[Intel-xe] [PATCH v2 1/1] drm/xe: Notify Userspace when engine/gt reset fails.
Ghimiray, Himal Prasad
himal.prasad.ghimiray at intel.com
Tue Jun 27 18:02:00 UTC 2023
> -----Original Message-----
> From: Ghimiray, Himal Prasad
> Sent: 16 May 2023 17:12
> To: Rodrigo Vivi <rodrigo.vivi at kernel.org>; Nilawar, Badal
> <badal.nilawar at intel.com>
> Cc: intel-xe at lists.freedesktop.org; Harrison, John C
> <john.c.harrison at intel.com>; Argenziano, Antonio
> <antonio.argenziano at intel.com>; Iddamsetty, Aravind
> <Aravind.Iddamsetty at intel.com>
> Subject: RE: [Intel-xe] [PATCH v2 1/1] drm/xe: Notify Userspace when
> engine/gt reset fails.
>
>
>
> > -----Original Message-----
> > From: Rodrigo Vivi <rodrigo.vivi at kernel.org>
> > Sent: 16 May 2023 06:18
> > To: Nilawar, Badal <badal.nilawar at intel.com>
> > Cc: Ghimiray, Himal Prasad <himal.prasad.ghimiray at intel.com>; intel-
> > xe at lists.freedesktop.org; Harrison, John C
> > <john.c.harrison at intel.com>; Argenziano, Antonio
> > <antonio.argenziano at intel.com>
> > Subject: Re: [Intel-xe] [PATCH v2 1/1] drm/xe: Notify Userspace when
> > engine/gt reset fails.
> >
> > On Mon, May 15, 2023 at 09:46:47PM +0530, Nilawar, Badal wrote:
> > >
> > >
> > > On 11-05-2023 17:13, Ghimiray, Himal Prasad wrote:
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Nilawar, Badal <badal.nilawar at intel.com>
> > > > > Sent: 11 May 2023 10:07
> > > > > To: Ghimiray, Himal Prasad <himal.prasad.ghimiray at intel.com>;
> > > > > intel- xe at lists.freedesktop.org
> > > > > Cc: Upadhyay, Tejas <tejas.upadhyay at intel.com>
> > > > > Subject: Re: [Intel-xe] [PATCH v2 1/1] drm/xe: Notify Userspace
> > > > > when engine/gt reset fails.
> > > > >
> > > > > Hi Himal,
> > > > >
> > > > > On 09-05-2023 13:55, Himal Prasad Ghimiray wrote:
> > > > > > Send uevent in case of engine reset or gt reset failure.
> > > > > > This intimation can be used by userspace monitoring tool to do
> > > > > > the device level reset/reboot when GT reset fails. udevadm can
> > > > > > be used to monitor the uevents.
> > > > > >
> > > > > > v2:
> > > > > > -Add NULL check for xe_gt_hw_engine return(Aravind) -Arrange
> > > > > > variables in Christmas tree order(Tejas) -Check
> > > > > > GUC_GSC_OTHER_CLASS(Tejas)
> > > > > >
> > > > > > Cc: Aravind Iddamsetty <aravind.iddamsetty at intel.com>
> > > > > > Cc: Tejas Upadhyay <tejas.upadhyay at intel.com>
> > > > > > Signed-off-by: Himal Prasad Ghimiray
> > > > > > <himal.prasad.ghimiray at intel.com>
> > > > > > ---
> > > > > > drivers/gpu/drm/xe/xe_gt.c | 18 ++++++++++++++++++
> > > > > > drivers/gpu/drm/xe/xe_guc.h | 19 +++++++++++++++++++
> > > > > > drivers/gpu/drm/xe/xe_guc_submit.c | 23
> > +++++++++++++++++++++++
> > > > > > include/uapi/drm/xe_drm.h | 8 ++++++++
> > > > > > 4 files changed, 68 insertions(+)
> > > > > >
> > > > > > diff --git a/drivers/gpu/drm/xe/xe_gt.c
> > > > > > b/drivers/gpu/drm/xe/xe_gt.c index 3afca3dd9657..2c3ffa1db74e
> > > > > > 100644
> > > > > > --- a/drivers/gpu/drm/xe/xe_gt.c
> > > > > > +++ b/drivers/gpu/drm/xe/xe_gt.c
> > > > > > @@ -8,6 +8,7 @@
> > > > > > #include <linux/minmax.h>
> > > > > >
> > > > > > #include <drm/drm_managed.h>
> > > > > > +#include <drm/xe_drm.h>
> > > > > >
> > > > > > #include "regs/xe_gt_regs.h"
> > > > > > #include "xe_bb.h"
> > > > > > @@ -590,6 +591,20 @@ static int do_gt_restart(struct xe_gt *gt)
> > > > > > return 0;
> > > > > > }
> > > > > >
> > > > > > +static void xe_uevent_gt_reset_failure(struct xe_device *xe,
> > > > > > +u8 id)
> > {
> > > > > > + char *reset_event[5];
> > > > > > +
> > > > > > + reset_event[0] = XE_RESET_FAILED_UEVENT "=1";
> > > > > > + reset_event[1] = "RESET_ENABLED=1";
> > > > > > + reset_event[2] = "RESET_UNIT=gt";
> > > > > > + reset_event[3] = kasprintf(GFP_KERNEL, "RESET_ID=%d",
> > id);
> > > > > > + reset_event[4] = NULL;
> > > > > > + kobject_uevent_env(&xe->drm.primary->kdev->kobj,
> > KOBJ_CHANGE,
> > > > > > +reset_event);
> > > > > > +
> > > > > > + kfree(reset_event[3]);
> > > > > > +}
> > > > > > +
> > > > > > static int gt_reset(struct xe_gt *gt)
> > > > > > {
> > > > > > struct xe_device *xe = gt_to_xe(gt); @@ -639,6 +654,9 @@
> > > > > > static int gt_reset(struct xe_gt *gt)
> > > > > > xe_device_mem_access_put(gt_to_xe(gt));
> > > > > > drm_err(&xe->drm, "GT reset failed, err=%d\n", err);
> > > > > >
> > > > > > + /* Notify userspace about gt reset failure */
> > > > > > + xe_uevent_gt_reset_failure(xe, gt->info.id);
> > > > > > +
> > > > > > return err;
> > > > > > }
> > > > > >
> > > > > > diff --git a/drivers/gpu/drm/xe/xe_guc.h
> > > > > > b/drivers/gpu/drm/xe/xe_guc.h index
> 74a74051f354..a3637eeaaa37
> > > > > > 100644
> > > > > > --- a/drivers/gpu/drm/xe/xe_guc.h
> > > > > > +++ b/drivers/gpu/drm/xe/xe_guc.h
> > > > > > @@ -56,4 +56,23 @@ static inline u16
> > > > > > xe_engine_class_to_guc_class(enum
> > > > > xe_engine_class class)
> > > > > > }
> > > > > > }
> > > > > >
> > > > > > +static inline u16 xe_guc_class_to_engine_class(u8 guc_class) {
> > > > > > + switch (guc_class) {
> > > > > > + case GUC_RENDER_CLASS:
> > > > > > + return XE_ENGINE_CLASS_RENDER;
> > > > > > + case GUC_VIDEO_CLASS:
> > > > > > + return XE_ENGINE_CLASS_VIDEO_DECODE;
> > > > > > + case GUC_VIDEOENHANCE_CLASS:
> > > > > > + return XE_ENGINE_CLASS_VIDEO_ENHANCE;
> > > > > > + case GUC_BLITTER_CLASS:
> > > > > > + return XE_ENGINE_CLASS_COPY;
> > > > > > + case GUC_COMPUTE_CLASS:
> > > > > > + return XE_ENGINE_CLASS_COMPUTE;
> > > > > > + case GUC_GSC_OTHER_CLASS:
> > > > > > + default:
> > > > > > + XE_WARN_ON(guc_class);
> > > > > > + return -1;
> > > > > > + }
> > > > > > +}
> > > > > > #endif
> > > > > > diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c
> > > > > > b/drivers/gpu/drm/xe/xe_guc_submit.c
> > > > > > index e857013070b9..d068af0ca7df 100644
> > > > > > --- a/drivers/gpu/drm/xe/xe_guc_submit.c
> > > > > > +++ b/drivers/gpu/drm/xe/xe_guc_submit.c
> > > > > > @@ -12,6 +12,7 @@
> > > > > > #include <linux/dma-fence-array.h>
> > > > > >
> > > > > > #include <drm/drm_managed.h>
> > > > > > +#include <drm/xe_drm.h>
> > > > > >
> > > > > > #include "regs/xe_lrc_layout.h"
> > > > > > #include "xe_device.h"
> > > > > > @@ -1589,10 +1590,26 @@ int
> > > > > xe_guc_engine_memory_cat_error_handler(struct xe_guc *guc, u32
> > > > > *msg,
> > > > > > return 0;
> > > > > > }
> > > > > >
> > > > > > +static void xe_uevent_engine_reset_failure(struct xe_device
> > > > > > +*xe, const char *name) {
> > > > > > + char *reset_event[5];
> > > > > > +
> > > > > > + reset_event[0] = XE_RESET_FAILED_UEVENT "=1";
> > > > > > + reset_event[1] = "RESET_ENABLED=1";
> > > > > > + reset_event[2] = "RESET_UNIT=engine";
> > > > > > + reset_event[3] = kasprintf(GFP_KERNEL, "RESET_ID=%s",
> > name);
> > > > > > + reset_event[4] = NULL;
> > > > > > + kobject_uevent_env(&xe->drm.primary->kdev->kobj,
> > KOBJ_CHANGE,
> > > > > > +reset_event);
> > > > > > +
> > > > > > + kfree(reset_event[3]);
> > > > > > +}
> > > > > > +
> > > > > > int xe_guc_engine_reset_failure_handler(struct xe_guc *guc,
> > > > > > u32 *msg,
> > > > > u32 len)
> > > > > > {
> > > > > > struct xe_device *xe = guc_to_xe(guc);
> > > > > > + struct xe_hw_engine *hwe;
> > > > > > u8 guc_class, instance;
> > > > > > + u16 engine_class;
> > > > > > u32 reason;
> > > > > >
> > > > > > if (unlikely(len != 3)) {
> > > > > > @@ -1608,6 +1625,12 @@ int
> > > > > > xe_guc_engine_reset_failure_handler(struct
> > > > > xe_guc *guc, u32 *msg, u32 len)
> > > > > > drm_err(&xe->drm, "GuC engine reset request failed on
> > %d:%d
> > > > > because 0x%08X",
> > > > > > guc_class, instance, reason);
> > > > > >
> > > > > > + /* Notify userspace about engine reset failure */
> > > > > > + engine_class = xe_guc_class_to_engine_class(guc_class);
> > > > > > + hwe = xe_gt_hw_engine(guc_to_gt(guc), engine_class,
> > instance,
> > > > > > +false);The The instance here is considered as physical
> > > > > > +instance of hwe but while
> > > > > submitting contexts kmd passes logical instance to guc. With
> > > > > that I assume guc might be passing logical instance or for this
> > > > > notification it is passing physical instance?
> > > > AFAIK for reset failure notification the instance passed by guc in
> > > > g2h
> > payload is physical instance .
> > > > @Harrison, John C Can you please confirm ?
> > >
> > > Spoke to @Antonio, he confirmed that to notify engine reset failure
> > > guc is reporting physical instance to host.
> > >
> > > Reviewed-by: Badal Nilawar <badal.nilawar at intel.com>
> >
> > I don't believe this is what we want here still. At least not from the
> > sysman perspective.
> >
> > After guc reset fails, we will try to reset the GT ourselves and if we
> > succeed we don't need to request user space interference.
> >
> > That should only happen if we also fail our gt_reset after the
> > guc_reset failed.
> >
> > That would be the similar point on when i915 gets wedged?
> > We probably also need the wedge point here before we can issue this
> > notification?
>
> Hi Rodrigo,
> You are right, In case of engine reset failure the driver tries to do the gt reset
> and if gt reset fails there are two notifications one for engine failure and
> another for gt reset failure. The changes to send uevent in case of gt failure
> are added in xe_uevent_gt_reset_failure.
>
> Level0 UMD does sends the RESET_REQUIRED uevent in case of only gt reset
> failure.
> Engine reset failure notification is not used for this purpose, it is to provide
> information that engine reset was required and it failed.
>
> >
> > Anyway, we also need IGT for the uevent itself.
> The IGT for uevent listener will be helpful only if we can inject errors in such
> a way that they lead to engine/gt reset failure.
> I am unable to find a mechanism to do so. Can you provide some input on
> the same ?
> BR
> Himal
As it was discussed there is no mechanism to ensure engine/gt reset failure from HW perspective to validate these changes.
The recommendation was to come up with a debugfs which can fake gt reset failure and use same with IGT.
The series covers the debugfs entry https://patchwork.freedesktop.org/series/119784/
and https://patchwork.freedesktop.org/series/119390/ covers IGT.
Requesting for further comments and conclusion on this patch.
> >
> > Thanks,
> > Rodrigo.
> >
> > >
> > > Regards,
> > > Badal
> > > >
> > > > >
> > > > > Regards,
> > > > > Badal
> > > > > > + if (hwe)
> > > > > > + xe_uevent_engine_reset_failure(xe, hwe->name);
> > > > > > +
> > > > > > xe_gt_reset_async(guc_to_gt(guc));
> > > > > >
> > > > > > return 0;
> > > > > > diff --git a/include/uapi/drm/xe_drm.h
> > > > > > b/include/uapi/drm/xe_drm.h index b0b80aae3ee8..79ef5947c172
> > > > > > 100644
> > > > > > --- a/include/uapi/drm/xe_drm.h
> > > > > > +++ b/include/uapi/drm/xe_drm.h
> > > > > > @@ -36,6 +36,14 @@ extern "C" {
> > > > > > * subject to backwards-compatibility constraints.
> > > > > > */
> > > > > >
> > > > > > +/*
> > > > > > + * Uevents generated by xe on it's device node.
> > > > > > + *
> > > > > > + * XE_RESET_FAILED_UEVENT - Event is generated when attempt
> > > > > > +to reset
> > > > > engine
> > > > > > + * or gt fails. The value supplied with the event is always 1.
> > > > > > + */
> > > > > > +#define XE_RESET_FAILED_UEVENT "RESET_FAILED"
> > > > > > +
> > > > > > /**
> > > > > > * struct xe_user_extension - Base class for defining a
> > > > > > chain of
> > extensions
> > > > > > *
More information about the Intel-xe
mailing list