From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id 6A87CC7EE22 for ; Tue, 9 May 2023 17:25:43 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 2A31710E3B4; Tue, 9 May 2023 17:25:43 +0000 (UTC) Received: from dfw.source.kernel.org (dfw.source.kernel.org [139.178.84.217]) by gabe.freedesktop.org (Postfix) with ESMTPS id 8B4AD10E3B4 for ; Tue, 9 May 2023 17:25:42 +0000 (UTC) Received: from smtp.kernel.org (relay.kernel.org [52.25.139.140]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by dfw.source.kernel.org (Postfix) with ESMTPS id 459EA62B5D; Tue, 9 May 2023 17:25:41 +0000 (UTC) Received: from rdvivi-mobl4 (unknown [192.55.54.50]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.kernel.org (Postfix) with ESMTPSA id 8AD75C433EF; Tue, 9 May 2023 17:25:38 +0000 (UTC) Date: Tue, 9 May 2023 13:25:36 -0400 From: Rodrigo Vivi To: "Ghimiray, Himal Prasad" Message-ID: References: <20230508134823.585186-1-himal.prasad.ghimiray@intel.com> <22d0494f-0db8-01b2-7611-8d43b2d79e2b@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Subject: Re: [Intel-xe] [RFC 1/1] drm/xe: Notify Userspace when engine/gt reset fails. X-BeenThere: intel-xe@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Intel Xe graphics driver List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: "intel-xe@lists.freedesktop.org" , "Vivi, Rodrigo" Errors-To: intel-xe-bounces@lists.freedesktop.org Sender: "Intel-xe" On Tue, May 09, 2023 at 06:28:07AM +0000, Ghimiray, Himal Prasad wrote: > > > > -----Original Message----- > > From: Iddamsetty, Aravind > > Sent: 09 May 2023 10:42 > > To: Ghimiray, Himal Prasad ; intel- > > xe@lists.freedesktop.org; Vivi, Rodrigo > > Subject: Re: [RFC 1/1] drm/xe: Notify Userspace when engine/gt reset fails. > > > > > > > > On 08-05-2023 19:18, Himal Prasad Ghimiray wrote: > > > Send uevent in case of engine reset or gt reset failure. > > > L0 sysman would like to generate device reset needed event when gt > > > reset fails > > > https://one-api.gitlab-pages.devtools.intel.com/level_zero/sysman/api. > > > html#_CPPv441ZES_EVENT_TYPE_FLAG_DEVICE_RESET_REQUIRED > > This could rather be an intimation to userpace monitoring tool to do the > > device level reset/reboot when GT reset fails. udevadm can be used to > > monitor the uevents. > Hmm. Makes sense. > > > > @rodrigo, your thoughts? It would be awesome if this is an indication that user space needs to call FLR/SBR... so we would do this when we are unable to reset ("wedged") but let's get some confirmation on the true meaning before moving on... > > > > > > > > > > Cc: Aravind Iddamsetty > > > Signed-off-by: Himal Prasad Ghimiray > > > --- > > > drivers/gpu/drm/xe/xe_gt.c | 18 ++++++++++++++++++ > > > drivers/gpu/drm/xe/xe_guc.h | 18 ++++++++++++++++++ > > > drivers/gpu/drm/xe/xe_guc_submit.c | 23 +++++++++++++++++++++++ > > > include/uapi/drm/xe_drm.h | 8 ++++++++ > > > 4 files changed, 67 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 > > > > > > #include > > > +#include > > > > > > #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..845c57b95562 100644 > > > --- a/drivers/gpu/drm/xe/xe_guc.h > > > +++ b/drivers/gpu/drm/xe/xe_guc.h > > > @@ -56,4 +56,22 @@ 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; > > > + 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..663908573d5c 100644 > > > --- a/drivers/gpu/drm/xe/xe_guc_submit.c > > > +++ b/drivers/gpu/drm/xe/xe_guc_submit.c > > > @@ -12,6 +12,7 @@ > > > #include > > > > > > #include > > > +#include > > > > > > #include "regs/xe_lrc_layout.h" > > > #include "xe_device.h" > > > @@ -1589,9 +1590,25 @@ 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; > > > + u16 engine_class; > > > u8 guc_class, instance; > > > u32 reason; > > > > > > @@ -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); > > > > > > + engine_class = xe_guc_class_to_engine_class(guc_class); > > > + hwe = xe_gt_hw_engine(guc_to_gt(guc), engine_class, instance, > > > +false); > > > > this can return NULL, so a check is needed here. > Will address in next patch. > > > > Thanks, > > Aravind. > > > + > > > + /* Notify userspace about engine reset failure */ > > > + 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 > > > *