Check the AFU state whenever an API is called. The hypervisor may
issue a reset of the adapter when it detects a fault. When it happens,
it launches an error recovery which will either move the AFU to a
permanent failure state, or in the disabled state.
If the AFU is found to be disabled, detach all existing contexts from
it before issuing a AFU reset to re-enable it.
Before detaching contexts, notify any kernel driver through the EEH
callbacks of the AFU pci device.
Co-authored-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Christophe Lombard <clombard@linux.vnet.ibm.com>
Reviewed-by: Manoj Kumar <manoj@linux.vnet.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The new of.c file contains code to parse the device tree to find out
about cxl adapters and AFUs.
guest.c implements the guest-specific callbacks for the backend API.
The process element ID is not known until the context is attached, so
we have to separate the context ID assigned by the cxl driver from the
process element ID visible to the user applications. In bare-metal,
the 2 IDs match.
Co-authored-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Christophe Lombard <clombard@linux.vnet.ibm.com>
Reviewed-by: Manoj Kumar <manoj@linux.vnet.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
[mpe: Fix SMP=n build, fix PSERIES=n build, minor whitespace fixes]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Introduce sub-structures containing the bare-metal specific fields in
the structures describing the adapter (struct cxl) and AFU (struct
cxl_afu).
Update all their references.
Co-authored-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Christophe Lombard <clombard@linux.vnet.ibm.com>
Reviewed-by: Manoj Kumar <manoj@linux.vnet.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
A few functions are mostly common between bare-metal and guest and
just need minor tuning. To avoid crowding the backend API, introduce a
few 'if' based on the CPU being in HV mode.
Co-authored-by: Christophe Lombard <clombard@linux.vnet.ibm.com>
Signed-off-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Christophe Lombard <clombard@linux.vnet.ibm.com>
Reviewed-by: Manoj Kumar <manoj@linux.vnet.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The backend API (in cxl.h) lists some low-level functions whose
implementation is different on bare-metal and in a guest. Each
environment implements its own functions, and the common code uses
them through function pointers, defined in cxl_backend_ops
Co-authored-by: Christophe Lombard <clombard@linux.vnet.ibm.com>
Signed-off-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Christophe Lombard <clombard@linux.vnet.ibm.com>
Reviewed-by: Manoj Kumar <manoj@linux.vnet.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Presently when a user-space process issues CXL_IOCTL_START_WORK ioctl we
store the pid of the current task_struct and use it to get pointer to
the mm_struct of the process, while processing page or segment faults
from the capi card. However this causes issues when the thread that had
originally issued the start-work ioctl exits in which case the stored
pid is no more valid and the cxl driver is unable to handle faults as
the mm_struct corresponding to process is no more accessible.
This patch fixes this issue by using the mm_struct of the next alive
task in the thread group. This is done by iterating over all the tasks
in the thread group starting from thread group leader and calling
get_task_mm on each one of them. When a valid mm_struct is obtained the
pid of the associated task is stored in the context replacing the
exiting one for handling future faults.
The patch introduces a new function named get_mem_context that checks if
the current task pointed to by ctx->pid is dead? If yes it performs the
steps described above. Also a new variable cxl_context.glpid is
introduced which stores the pid of the thread group leader associated
with the context owning task.
Reported-by: Matthew R. Ochs <mrochs@linux.vnet.ibm.com>
Reported-by: Frank Haverkamp <HAVERKAM@de.ibm.com>
Suggested-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Vaibhav Jain <vaibhav@linux.vnet.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Reviewed-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Reviewed-by: Matthew R. Ochs <mrochs@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
An idr warning is reported when a context is release after the capi card
is unbound from the cxl driver via sysfs. Below are the steps to
reproduce:
1. Create multiple afu contexts in an user-space application using libcxl.
2. Unbind capi card from cxl using command of form
echo <capi-card-pci-addr> > /sys/bus/pci/drivers/cxl-pci/unbind
3. Exit/kill the application owning afu contexts.
After above steps a warning message is usually seen in the kernel logs
of the form "idr_remove called for id=<context-id> which is not
allocated."
This is caused by the function cxl_release_afu which destroys the
contexts_idr table. So when a context is release no entry for context pe
is found in the contexts_idr table and idr code prints this warning.
This patch fixes this issue by increasing & decreasing the ref-count on
the afu device when a context is initialized or when its freed
respectively. This prevents the afu from being released until all the
afu contexts have been released. The patch introduces two new functions
namely cxl_afu_get/put that manage the ref-count on the afu device.
Also the patch removes code inside cxl_dev_context_init that increases ref
on the afu device as its guaranteed to be alive during this function.
Reported-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Vaibhav Jain <vaibhav@linux.vnet.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
At present, ctx->irq_bitmap is freed in afu_release_irqs(), which is called
from afu_release() via cxl_context_detach().
Move the freeing of ctx->irq_bitmap from afu_release_irqs() to
reclaim_ctx() (called through cxl_context_free()) so it's freed when
releasing a context via the kernel API (cxl_release_context()) or the
userspace API (afu_release()).
Reported-by: Matthew R. Ochs <mrochs@linux.vnet.ibm.com>
Fixes: 6f7f0b3df6 ("cxl: Add AFU virtual PHB and kernel API")
Signed-off-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Reviewed-by: Matthew R. Ochs <mrochs@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The cxl user api uses the address_space associated with the file when we
need to force unmap all cxl mmap regions (e.g. on eeh, driver detach,
etc). Currently, contexts allocated through the kernel api do not do
this and instead skip the mmap invalidation, potentially allowing them
to poke at the hardware after such an event, which may cause all sorts
of trouble.
This patch allocates an address_space for cxl contexts allocated through
the kernel api so that the same invalidate path will for these contexts
as well. We don't use the anonymous inode's address_space, as doing so
could invalidate any mmaps of completely unrelated drivers using
anonymous file descriptors.
This patch also introduces a kernelapi flag, so we know when freeing the
context if the address_space was allocated by us and needs to be freed.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
userspace programs using cxl currently have to use two strategies for
dealing with MMIO errors simultaneously. They have to check every read
for a return of all Fs in case the adapter has gone away and the kernel
has not yet noticed, and they have to deal with SIGBUS in case the
kernel has already noticed, invalidated the mapping and marked the
context as failed.
In order to simplify things, this patch adds an alternative approach
where the kernel will return a page filled with Fs instead of delivering
a SIGBUS. This allows userspace to only need to deal with one of these
two error paths, and is intended for use in libraries that use cxl
transparently and may not be able to safely install a signal handler.
This approach will only work if certain constraints are met. Namely, if
the application is both reading and writing to an address in the problem
state area it cannot assume that a non-FF read is OK, as it may just be
reading out a value it has previously written. Further - since only one
page is used per context a write to a given offset would be visible when
reading the same offset from a different page in the mapping (this only
applies within a single context, not between contexts).
An application could deal with this by e.g. making sure it also reads
from a read-only offset after any reads to a read/write offset.
Due to these constraints, this functionality must be explicitly
requested by userspace when starting the context by passing in the
CXL_START_WORK_ERR_FF flag.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
If the PCI channel has gone down, don't attempt to poke the hardware.
We need to guard every time cxl_whatever_(read|write) is called. This
is because a call to those functions will dereference an offset into an
mmio register, and the mmio mappings get invalidated in the EEH
teardown.
Check in the read/write functions in the header.
We give them the same semantics as usual PCI operations:
- a write to a channel that is down is ignored.
- a read from a channel that is down returns all fs.
Also, we try to access the MMIO space of a vPHB device as part of the
PCI disable path. Because that's a read that bypasses most of our usual
checks, we handle it explicitly.
As far as user visible warnings go:
- Check link state in file ops, return -EIO if down.
- Be reasonably quiet if there's an error in a teardown path,
or when we already know the hardware is going down.
- Throw a big WARN if someone tries to start a CXL operation
while the card is down. This gives a useful stacktrace for
debugging whatever is doing that.
Signed-off-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
It was discovered that if a process mmaped their problem state area they
were able to access one page more than expected, potentially allowing
them to access the problem state area of an unrelated process.
This was due to a simple off by one error in the mmap fault handler
introduced in 0712dc7e73 ("cxl: Fix issues
when unmapping contexts"), which is fixed in this patch.
Cc: stable@vger.kernel.org
Fixes: 0712dc7e73 ("cxl: Fix issues when unmapping contexts")
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This patch makes the mmap call fail outright if the requested region is
larger than the problem state area assigned to the context so the error
is reported immediately rather than waiting for an attempt to access an
address out of bounds.
Although we never expect users to map more than the assigned problem
state area and are not aware of anyone doing this (other than for
testing), this does have the potential to break users if someone has
used a larger range regardless. I'm submitting it for consideration, but
if this change is not considered acceptable the previous patch is
sufficient to prevent access out of bounds without breaking anyone.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This reworks contexts lifetimes a bit to enable the kernel API where we may
want to reuse contexts. Here we will want to start and stop contexts without
freeing them.
Start context does the get pid & ctx so stop context will need to do the puts.
Here we move put pid & ctx to the detach context path which will become part of
the stop context path.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Rework __detach_context() and cxl_context_detach() so we can reuse them in the
kernel API.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Add cookie parameter to afu_release_irqs() so that we can pass in a different
cookie than the context structure. This will be useful for other kernel
drivers that want to call this but get their own cookie back in the interrupt
handler.
Update all existing call sites.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The afu fd release path was identified as a significant bottleneck in
the overall performance of cxl. While an optimal AFU design would
minimise the need to close & reopen the AFU fd, it is not always
practical to avoid.
The bottleneck seems to be down to the call to synchronize_rcu(), which
will block until every other thread is guaranteed to be out of an RCU
critical section. Replace it with call_rcu() to free the context
structures later so we can return to the application sooner.
This reduces the time spent in the fd release path from 13356 usec to
13.3 usec - about a 100x speed up.
Reported-by: Fei K Chen <uchen@cn.ibm.com>
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
An issue was introduced with "cxl: Unmap MMIO regions when detaching a
context" (b123429e6a) where closing a
context normally could also unmap the problem state area of other
contexts currently using the AFU.
It was also discovered that after a context's MMIO space had been
unmapped it would read 0s when accessing it, whereas the expected
behaviour was for the access to fail altogether.
In order to address these issues, this patch does two things:
- Forced mmap unmapping is only done when we are forcefully detaching
all contexts, and not in the normal detach path. Since the normal
context close path is tied to the file release any mmaps must have
already been released so we don't need to worry in that case.
- The mmap path now uses a vm_operations_struct with a fault handler.
The fault handler ensures that the context is in started state,
otherwise it fails the access attempt with a SIGBUS.
Fixes: b123429e6a ("cxl: Unmap MMIO regions when detaching a context")
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
If we need to force detach a context (e.g. due to EEH or simply force
unbinding the driver) we should prevent the userspace contexts from
being able to access the Problem State Area MMIO region further, which
they may have mapped with mmap().
This patch unmaps any mapped MMIO regions when detaching a userspace
context.
Cc: stable@vger.kernel.org
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
We had a known sleep while atomic bug if a CXL device was forcefully
unbound while it was in use. This could occur as a result of EEH, or
manually induced with something like this while the device was in use:
echo 0000:01:00.0 > /sys/bus/pci/drivers/cxl-pci/unbind
The issue was that in this code path we iterated over each context and
forcefully detached it with the contexts_lock spin lock held, however
the detach also needed to take the spu_mutex, and call schedule.
This patch changes the contexts_lock to a mutex so that we are not in
atomic context while doing the detach, thereby avoiding the sleep while
atomic.
Also delete the related TODO comment, which suggested an alternate
solution which turned out to not be workable.
Cc: stable@vger.kernel.org
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This is the core of the cxl driver.
It adds support for using cxl cards in the powernv environment only (ie POWER8
bare metal). It allows access to cxl accelerators by userspace using the
/dev/cxl/afuM.N char devices.
The kernel driver has no knowledge of the function implemented by the
accelerator. It provides services to userspace via the /dev/cxl/afuM.N
devices. When a program opens this device and runs the start work IOCTL, the
accelerator will have coherent access to that processes memory using the same
virtual addresses. That process may mmap the device to access any MMIO space
the accelerator provides. Also, reads on the device will allow interrupts to
be received. These services are further documented in a later patch in
Documentation/powerpc/cxl.txt.
Documentation of the cxl hardware architecture and userspace API is provided in
subsequent patches.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Christophe Lombard
Frederic Barrat
Vaibhav Jain
Andrew Donnellan
Ian Munsie
Daniel Axtens
Michael Neuling