From cb7a3f949aa4804ed8e0553d0196a4ce048ff7b8 Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Fri, 1 Aug 2025 16:28:21 +0200 Subject: [PATCH] drm/xe/pf: Make sure PF is ready to configure VFs MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit The PF driver might be resumed just to configure VFs, but since it is doing some asynchronous GuC reconfigurations after fresh reset, we should wait until all pending works are completed. This is especially important in case of LMEM provisioning, since we also need to update the LMTT and send invalidation requests to all GuCs, which are expected to be already in the VGT mode. Fixes: 68ae022278a1 ("drm/xe/pf: Force GuC virtualization mode") Signed-off-by: Michal Wajdeczko Reviewed-by: Piotr Piórkowski Cc: Jonathan Cavitt Link: https://lore.kernel.org/r/20250801142822.180530-3-michal.wajdeczko@intel.com (cherry picked from commit c6c86441c465ea440dfb5039f1c26e629a6fd64c) Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/xe/xe_gt_sriov_pf.c | 25 +++++++++++++++++++ drivers/gpu/drm/xe/xe_gt_sriov_pf.h | 1 + drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c | 4 ++- drivers/gpu/drm/xe/xe_pci_sriov.c | 7 +++++- drivers/gpu/drm/xe/xe_sriov_pf.c | 27 +++++++++++++++++++++ drivers/gpu/drm/xe/xe_sriov_pf.h | 1 + 6 files changed, 63 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_pf.c b/drivers/gpu/drm/xe/xe_gt_sriov_pf.c index 2ea81d81c0aeb..bdbd15f3afe38 100644 --- a/drivers/gpu/drm/xe/xe_gt_sriov_pf.c +++ b/drivers/gpu/drm/xe/xe_gt_sriov_pf.c @@ -16,6 +16,7 @@ #include "xe_gt_sriov_pf_migration.h" #include "xe_gt_sriov_pf_service.h" #include "xe_gt_sriov_printk.h" +#include "xe_guc_submit.h" #include "xe_mmio.h" #include "xe_pm.h" @@ -260,3 +261,27 @@ void xe_gt_sriov_pf_restart(struct xe_gt *gt) { pf_queue_restart(gt); } + +static void pf_flush_restart(struct xe_gt *gt) +{ + xe_gt_assert(gt, IS_SRIOV_PF(gt_to_xe(gt))); + flush_work(>->sriov.pf.workers.restart); +} + +/** + * xe_gt_sriov_pf_wait_ready() - Wait until per-GT PF SR-IOV support is ready. + * @gt: the &xe_gt + * + * This function can only be called on PF. + * + * Return: 0 on success or a negative error code on failure. + */ +int xe_gt_sriov_pf_wait_ready(struct xe_gt *gt) +{ + /* don't wait if there is another ongoing reset */ + if (xe_guc_read_stopped(>->uc.guc)) + return -EBUSY; + + pf_flush_restart(gt); + return 0; +} diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_pf.h b/drivers/gpu/drm/xe/xe_gt_sriov_pf.h index e2b2ff8132dc5..e7fde3f9937af 100644 --- a/drivers/gpu/drm/xe/xe_gt_sriov_pf.h +++ b/drivers/gpu/drm/xe/xe_gt_sriov_pf.h @@ -11,6 +11,7 @@ struct xe_gt; #ifdef CONFIG_PCI_IOV int xe_gt_sriov_pf_init_early(struct xe_gt *gt); int xe_gt_sriov_pf_init(struct xe_gt *gt); +int xe_gt_sriov_pf_wait_ready(struct xe_gt *gt); void xe_gt_sriov_pf_init_hw(struct xe_gt *gt); void xe_gt_sriov_pf_sanitize_hw(struct xe_gt *gt, unsigned int vfid); void xe_gt_sriov_pf_stop_prepare(struct xe_gt *gt); diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c b/drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c index bf679b21f4852..3ed245e04d0cf 100644 --- a/drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c +++ b/drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c @@ -22,6 +22,7 @@ #include "xe_gt_sriov_pf_policy.h" #include "xe_gt_sriov_pf_service.h" #include "xe_pm.h" +#include "xe_sriov_pf.h" /* * /sys/kernel/debug/dri/0/ @@ -205,7 +206,8 @@ static int CONFIG##_set(void *data, u64 val) \ return -EOVERFLOW; \ \ xe_pm_runtime_get(xe); \ - err = xe_gt_sriov_pf_config_set_##CONFIG(gt, vfid, val); \ + err = xe_sriov_pf_wait_ready(xe) ?: \ + xe_gt_sriov_pf_config_set_##CONFIG(gt, vfid, val); \ xe_pm_runtime_put(xe); \ \ return err; \ diff --git a/drivers/gpu/drm/xe/xe_pci_sriov.c b/drivers/gpu/drm/xe/xe_pci_sriov.c index 8813efdcafbb1..447a7867eecba 100644 --- a/drivers/gpu/drm/xe/xe_pci_sriov.c +++ b/drivers/gpu/drm/xe/xe_pci_sriov.c @@ -12,6 +12,7 @@ #include "xe_pci_sriov.h" #include "xe_pm.h" #include "xe_sriov.h" +#include "xe_sriov_pf.h" #include "xe_sriov_pf_helpers.h" #include "xe_sriov_printk.h" @@ -138,6 +139,10 @@ static int pf_enable_vfs(struct xe_device *xe, int num_vfs) xe_assert(xe, num_vfs <= total_vfs); xe_sriov_dbg(xe, "enabling %u VF%s\n", num_vfs, str_plural(num_vfs)); + err = xe_sriov_pf_wait_ready(xe); + if (err) + goto out; + /* * We must hold additional reference to the runtime PM to keep PF in D0 * during VFs lifetime, as our VFs do not implement the PM capability. @@ -169,7 +174,7 @@ static int pf_enable_vfs(struct xe_device *xe, int num_vfs) failed: pf_unprovision_vfs(xe, num_vfs); xe_pm_runtime_put(xe); - +out: xe_sriov_notice(xe, "Failed to enable %u VF%s (%pe)\n", num_vfs, str_plural(num_vfs), ERR_PTR(err)); return err; diff --git a/drivers/gpu/drm/xe/xe_sriov_pf.c b/drivers/gpu/drm/xe/xe_sriov_pf.c index afbdd894bd6ee..27ddf3cc80e9c 100644 --- a/drivers/gpu/drm/xe/xe_sriov_pf.c +++ b/drivers/gpu/drm/xe/xe_sriov_pf.c @@ -9,6 +9,7 @@ #include "xe_assert.h" #include "xe_device.h" +#include "xe_gt_sriov_pf.h" #include "xe_module.h" #include "xe_sriov.h" #include "xe_sriov_pf.h" @@ -102,6 +103,32 @@ int xe_sriov_pf_init_early(struct xe_device *xe) return 0; } +/** + * xe_sriov_pf_wait_ready() - Wait until PF is ready to operate. + * @xe: the &xe_device to test + * + * This function can only be called on PF. + * + * Return: 0 on success or a negative error code on failure. + */ +int xe_sriov_pf_wait_ready(struct xe_device *xe) +{ + struct xe_gt *gt; + unsigned int id; + int err; + + if (xe_device_wedged(xe)) + return -ECANCELED; + + for_each_gt(gt, xe, id) { + err = xe_gt_sriov_pf_wait_ready(gt); + if (err) + return err; + } + + return 0; +} + /** * xe_sriov_pf_print_vfs_summary - Print SR-IOV PF information. * @xe: the &xe_device to print info from diff --git a/drivers/gpu/drm/xe/xe_sriov_pf.h b/drivers/gpu/drm/xe/xe_sriov_pf.h index c392c3fcf0852..e3b34f8f5e046 100644 --- a/drivers/gpu/drm/xe/xe_sriov_pf.h +++ b/drivers/gpu/drm/xe/xe_sriov_pf.h @@ -15,6 +15,7 @@ struct xe_device; #ifdef CONFIG_PCI_IOV bool xe_sriov_pf_readiness(struct xe_device *xe); int xe_sriov_pf_init_early(struct xe_device *xe); +int xe_sriov_pf_wait_ready(struct xe_device *xe); void xe_sriov_pf_debugfs_register(struct xe_device *xe, struct dentry *root); void xe_sriov_pf_print_vfs_summary(struct xe_device *xe, struct drm_printer *p); #else -- 2.51.0