During IGT testing it has been shown that the specification
defined polling time of 1 us for FCLK_DONE, is sometimes not
enough. The issue is still reproducible while disabling
C-states through the PM QoS framework and also while disabling
preemtion. From this the most plausible explanation is that the
issue is due to a firmware flaw.
As a workaround, it is better to wait a little bit longer for
the FCLK_DONE to come around, than to leave with an DRM_ERROR
and having FCLK_DONE at a randome time after.
While spinning a list of igt tests prone to reproduce the issue
the FCLK_DONE poll failed at approximately 2% of the invocations
of the bdw_set_cdclk function. The longest poll time during this
testing was measured to ~7us. So, the suggested new poll time of
100us is on the safe side.
v2: Added more documentation about investigations done.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102243
Signed-off-by: Marta Lofstedt <marta.lofstedt@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170908132829.6312-1-marta.lofstedt@intel.com
If we know that we will completely fill a pagetable (i.e. we are
inserting a complete set of 512 pages), we can skip prefilling that PT
with scratch entries. If we have to abort the insertion prior to writing
the real entries, we will teardown the pagetable and remove it from the
page directory (so that we will restart the allocation next time).
We could do similar tricks for the PD and PDP, but the likelihood of a
single insertion covering the entire 512 entries diminishes, as do the
cycle savings. The saving are even greater (relatively) when we are
preallocating page tables for huge pages, as then we never need to fill
the page table.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170908181622.17791-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Looking at our virtual PCI device, we can see surprising Region 4 and Region 5.
00:10.0 VGA compatible controller: Intel Corporation Sky Lake Integrated Graphics (rev 06) (prog-if 00 [VGA controller])
....
Region 0: Memory at 140000000 (64-bit, non-prefetchable) [size=16M]
Region 2: Memory at 180000000 (64-bit, prefetchable) [size=1G]
Region 4: Memory at <ignored> (32-bit, non-prefetchable)
Region 5: Memory at <ignored> (32-bit, non-prefetchable)
Expansion ROM at febd6000 [disabled] [size=2K]
The fact is that we only implemented BAR0 and BAR2. Surprising Region 4 and
Region 5 are shown because we report their size as 0xffffffff. They should
report size 0 instead.
BTW, the physical GPU has a PIO BAR. GVTg hasn't implemented PIO access, so
we ignored this BAR for vGPU device.
v2: fix BAR size value calculation.
Link: https://bugzilla.redhat.com/show_bug.cgi?id=1458032
Signed-off-by: Changbin Du <changbin.du@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
fix the wrong return type and return error once the unknown
command is scanned.
v2:
- separate this error handle from healthy rating code. (Zhenyu)
Signed-off-by: fred gao <fred.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
When an error occurs in dispatch_workload, this patch is to do the
proper cleanup and rollback to the original states before the workload
is abandoned.
v2:
- split the mixed several error paths for better review. (Zhenyu)
v3:
- original PTR_ERR(cs) is good and code cleanup. (Zhenyu)
v4:
- reuse the existing i915_add_request for error handling. (Zhenyu)
v5:
- remove the duplicate error handling release_shadow_wa_ctx and
move the engine->context_unpin upper. (Zhenyu)
v6:
- keep the old label "out". (Zhenyu)
Signed-off-by: fred gao <fred.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
When it is failed in shadow_mm, the pin_count should rollback
to the original states before return.
v2:
- split the mixed several error paths for better review. (Zhenyu)
v3:
increase the pincount after shadow success. (Zhenyu)
Signed-off-by: fred gao <fred.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
refine the error handling for prepare_execlist_workload to restore to the
original states once error occurs.
only release the shadowed batch buffer and wa ctx when the workload is
completed successfully.
v2:
- split the mixed several error paths for better review. (Zhenyu)
v3:
- handle prepare batch buffer/wa ctx pin errors and
- emulate_schedule_in null issue. (Zhenyu)
v4:
- no need to handle emulate_schedule_in null issue. (Zhenyu)
v5:
- release the shadowed batch buffer and wa ctx only for the
successful workload. (Zhenyu)
v6:
- polish the return style. (Zhenyu)
Signed-off-by: fred gao <fred.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
When an error occurs after shadow_indirect_ctx, this patch is to do the
proper cleanup and rollback to the original states for shadowed indirect
context before the workload is abandoned.
v2:
- split the mixed several error paths for better review. (Zhenyu)
v3:
- no return check for clean up functions. (Changbin)
v4:
- expose and reuse the existing release_shadow_wa_ctx. (Zhenyu)
v5:
- move the release function to scheduler.c file. (Zhenyu)
v6:
- move error handling code of intel_gvt_scan_and_shadow_workload
to here. (Zhenyu)
Signed-off-by: fred gao <fred.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Currently i915 request structure and shadow ring buffer are allocated
before command scan, so it will have to restore to previous states once
any error happens afterwards in the long dispatch_workload path.
This patch is to introduce a reserved ring buffer created at the beginning
of vGPU initialization. Workload will be coped to this reserved buffer and
be scanned first, the i915 request and shadow ring buffer are only
allocated after the result of scan is successful.
To balance the memory usage and buffer alloc time, the coming bigger ring
buffer will be reallocated and kept until more bigger buffer is coming.
v2:
- use kmalloc for the smaller ring buffer, realloc if required. (Zhenyu)
v3:
- remove the dynamically allocated ring buffer. (Zhenyu)
v4:
- code style polish.
- kfree previous allocated buffer once kmalloc failed. (Zhenyu)
Signed-off-by: fred gao <fred.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
For vfio-pci, if the region support MMAP then it should support both
mmap and normal file access. The user-space is free to choose which is
being used. For qemu, we just need add 'x-no-mmap=on' for vfio-pci
option.
Currently GVTg only support MMAP for BAR2. So GVTg will not work when
user turn on x-no-mmap option.
This patch added file style access for BAR2, aka the GPU aperture. We
map the entire aperture partition of active vGPU to kernel space when
guest driver try to enable PCI Memory Space. Then we redirect the file
RW operation from kvmgt to this mapped area.
Link: https://bugzilla.redhat.com/show_bug.cgi?id=1458032
Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
For PCI, 64bit bar consumes two BAR registers, but this doesn't mean
both of two BAR are valid. Actually the second BAR is regarded as
reserved in this case. So we shouldn't emulate the second BAR.
Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
drm_pci_alloc() refuses to cooperate if the passed alignment exceeds the
object size. So round up the obj size to the next power of two as well
to make this actually work.
Obviously things work just fine as long as the size was a power of two
to begin with. However kms_cursor_crc doesn't always use power of two
sizes so we hit a failure when we try to allocate the phys memory.
Testcase: igt/kms_cursor_crc
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170907143203.13055-1-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
This patch adds IPC support. This patch also enables IPC in all supported
platforms based on has_ipc flag.
IPC (Isochronous Priority Control) is the hardware feature, which
dynamically controls the memory read priority of Display.
When IPC is enabled, plane read requests are sent at high priority until
filling above the transition watermark, then the requests are sent at
lower priority until dropping below the level 0 watermark.
The lower priority requests allow other memory clients to have better
memory access. When IPC is disabled, all plane read requests are sent at
high priority.
Changes since V1:
- Remove commandline parameter to disable ipc
- Address Paulo's comments
Changes since V2:
- Address review comments
- Set ipc_enabled flag
Changes since V3:
- move ipc_enabled flag assignment inside intel_ipc_enable function
Changes since V4:
- Re-enable IPC after suspend/resume
Changes since V5:
- Enable IPC for all gen >=9 except SKL
Changes since V6:
- fix commit msg
- after resume program IPC based on SW state.
Changes since V7:
- Modify IPC support check based on HAS_IPC macro (suggested by Chris)
Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-8-mahesh1.kumar@intel.com
GEN > 9 require transition WM to be programmed if IPC is enabled.
This patch calculates & enable transition WM for supported platforms.
If transition WM is enabled, Plane read requests are sent at high
priority until filling above the transition watermark, then the
requests are sent at lower priority until dropping below the level-0 WM.
The lower priority requests allow other memory clients to have better
memory access.
transition minimum is the minimum amount needed for trans_wm to work to
ensure the demote does not happen before enough data has been read to
meet the level 0 watermark requirements.
transition amount is configurable value. Higher values will
tend to cause longer periods of high priority reads followed by longer
periods of lower priority reads. Tuning to lower values will tend to
cause shorter periods of high and lower priority reads.
Keeping transition amount to 10 in this patch, as suggested by HW team.
Changes since V1:
- Address review comments from Maarten
Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-4-mahesh1.kumar@intel.com
