jun 2020 12:31:13 +0530 Subject: [PATCH 2/4] powerpc/pseries/svm: Fix incorrect check for shared_lppaca_size Early secure guest boot hits the below crash while booting with vcpus numbers aligned with page boundary for PAGE size of 64k and LPPACA size of 1k i.e 64, 128 etc. Partition configured for 64 cpus. CPU maps initialized for 1 thread per core ------------[ cut here ]------------ kernel BUG at arch/powerpc/kernel/paca.c:89! Oops: Exception in kernel mode, sig: 5 [#1] LE PAGE_SIZE=64K MMU=Radix SMP NR_CPUS=2048 NUMA pSeries This is due to the BUG_ON() for shared_lppaca_total_size equal to shared_lppaca_size. Instead the code should only BUG_ON() if we have exceeded the total_size, which indicates we've overflowed the array. Fixes: bd104e6db6f0 ("powerpc/pseries/svm: Use shared memory for LPPACA structures") Cc: stable@vger.kernel.org # v5.4+ Signed-off-by: Satheesh Rajendran Reviewed-by: Laurent Dufour Reviewed-by: Thiago Jung Bauermann [mpe: Reword change log to clarify we're fixing not removing the check] Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20200619070113.16696-1-sathnaga@linux.vnet.ibm.com --- arch/powerpc/kernel/paca.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/powerpc/kernel/paca.c b/arch/powerpc/kernel/paca.c index 2168372b792d..74da65aacbc9 100644 --- a/arch/powerpc/kernel/paca.c +++ b/arch/powerpc/kernel/paca.c @@ -87,7 +87,7 @@ static void *__init alloc_shared_lppaca(unsigned long size, unsigned long align, * This is very early in boot, so no harm done if the kernel crashes at * this point. */ - BUG_ON(shared_lppaca_size >= shared_lppaca_total_size); + BUG_ON(shared_lppaca_size > shared_lppaca_total_size); return ptr; } From 6068e1a4427e88f5cc62f238d1baf94a8b824ef4 Mon Sep 17 00:00:00 2001 From: Haren Myneni Date: Fri, 10 Jul 2020 16:47:19 -0700 Subject: [PATCH 3/4] powerpc/vas: Report proper error code for address translation failure P9 DD2 NX workbook (Table 4-36) says DMA controller uses CC=5 internally for translation fault handling. NX reserves CC=250 for OS to notify user space when NX encounters address translation failure on the request buffer. Not an issue in earlier releases as NX does not get faults on kernel addresses. This patch defines CSB_CC_FAULT_ADDRESS(250) and updates CSB.CC with this proper error code for user space. Fixes: c96c4436aba4 ("powerpc/vas: Update CSB and notify process for fault CRBs") Signed-off-by: Haren Myneni [mpe: Added Fixes tag and fix typo in comment] Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/019fd53e7538c6f8f332d175df74b1815ef5aa8c.camel@linux.ibm.com --- Documentation/powerpc/vas-api.rst | 2 +- arch/powerpc/include/asm/icswx.h | 2 ++ arch/powerpc/platforms/powernv/vas-fault.c | 2 +- 3 files changed, 4 insertions(+), 2 deletions(-) diff --git a/Documentation/powerpc/vas-api.rst b/Documentation/powerpc/vas-api.rst index 1217c2f1595e..788dc8375a0e 100644 --- a/Documentation/powerpc/vas-api.rst +++ b/Documentation/powerpc/vas-api.rst @@ -213,7 +213,7 @@ request buffers are not in memory. The operating system handles the fault by updating CSB with the following data: csb.flags = CSB_V; - csb.cc = CSB_CC_TRANSLATION; + csb.cc = CSB_CC_FAULT_ADDRESS; csb.ce = CSB_CE_TERMINATION; csb.address = fault_address; diff --git a/arch/powerpc/include/asm/icswx.h b/arch/powerpc/include/asm/icswx.h index 965b1f39b2a5..b0c70a35fd0e 100644 --- a/arch/powerpc/include/asm/icswx.h +++ b/arch/powerpc/include/asm/icswx.h @@ -77,6 +77,8 @@ struct coprocessor_completion_block { #define CSB_CC_CHAIN (37) #define CSB_CC_SEQUENCE (38) #define CSB_CC_HW (39) +/* P9 DD2 NX Workbook 3.2 (Table 4-36): Address translation fault */ +#define CSB_CC_FAULT_ADDRESS (250) #define CSB_SIZE (0x10) #define CSB_ALIGN CSB_SIZE diff --git a/arch/powerpc/platforms/powernv/vas-fault.c b/arch/powerpc/platforms/powernv/vas-fault.c index 266a6ca5e15e..3d21fce254b7 100644 --- a/arch/powerpc/platforms/powernv/vas-fault.c +++ b/arch/powerpc/platforms/powernv/vas-fault.c @@ -79,7 +79,7 @@ static void update_csb(struct vas_window *window, csb_addr = (void __user *)be64_to_cpu(crb->csb_addr); memset(&csb, 0, sizeof(csb)); - csb.cc = CSB_CC_TRANSLATION; + csb.cc = CSB_CC_FAULT_ADDRESS; csb.ce = CSB_CE_TERMINATION; csb.cs = 0; csb.count = 0; From f0479c4bcbd92d1a457d4a43bcab79f29d11334a Mon Sep 17 00:00:00 2001 From: Haren Myneni Date: Fri, 10 Jul 2020 16:49:58 -0700 Subject: [PATCH 4/4] selftests/powerpc: Use proper error code to check fault address ERR_NX_TRANSLATION(CSB.CC=5) is for internal to VAS for fault handling and should not used by OS. ERR_NX_AT_FAULT(CSB.CC=250) is the proper error code should be reported by OS when NX encounters address translation failure. This patch uses CC=250 to determine the fault address when the request is not successful. Signed-off-by: Haren Myneni Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/0315251705baff94f678c33178491b5008723511.camel@linux.ibm.com --- tools/testing/selftests/powerpc/nx-gzip/gunz_test.c | 4 ++-- tools/testing/selftests/powerpc/nx-gzip/gzfht_test.c | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/tools/testing/selftests/powerpc/nx-gzip/gunz_test.c b/tools/testing/selftests/powerpc/nx-gzip/gunz_test.c index 6ee0fded0391..7c23d3dd7d6d 100644 --- a/tools/testing/selftests/powerpc/nx-gzip/gunz_test.c +++ b/tools/testing/selftests/powerpc/nx-gzip/gunz_test.c @@ -698,13 +698,13 @@ restart_nx: switch (cc) { - case ERR_NX_TRANSLATION: + case ERR_NX_AT_FAULT: /* We touched the pages ahead of time. In the most common case * we shouldn't be here. But may be some pages were paged out. * Kernel should have placed the faulting address to fsaddr. */ - NXPRT(fprintf(stderr, "ERR_NX_TRANSLATION %p\n", + NXPRT(fprintf(stderr, "ERR_NX_AT_FAULT %p\n", (void *)cmdp->crb.csb.fsaddr)); if (pgfault_retries == NX_MAX_FAULTS) { diff --git a/tools/testing/selftests/powerpc/nx-gzip/gzfht_test.c b/tools/testing/selftests/powerpc/nx-gzip/gzfht_test.c index 7496a83f9c9d..02dffb65de48 100644 --- a/tools/testing/selftests/powerpc/nx-gzip/gzfht_test.c +++ b/tools/testing/selftests/powerpc/nx-gzip/gzfht_test.c @@ -306,13 +306,13 @@ int compress_file(int argc, char **argv, void *handle) lzcounts, cmdp, handle); if (cc != ERR_NX_OK && cc != ERR_NX_TPBC_GT_SPBC && - cc != ERR_NX_TRANSLATION) { + cc != ERR_NX_AT_FAULT) { fprintf(stderr, "nx error: cc= %d\n", cc); exit(-1); } /* Page faults are handled by the user code */ - if (cc == ERR_NX_TRANSLATION) { + if (cc == ERR_NX_AT_FAULT) { NXPRT(fprintf(stderr, "page fault: cc= %d, ", cc)); NXPRT(fprintf(stderr, "try= %d, fsa= %08llx\n", fault_tries,