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 4A415CD8CA8 for ; Fri, 12 Jun 2026 11:08:51 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id F3B0310F450; Fri, 12 Jun 2026 11:08:50 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="X1wSdG3e"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [198.175.65.21]) by gabe.freedesktop.org (Postfix) with ESMTPS id 5E8E210E9E2 for ; Fri, 12 Jun 2026 11:06:58 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1781262419; x=1812798419; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=IGL9T/QxifaaQmRKft7m+/WrsnCS+PO2BBr5Q9G8cHM=; b=X1wSdG3ejKGKILiK40YHTqN204O8p9Y+SPUOxU5DOtZHwybIpqeBzqAN wcR2DrjVv5JA4Ha/vjmxnwCZ2NZtMcxwI7rmLXxkV3qN9HWgsIXRVAg8u yMIpZa1pguZBylfSIa4Uj7HtibLwcATxJerDysC+4I4kv7yb/212vsJKO 43ywxNIIhFAUp9XGxR7Fr1dnzGOU4cHaHrrhNDKMphFnjqwSdgaQrQrPA H9pLhmT+OLtLoDgBpIdA5hDxyu3yxd1tN4tSb1fDK8PQVaPrQZKXaMu4v /i8VnuTkE0EfTVw6NwGZF6tdNdOJf5ljOAkYtQUvpimy6jFDUBqRWUU/1 Q==; X-CSE-ConnectionGUID: pTAWISoySIa4A2ZdLeL4xg== X-CSE-MsgGUID: ec6wtLg7ShWDJQhFV8oQbQ== X-IronPort-AV: E=McAfee;i="6800,10657,11813"; a="81997604" X-IronPort-AV: E=Sophos;i="6.24,200,1774335600"; d="scan'208";a="81997604" Received: from orviesa003.jf.intel.com ([10.64.159.143]) by orvoesa113.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 12 Jun 2026 04:06:59 -0700 X-CSE-ConnectionGUID: FUsb8aUkRWC6Qvnl5JMO3g== X-CSE-MsgGUID: weEqUlTIS+WI8O6gGdLqkw== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.24,200,1774335600"; d="scan'208";a="250717735" Received: from slindbla-desk.ger.corp.intel.com (HELO fedora) ([10.245.245.68]) by ORVIESA003-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 12 Jun 2026 04:06:56 -0700 From: =?UTF-8?q?Thomas=20Hellstr=C3=B6m?= To: igt-dev@lists.freedesktop.org Cc: =?UTF-8?q?Thomas=20Hellstr=C3=B6m?= , Matthew Brost , Maarten Lankhorst , Michal Mrozek , John Falkowski , Rodrigo Vivi , Lahtinen Joonas Subject: [PATCH i-g-t 1/4] lib/xe: add xe_vm_restart ioctl helper Date: Fri, 12 Jun 2026 13:06:16 +0200 Message-ID: <20260612110619.103198-2-thomas.hellstrom@linux.intel.com> X-Mailer: git-send-email 2.54.0 In-Reply-To: <20260612110619.103198-1-thomas.hellstrom@linux.intel.com> References: <20260612110619.103198-1-thomas.hellstrom@linux.intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-BeenThere: igt-dev@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Development mailing list for IGT GPU Tools List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: igt-dev-bounces@lists.freedesktop.org Sender: "igt-dev" Add DRM_XE_VM_RESTART (ioctl 0x10) and struct drm_xe_vm_restart to the Xe DRM UAPI header, taken from the xe_event kernel branch. Add DRM_XE_VM_CREATE_FLAG_RESTARTABLE (bit 4) to allow VMs to opt in to the restart mechanism. Add __xe_vm_restart() (failable, with bounded EAGAIN retry) and xe_vm_restart() (asserting wrapper) to lib/xe/xe_ioctl. Both take an optional CLOCK_MONOTONIC timestamp_ns which is forwarded to the IOCTL so the driver can log event-to-restart latency. Assisted-by: GitHub Copilot:claude-sonnet-4.6 --- include/drm-uapi/xe_drm.h | 40 +++++++++++++++++++++++++++++++++++ lib/xe/xe_ioctl.c | 44 +++++++++++++++++++++++++++++++++++++++ lib/xe/xe_ioctl.h | 2 ++ 3 files changed, 86 insertions(+) diff --git a/include/drm-uapi/xe_drm.h b/include/drm-uapi/xe_drm.h index 5a96a7910..43b65b1d9 100644 --- a/include/drm-uapi/xe_drm.h +++ b/include/drm-uapi/xe_drm.h @@ -84,6 +84,7 @@ extern "C" { * - &DRM_IOCTL_XE_MADVISE * - &DRM_IOCTL_XE_VM_QUERY_MEM_RANGE_ATTRS * - &DRM_IOCTL_XE_VM_GET_PROPERTY + * - &DRM_IOCTL_XE_VM_RESTART */ /* @@ -109,6 +110,7 @@ extern "C" { #define DRM_XE_VM_QUERY_MEM_RANGE_ATTRS 0x0d #define DRM_XE_EXEC_QUEUE_SET_PROPERTY 0x0e #define DRM_XE_VM_GET_PROPERTY 0x0f +#define DRM_XE_VM_RESTART 0x10 /* Must be kept compact -- no holes */ @@ -128,6 +130,7 @@ extern "C" { #define DRM_IOCTL_XE_VM_QUERY_MEM_RANGE_ATTRS DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_VM_QUERY_MEM_RANGE_ATTRS, struct drm_xe_vm_query_mem_range_attr) #define DRM_IOCTL_XE_EXEC_QUEUE_SET_PROPERTY DRM_IOW(DRM_COMMAND_BASE + DRM_XE_EXEC_QUEUE_SET_PROPERTY, struct drm_xe_exec_queue_set_property) #define DRM_IOCTL_XE_VM_GET_PROPERTY DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_VM_GET_PROPERTY, struct drm_xe_vm_get_property) +#define DRM_IOCTL_XE_VM_RESTART DRM_IOW(DRM_COMMAND_BASE + DRM_XE_VM_RESTART, struct drm_xe_vm_restart) /** * DOC: Xe IOCTL Extensions @@ -991,6 +994,7 @@ struct drm_xe_vm_create { #define DRM_XE_VM_CREATE_FLAG_LR_MODE (1 << 1) #define DRM_XE_VM_CREATE_FLAG_FAULT_MODE (1 << 2) #define DRM_XE_VM_CREATE_FLAG_NO_VM_OVERCOMMIT (1 << 3) +#define DRM_XE_VM_CREATE_FLAG_RESTARTABLE (1 << 4) /** @flags: Flags */ __u32 flags; @@ -2609,6 +2613,42 @@ enum drm_xe_ras_error_component { [DRM_XE_RAS_ERR_COMP_SOC_INTERNAL] = "soc-internal" \ } +/** + * DOC: DRM_XE_VM_RESTART + * + * Synchronously restart a VM by running its preempt-rebind worker in the + * calling context. The VM must be in preempt-fence mode (i.e. it must have + * been created with exec queues that use preempt fences). + * + * On return the rebind attempt has completed or a retriable error was + * encountered. Any non-retriable error is surfaced through the event + * mechanism if the caller has subscribed to %DRM_XE_EVENT_MASK_VM_ERR. + * The IOCTL may return -EAGAIN if userptr memory needs to be repinned; + * callers should retry in that case. + */ + +/** + * struct drm_xe_vm_restart - restart a VM's preempt-rebind worker + * + * Used with %DRM_IOCTL_XE_VM_RESTART. + */ +struct drm_xe_vm_restart { + /** @vm_id: ID of the VM to restart */ + __u32 vm_id; + /** @pad: reserved, must be zero */ + __u32 pad; + /** + * @timestamp_ns: optional CLOCK_MONOTONIC timestamp in nanoseconds. + * When non-zero, the driver logs the delay between this timestamp and + * the point the rebind completes, which can be used to measure the + * response latency from event delivery to VM restart. Pass zero to + * disable the logging. + */ + __u64 timestamp_ns; + /** @reserved: reserved, must be zero */ + __u64 reserved; +}; + #if defined(__cplusplus) } #endif diff --git a/lib/xe/xe_ioctl.c b/lib/xe/xe_ioctl.c index c8ed99182..f102fe34e 100644 --- a/lib/xe/xe_ioctl.c +++ b/lib/xe/xe_ioctl.c @@ -337,6 +337,50 @@ void xe_vm_get_property(int fd, uint32_t vm, struct drm_xe_vm_get_property *quer igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_VM_GET_PROPERTY, query), 0); } +/** + * __xe_vm_restart() - restart a VM's preempt-rebind worker (failable) + * @fd: open Xe DRM device file descriptor + * @vm: VM id to restart + * @timestamp_ns: CLOCK_MONOTONIC timestamp from the triggering event, or 0 + * + * Calls %DRM_IOCTL_XE_VM_RESTART, retrying up to 10 times on -EAGAIN as + * required when userptr memory needs repinning. + * + * Return: 0 on success, negative errno on failure. + */ +int __xe_vm_restart(int fd, uint32_t vm, uint64_t timestamp_ns) +{ + struct drm_xe_vm_restart restart = { + .vm_id = vm, + .timestamp_ns = timestamp_ns, + }; + int err, tries = 10; + + do { + err = igt_ioctl(fd, DRM_IOCTL_XE_VM_RESTART, &restart); + if (err) { + err = -errno; + igt_assume(err); + errno = 0; + } + } while (err == -EAGAIN && --tries > 0); + + return err; +} + +/** + * xe_vm_restart() - restart a VM's preempt-rebind worker + * @fd: open Xe DRM device file descriptor + * @vm: VM id to restart + * @timestamp_ns: CLOCK_MONOTONIC timestamp from the triggering event, or 0 + * + * Calls __xe_vm_restart() and asserts success. + */ +void xe_vm_restart(int fd, uint32_t vm, uint64_t timestamp_ns) +{ + igt_assert_eq(__xe_vm_restart(fd, vm, timestamp_ns), 0); +} + void xe_vm_destroy(int fd, uint32_t vm) { struct drm_xe_vm_destroy destroy = { diff --git a/lib/xe/xe_ioctl.h b/lib/xe/xe_ioctl.h index bf40fb6bd..95f2ec3d8 100644 --- a/lib/xe/xe_ioctl.h +++ b/lib/xe/xe_ioctl.h @@ -66,6 +66,8 @@ void xe_vm_unbind_all_async(int fd, uint32_t vm, uint32_t exec_queue, uint32_t bo, struct drm_xe_sync *sync, uint32_t num_syncs); void xe_vm_get_property(int fd, uint32_t vm, struct drm_xe_vm_get_property *query); +int __xe_vm_restart(int fd, uint32_t vm, uint64_t timestamp_ns); +void xe_vm_restart(int fd, uint32_t vm, uint64_t timestamp_ns); void xe_vm_destroy(int fd, uint32_t vm); uint32_t __xe_bo_create(int fd, uint32_t vm, uint64_t size, uint32_t placement, uint32_t flags, void *ext, uint32_t *handle); -- 2.54.0