Konstantin observed that when descriptors are submitted, the descriptor is
added to the pending list after the submission. This creates a race window
with the slight possibility that the descriptor can complete before it
gets added to the pending list and this window would cause the completion
handler to miss processing the descriptor.
To address the issue, the addition of the descriptor to the pending list
must be done before it gets submitted to the hardware. However, submitting
to swq with ENQCMDS instruction can cause a failure with the condition of
either wq is full or wq is not "active".
With the descriptor allocation being the gate to the wq capacity, it is not
possible to hit a retry with ENQCMDS submission to the swq. The only
possible failure can happen is when wq is no longer "active" due to hw
error and therefore we are moving towards taking down the portal. Given
this is a rare condition and there's no longer concern over I/O
performance, the driver can walk the completion lists in order to retrieve
and abort the descriptor.
The error path will set the descriptor to aborted status. It will take the
work list lock to prevent further processing of worklist. It will do a
delete_all on the pending llist to retrieve all descriptors on the pending
llist. The delete_all action does not require a lock. It will walk through
the acquired llist to find the aborted descriptor while add all remaining
descriptors to the work list since it holds the lock. If it does not find
the aborted descriptor on the llist, it will walk through the work
list. And if it still does not find the descriptor, then it means the
interrupt handler has removed the desc from the llist but is pending on
the work list lock and will process it once the error path releases the
lock.
Fixes: eb15e7154f ("dmaengine: idxd: add interrupt handle request and release support")
Reported-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/162628855747.360485.10101925573082466530.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add the code needed in the main IDXD driver to interface with the IDXD
perfmon implementation.
[ Based on work originally by Jing Lin. ]
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Link: https://lore.kernel.org/r/a5564a5583911565d31c2af9234218c5166c4b2c.1619276133.git.zanussi@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Remove interrupt masking and just let the hard irq handler keep
firing for new events. This is less of a performance impact vs
the MMIO readback inside the pci_msi_{mask,unmas}_irq(). Especially
with a loaded system those flushes can be stuck behind large amounts
of MMIO writes to flush. When guest kernel is running on top of VFIO
mdev, mask/unmask causes a vmexit each time and is not desirable.
Suggested-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/161894523436.3210025.1834640110556139277.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The char device setup and cleanup has device lifetime issues regarding when
parts are initialized and cleaned up. The initialization of struct device is
done incorrectly. device_initialize() needs to be called on the 'struct
device' and then additional changes can be added. The ->release() function
needs to be setup via device_type before dev_set_name() to allow proper
cleanup. The change re-parents the cdev under the wq->conf_dev to get
natural reference inheritance. No known dependency on the old device path exists.
Reported-by: Jason Gunthorpe <jgg@nvidia.com>
Fixes: 42d279f913 ("dmaengine: idxd: add char driver to expose submission portal to userland")
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/161852987721.2203940.1478218825576630810.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Remove devm_* allocation and fix wq->conf_dev 'struct device' lifetime.
Address issues flagged by CONFIG_DEBUG_KOBJECT_RELEASE. Add release
functions in order to free the allocated memory for the wq context at
device destruction time.
Reported-by: Jason Gunthorpe <jgg@nvidia.com>
Fixes: bfe1d56091 ("dmaengine: idxd: Init and probe for Intel data accelerators")
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/161852985907.2203940.6840120734115043753.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Current code blindly writes over the SWERR and the OVERFLOW bits. Write
back the bits actually read instead so the driver avoids clobbering the
OVERFLOW bit that comes after the register is read.
Fixes: bfe1d56091 ("dmaengine: idxd: Init and probe for Intel data accelerators")
Reported-by: Sanjay Kumar <sanjay.k.kumar@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/161352082229.3511254.1002151220537623503.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Nikhil reported the misc interrupt handler can sometimes miss handling
the command interrupt when an error interrupt happens near the same time.
Have the irq handling thread continue to process the misc interrupts until
all interrupts are processed. This is a low usage interrupt and is not
expected to handle high volume traffic. Therefore there is no concern of
this thread running for a long time.
Fixes: 0d5c10b4c8 ("dmaengine: idxd: add work queue drain support")
Reported-by: Nikhil Rao <nikhil.rao@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/161074755329.2183844.13295528344116907983.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add code to "complete" a descriptor when the descriptor or its completion
address hit a fault error when SVA mode is being used. This error can be
triggered due to bad programming by the user. A lock is introduced in order
to protect the descriptor completion lists since the fault handler will run
from the system work queue after being scheduled in the interrupt handler.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/160382008092.3911367.12766483427643278985.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Switch driver to use MSIX mask and unmask instead of the ignore bit.
When ignore bit is cleared, we must issue an MMIO read to ensure writes
have all arrived and check and process any additional completions. The
ignore bit does not queue up any pending MSIX interrupts. The mask bit
however does. Use API call from interrupt subsystem to mask MSIX
interrupt since the hardware does not have convenient mask bit register.
Suggested-by: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/159319517621.70410.11816465052708900506.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add wq drain support. When a wq is being released, it needs to wait for
all in-flight operation to complete. A device control function
idxd_wq_drain() has been added to facilitate this. A wq drain call
is added to the char dev on release to make sure all user operations are
complete. A wq drain is also added before the wq is being disabled.
A drain command can take an unpredictable period of time. Interrupt support
for device commands is added to allow waiting on the command to
finish. If a previous command is in progress, the new submitter can block
until the current command is finished before proceeding. The interrupt
based submission will submit the command and then wait until a command
completion interrupt happens to complete. All commands are moved to the
interrupt based command submission except for the device reset during
probe, which will be polled.
Fixes: 42d279f913 ("dmaengine: idxd: add char driver to expose submission portal to userland")
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/159319502515.69593.13451647706946040301.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Fix unmasking of misc interrupt handler when completing normal. It exits
early and skips the unmasking with the current implementation. Fix to
unmask interrupt when exiting normally.
Fixes: bfe1d56091 ("dmaengine: idxd: Init and probe for Intel data accelerators")
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/159311256528.855.11527922406329728512.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The current implementation may miss completions after we unmask the
interrupt. In order to make sure we process all competions, we need to:
1. Do an MMIO read from the device as a barrier to ensure that all PCI
writes for completions have arrived.
2. Check for any additional completions that we missed.
Fixes: 8f47d1a5e5 ("dmaengine: idxd: connect idxd to dmaengine subsystem")
Reported-by: Sanjay Kumar <sanjay.k.kumar@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/158834641769.35613.1341160109892008587.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Create a char device region that will allow acquisition of user portals in
order to allow applications to submit DMA operations. A char device will be
created per work queue that gets exposed. The workqueue type "user"
is used to mark a work queue for user char device. For example if the
workqueue 0 of DSA device 0 is marked for char device, then a device node
of /dev/dsa/wq0.0 will be created.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/157965026985.73301.976523230037106742.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add plumbing for dmaengine subsystem connection. The driver register a DMA
device per DSA device. The channels are dynamically registered when a
workqueue is configured to be "kernel:dmanegine" type.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/157965026376.73301.13867988830650740445.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The idxd driver introduces the Intel Data Stream Accelerator [1] that will
be available on future Intel Xeon CPUs. One of the kernel access
point for the driver is through the dmaengine subsystem. It will initially
provide the DMA copy service to the kernel.
Some of the main functionality introduced with this accelerator
are: shared virtual memory (SVM) support, and descriptor submission using
Intel CPU instructions movdir64b and enqcmds. There will be additional
accelerator devices that share the same driver with variations to
capabilities.
This commit introduces the probe and initialization component of the
driver.
[1]: https://software.intel.com/en-us/download/intel-data-streaming-accelerator-preliminary-architecture-specification
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/157965023991.73301.6186843973135311580.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Dave Jiang