Although radeon card fence and wait for gpu to finish processing current batch rings,
there is still a corner case that radeon lockup work queue may not be fully flushed,
and meanwhile the radeon_suspend_kms() function has called pci_set_power_state() to
put device in D3hot state.
Per PCI spec rev 4.0 on 5.3.1.4.1 D3hot State.
> Configuration and Message requests are the only TLPs accepted by a Function in
> the D3hot state. All other received Requests must be handled as Unsupported Requests,
> and all received Completions may optionally be handled as Unexpected Completions.
This issue will happen in following logs:
Unable to handle kernel paging request at virtual address 00008800e0008010
CPU 0 kworker/0:3(131): Oops 0
pc = [<ffffffff811bea5c>] ra = [<ffffffff81240844>] ps = 0000 Tainted: G W
pc is at si_gpu_check_soft_reset+0x3c/0x240
ra is at si_dma_is_lockup+0x34/0xd0
v0 = 0000000000000000 t0 = fff08800e0008010 t1 = 0000000000010000
t2 = 0000000000008010 t3 = fff00007e3c00000 t4 = fff00007e3c00258
t5 = 000000000000ffff t6 = 0000000000000001 t7 = fff00007ef078000
s0 = fff00007e3c016e8 s1 = fff00007e3c00000 s2 = fff00007e3c00018
s3 = fff00007e3c00000 s4 = fff00007fff59d80 s5 = 0000000000000000
s6 = fff00007ef07bd98
a0 = fff00007e3c00000 a1 = fff00007e3c016e8 a2 = 0000000000000008
a3 = 0000000000000001 a4 = 8f5c28f5c28f5c29 a5 = ffffffff810f4338
t8 = 0000000000000275 t9 = ffffffff809b66f8 t10 = ff6769c5d964b800
t11= 000000000000b886 pv = ffffffff811bea20 at = 0000000000000000
gp = ffffffff81d89690 sp = 00000000aa814126
Disabling lock debugging due to kernel taint
Trace:
[<ffffffff81240844>] si_dma_is_lockup+0x34/0xd0
[<ffffffff81119610>] radeon_fence_check_lockup+0xd0/0x290
[<ffffffff80977010>] process_one_work+0x280/0x550
[<ffffffff80977350>] worker_thread+0x70/0x7c0
[<ffffffff80977410>] worker_thread+0x130/0x7c0
[<ffffffff80982040>] kthread+0x200/0x210
[<ffffffff809772e0>] worker_thread+0x0/0x7c0
[<ffffffff80981f8c>] kthread+0x14c/0x210
[<ffffffff80911658>] ret_from_kernel_thread+0x18/0x20
[<ffffffff80981e40>] kthread+0x0/0x210
Code: ad3e0008 43f0074a ad7e0018 ad9e0020 8c3001e8 40230101
<88210000> 4821ed21
So force lockup work queue flush to fix this problem.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Zhenneng Li <lizhenneng@kylinos.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
dml_wrapper* files were removed in commit 01b537eeb0
("drm/amd/display: Remove unused code"), as they are not used anywhere.
However, the header file wasn't removed, so remove the header as well.
Signed-off-by: Magali Lemes <magalilemes00@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The file amdgpu_dm_plane.c missed the header amdgpu_dm_plane.h, which
resulted on the following warning:
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:1046:5:
warning: no previous prototype for 'fill_dc_scaling_info'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:1222:6:
warning: no previous prototype for 'handle_cursor_update'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:152:6:
warning: no previous prototype for 'modifier_has_dcc'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:1576:5:
warning: no previous prototype for 'amdgpu_dm_plane_init'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:157:10:
warning: no previous prototype for 'modifier_gfx9_swizzle_mode'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:752:5:
warning: no previous prototype for 'fill_plane_buffer_attributes'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:83:31:
warning: no previous prototype for 'amd_get_format_info'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:88:6:
warning: no previous prototype for 'fill_blending_from_plane_state'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:992:5:
warning: no previous prototype for 'dm_plane_helper_check_state'
[-Wmissing-prototypes]
Therefore, include the missing header on the file and turn global functions
that are not used outside of the file into static functions.
Fixes: 5d945cbcd4 ("drm/amd/display: Create a file dedicated to planes")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Maíra Canal <mairacanal@riseup.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Addresses the following warning:
drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn30/display_mode_vba_30.c:3596:6: error: stack frame size (2092) exceeds limit (2048) in 'dml30_ModeSupportAndSystemConfigurationFull' [-Werror,-Wframe-larger-than]
void dml30_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_lib)
^
UseMinimumDCFCLK() is eating away at
dml30_ModeSupportAndSystemConfigurationFull()'s stack space, so use a
pointer to struct vba_vars_st instead of passing lots of large arrays
as parameters by value.
Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
V1:
The amdgpu_xgmi_remove_device function will send unload command
to psp through psp ring to terminate xgmi, but psp ring has been
destroyed in psp_hw_fini.
V2:
1. Change the commit title.
2. Restore amdgpu_xgmi_remove_device to its original calling location.
Move psp_xgmi_terminate call from amdgpu_xgmi_remove_device to
psp_hw_fini.
Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We are adding two new callbacks to ttm resource manager
function to handle intersection and compatibility of
placement and resources.
v2: move the amdgpu and ttm_range_manager changes to
separate patches (Christian)
v3: rename "intersect" to "intersects" (Matthew)
v4: move !place check to the !res if and return false
in ttm_resource_compatible() function (Christian)
v5: move bits of code from patch number 6 to avoid
temporary driver breakup (Christian)
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220820073304.178444-1-Arunpravin.PaneerSelvam@amd.com
Commit 20f85ef89d ("drm/i915/backlight: use unique backlight device
names") added support for multiple backlight devices on dual panel
systems, but did so with error handling on -EEXIST from
backlight_device_register(). Unfortunately, that triggered a warning in
dmesg all the way down from sysfs_add_file_mode_ns() and
sysfs_warn_dup().
Instead of optimistically always attempting to register with the default
name ("intel_backlight", which we have to retain for backward
compatibility), check if a backlight device with the name exists first,
and, if so, use the card and connector based name.
v2: reworked on top of the patch commit 20f85ef89d
("drm/i915/backlight: use unique backlight device names")
v3: fixed the ref count leak(Jani N)
Fixes: 20f85ef89d ("drm/i915/backlight: use unique backlight device names")
Signed-off-by: Arun R Murthy <arun.r.murthy@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220808035750.3111046-1-arun.r.murthy@intel.com
Drivers' .remove and .shutdown callbacks are executed on different code
paths. The former is called when a device is removed from the bus, while
the latter is called at system shutdown time to quiesce the device.
This means that some overlap exists between the two, because both have to
take care of properly shutting down the hardware. But currently the logic
used in these two callbacks isn't consistent in msm drivers, which could
lead to kernel panic.
For example, on .remove the component is deleted and its .unbind callback
leads to the hardware being shutdown but only if the DRM device has been
marked as registered.
That check doesn't exist in the .shutdown logic and this can lead to the
driver calling drm_atomic_helper_shutdown() for a DRM device that hasn't
been properly initialized.
A situation like this can happen if drivers for expected sub-devices fail
to probe, since the .bind callback will never be executed. If that is the
case, drm_atomic_helper_shutdown() will attempt to take mutexes that are
only initialized if drm_mode_config_init() is called during a device bind.
This bug was attempted to be fixed in commit 623f279c77 ("drm/msm: fix
shutdown hook in case GPU components failed to bind"), but unfortunately
it still happens in some cases as the one mentioned above, i.e:
systemd-shutdown[1]: Powering off.
kvm: exiting hardware virtualization
platform wifi-firmware.0: Removing from iommu group 12
platform video-firmware.0: Removing from iommu group 10
------------[ cut here ]------------
WARNING: CPU: 6 PID: 1 at drivers/gpu/drm/drm_modeset_lock.c:317 drm_modeset_lock_all_ctx+0x3c4/0x3d0
...
Hardware name: Google CoachZ (rev3+) (DT)
pstate: a0400009 (NzCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : drm_modeset_lock_all_ctx+0x3c4/0x3d0
lr : drm_modeset_lock_all_ctx+0x48/0x3d0
sp : ffff80000805bb80
x29: ffff80000805bb80 x28: ffff327c00128000 x27: 0000000000000000
x26: 0000000000000000 x25: 0000000000000001 x24: ffffc95d820ec030
x23: ffff327c00bbd090 x22: ffffc95d8215eca0 x21: ffff327c039c5800
x20: ffff327c039c5988 x19: ffff80000805bbe8 x18: 0000000000000034
x17: 000000040044ffff x16: ffffc95d80cac920 x15: 0000000000000000
x14: 0000000000000315 x13: 0000000000000315 x12: 0000000000000000
x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
x8 : ffff80000805bc28 x7 : 0000000000000000 x6 : 0000000000000000
x5 : 0000000000000000 x4 : 0000000000000000 x3 : 0000000000000000
x2 : ffff327c00128000 x1 : 0000000000000000 x0 : ffff327c039c59b0
Call trace:
drm_modeset_lock_all_ctx+0x3c4/0x3d0
drm_atomic_helper_shutdown+0x70/0x134
msm_drv_shutdown+0x30/0x40
platform_shutdown+0x28/0x40
device_shutdown+0x148/0x350
kernel_power_off+0x38/0x80
__do_sys_reboot+0x288/0x2c0
__arm64_sys_reboot+0x28/0x34
invoke_syscall+0x48/0x114
el0_svc_common.constprop.0+0x44/0xec
do_el0_svc+0x2c/0xc0
el0_svc+0x2c/0x84
el0t_64_sync_handler+0x11c/0x150
el0t_64_sync+0x18c/0x190
---[ end trace 0000000000000000 ]---
Unable to handle kernel NULL pointer dereference at virtual address 0000000000000018
Mem abort info:
ESR = 0x0000000096000004
EC = 0x25: DABT (current EL), IL = 32 bits
SET = 0, FnV = 0
EA = 0, S1PTW = 0
FSC = 0x04: level 0 translation fault
Data abort info:
ISV = 0, ISS = 0x00000004
CM = 0, WnR = 0
user pgtable: 4k pages, 48-bit VAs, pgdp=000000010eab1000
[0000000000000018] pgd=0000000000000000, p4d=0000000000000000
Internal error: Oops: 96000004 [#1] PREEMPT SMP
...
Hardware name: Google CoachZ (rev3+) (DT)
pstate: a0400009 (NzCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : ww_mutex_lock+0x28/0x32c
lr : drm_modeset_lock_all_ctx+0x1b0/0x3d0
sp : ffff80000805bb50
x29: ffff80000805bb50 x28: ffff327c00128000 x27: 0000000000000000
x26: 0000000000000000 x25: 0000000000000001 x24: 0000000000000018
x23: ffff80000805bc10 x22: ffff327c039c5ad8 x21: ffff327c039c5800
x20: ffff80000805bbe8 x19: 0000000000000018 x18: 0000000000000034
x17: 000000040044ffff x16: ffffc95d80cac920 x15: 0000000000000000
x14: 0000000000000315 x13: 0000000000000315 x12: 0000000000000000
x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
x8 : ffff80000805bc28 x7 : 0000000000000000 x6 : 0000000000000000
x5 : 0000000000000000 x4 : 0000000000000000 x3 : 0000000000000000
x2 : ffff327c00128000 x1 : 0000000000000000 x0 : 0000000000000018
Call trace:
ww_mutex_lock+0x28/0x32c
drm_modeset_lock_all_ctx+0x1b0/0x3d0
drm_atomic_helper_shutdown+0x70/0x134
msm_drv_shutdown+0x30/0x40
platform_shutdown+0x28/0x40
device_shutdown+0x148/0x350
kernel_power_off+0x38/0x80
__do_sys_reboot+0x288/0x2c0
__arm64_sys_reboot+0x28/0x34
invoke_syscall+0x48/0x114
el0_svc_common.constprop.0+0x44/0xec
do_el0_svc+0x2c/0xc0
el0_svc+0x2c/0x84
el0t_64_sync_handler+0x11c/0x150
el0t_64_sync+0x18c/0x190
Code: aa0103f4 d503201f d2800001 aa0103e3 (c8e37c02)
---[ end trace 0000000000000000 ]---
Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
Kernel Offset: 0x495d77c00000 from 0xffff800008000000
PHYS_OFFSET: 0xffffcd8500000000
CPU features: 0x800,00c2a015,19801c82
Memory Limit: none
---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ]---
Fixes: 9d5cbf5fe4 ("drm/msm: add shutdown support for display platform_driver")
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220816134612.916527-1-javierm@redhat.com
Although radeon card fence and wait for gpu to finish processing current batch rings,
there is still a corner case that radeon lockup work queue may not be fully flushed,
and meanwhile the radeon_suspend_kms() function has called pci_set_power_state() to
put device in D3hot state.
Per PCI spec rev 4.0 on 5.3.1.4.1 D3hot State.
> Configuration and Message requests are the only TLPs accepted by a Function in
> the D3hot state. All other received Requests must be handled as Unsupported Requests,
> and all received Completions may optionally be handled as Unexpected Completions.
This issue will happen in following logs:
Unable to handle kernel paging request at virtual address 00008800e0008010
CPU 0 kworker/0:3(131): Oops 0
pc = [<ffffffff811bea5c>] ra = [<ffffffff81240844>] ps = 0000 Tainted: G W
pc is at si_gpu_check_soft_reset+0x3c/0x240
ra is at si_dma_is_lockup+0x34/0xd0
v0 = 0000000000000000 t0 = fff08800e0008010 t1 = 0000000000010000
t2 = 0000000000008010 t3 = fff00007e3c00000 t4 = fff00007e3c00258
t5 = 000000000000ffff t6 = 0000000000000001 t7 = fff00007ef078000
s0 = fff00007e3c016e8 s1 = fff00007e3c00000 s2 = fff00007e3c00018
s3 = fff00007e3c00000 s4 = fff00007fff59d80 s5 = 0000000000000000
s6 = fff00007ef07bd98
a0 = fff00007e3c00000 a1 = fff00007e3c016e8 a2 = 0000000000000008
a3 = 0000000000000001 a4 = 8f5c28f5c28f5c29 a5 = ffffffff810f4338
t8 = 0000000000000275 t9 = ffffffff809b66f8 t10 = ff6769c5d964b800
t11= 000000000000b886 pv = ffffffff811bea20 at = 0000000000000000
gp = ffffffff81d89690 sp = 00000000aa814126
Disabling lock debugging due to kernel taint
Trace:
[<ffffffff81240844>] si_dma_is_lockup+0x34/0xd0
[<ffffffff81119610>] radeon_fence_check_lockup+0xd0/0x290
[<ffffffff80977010>] process_one_work+0x280/0x550
[<ffffffff80977350>] worker_thread+0x70/0x7c0
[<ffffffff80977410>] worker_thread+0x130/0x7c0
[<ffffffff80982040>] kthread+0x200/0x210
[<ffffffff809772e0>] worker_thread+0x0/0x7c0
[<ffffffff80981f8c>] kthread+0x14c/0x210
[<ffffffff80911658>] ret_from_kernel_thread+0x18/0x20
[<ffffffff80981e40>] kthread+0x0/0x210
Code: ad3e0008 43f0074a ad7e0018 ad9e0020 8c3001e8 40230101
<88210000> 4821ed21
So force lockup work queue flush to fix this problem.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Zhenneng Li <lizhenneng@kylinos.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The file amdgpu_dm_plane.c missed the header amdgpu_dm_plane.h, which
resulted on the following warning:
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:1046:5:
warning: no previous prototype for 'fill_dc_scaling_info'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:1222:6:
warning: no previous prototype for 'handle_cursor_update'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:152:6:
warning: no previous prototype for 'modifier_has_dcc'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:1576:5:
warning: no previous prototype for 'amdgpu_dm_plane_init'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:157:10:
warning: no previous prototype for 'modifier_gfx9_swizzle_mode'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:752:5:
warning: no previous prototype for 'fill_plane_buffer_attributes'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:83:31:
warning: no previous prototype for 'amd_get_format_info'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:88:6:
warning: no previous prototype for 'fill_blending_from_plane_state'
[-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_plane.c:992:5:
warning: no previous prototype for 'dm_plane_helper_check_state'
[-Wmissing-prototypes]
Therefore, include the missing header on the file and turn global functions
that are not used outside of the file into static functions.
Fixes: 5d945cbcd4 ("drm/amd/display: Create a file dedicated to planes")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Maíra Canal <mairacanal@riseup.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
V1:
The amdgpu_xgmi_remove_device function will send unload command
to psp through psp ring to terminate xgmi, but psp ring has been
destroyed in psp_hw_fini.
V2:
1. Change the commit title.
2. Restore amdgpu_xgmi_remove_device to its original calling location.
Move psp_xgmi_terminate call from amdgpu_xgmi_remove_device to
psp_hw_fini.
Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The docs explicitly say the drm_gem_object_release function already calls this,
and this does not appear to be a prerequisite for the call to
etnaviv_gem_ops.release.
Signed-off-by: T.J. Mercier <tjmercier@google.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Track the pid per submit, so we can print the name and cmdline of
the task which submitted the batch that caused the gpu to hang.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
MTL introduces a second FBC engine. The two FBC engines can
operate entirely independently, FBC A serving pipe A and
FBC B serving pipe B.
The one place where things might go a bit wrong is the CFB
allocation from stolen. We might have to consider some change
to the allocation strategy to have a better chance of both
engines being able to allocate its CFB. Maybe FBC A should
allocate bottom up and FBC B top down, or something? For the
moment the allocation strategy is DRM_MM_INSERT_BEST for both.
Cc: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220817122624.213889-1-jani.nikula@intel.com
Add a delay, configurable via debugfs (default 34ms), to disable
scheduling of a context after the pin count goes to zero. Disable
scheduling is a costly operation as it requires synchronizing with
the GuC. So the idea is that a delay allows the user to resubmit
something before doing this operation. This delay is only done if
the context isn't closed and less than a given threshold
(default is 3/4) of the guc_ids are in use.
As temporary WA disable this feature for the selftests. Selftests are
very timing sensitive and any change in timing can cause failure. A
follow up patch will fixup the selftests to understand this delay.
Alan Previn: Matt Brost first introduced this series back in Oct 2021.
However no real world workload with measured performance impact was
available to prove the intended results. Today, this series is being
republished in response to a real world workload that benefited greatly
from it along with measured performance improvement.
Workload description: 36 containers were created on a DG2 device where
each container was performing a combination of 720p 3d game rendering
and 30fps video encoding. The workload density was configured in a way
that guaranteed each container to ALWAYS be able to render and
encode no less than 30fps with a predefined maximum render + encode
latency time. That means the totality of all 36 containers and their
workloads were not saturating the engines to their max (in order to
maintain just enough headrooom to meet the min fps and max latencies
of incoming container submissions).
Problem statement: It was observed that the CPU core processing the i915
soft IRQ work was experiencing severe load. Using tracelogs and an
instrumentation patch to count specific i915 IRQ events, it was confirmed
that the majority of the CPU cycles were caused by the
gen11_other_irq_handler() -> guc_irq_handler() code path. The vast
majority of the cycles was determined to be processing a specific G2H
IRQ: i.e. INTEL_GUC_ACTION_SCHED_CONTEXT_MODE_DONE. These IRQs are sent
by GuC in response to i915 KMD sending H2G requests:
INTEL_GUC_ACTION_SCHED_CONTEXT_MODE_SET. Those H2G requests are sent
whenever a context goes idle so that we can unpin the context from GuC.
The high CPU utilization % symptom was limiting density scaling.
Root Cause Analysis: Because the incoming execution buffers were spread
across 36 different containers (each with multiple contexts) but the
system in totality was NOT saturated to the max, it was assumed that each
context was constantly idling between submissions. This was causing
a thrashing of unpinning contexts from GuC at one moment, followed quickly
by repinning them due to incoming workload the very next moment. These
event-pairs were being triggered across multiple contexts per container,
across all containers at the rate of > 30 times per sec per context.
Metrics: When running this workload without this patch, we measured an
average of ~69K INTEL_GUC_ACTION_SCHED_CONTEXT_MODE_DONE events every 10
seconds or ~10 million times over ~25+ mins. With this patch, the count
reduced to ~480 every 10 seconds or about ~28K over ~10 mins. The
improvement observed is ~99% for the average counts per 10 seconds.
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Alan Previn <alan.previn.teres.alexis@intel.com>
Reviewed-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220817020511.2180747-3-alan.previn.teres.alexis@intel.com
The current code tries to handle the case where CONFIG_PM isn't selected
by first calling our runtime_resume implementation and then properly
report the power state to the runtime_pm core.
This allows to have a functionning device even if pm_runtime_get_*
functions are nops.
However, the device power state if CONFIG_PM is enabled is
RPM_SUSPENDED, and thus our vc4_hdmi_write() and vc4_hdmi_read() calls
in the runtime_pm hooks will now report a warning since the device might
not be properly powered.
Even more so, we need CONFIG_PM enabled since the previous RaspberryPi
have a power domain that needs to be powered up for the HDMI controller
to be usable.
The previous patch has created a dependency on CONFIG_PM, now we can
just assume it's there and only call pm_runtime_resume_and_get() to make
sure our device is powered in bind.
Link: https://lore.kernel.org/r/20220629123510.1915022-39-maxime@cerno.tech
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
(cherry picked from commit 53565c28e6)
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
