From 529bf8d1118bbaa1aa835563a22b0b5c64ca9d68 Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Wed, 28 Aug 2024 14:51:54 -0700 Subject: drm/xe/gsc: Do not attempt to load the GSC multiple times The GSC HW is only reset by driver FLR or D3cold entry. We don't support the former at runtime, while the latter is only supported on DGFX, for which we don't support GSC. Therefore, if GSC failed to load previously there is no need to try again because the HW is stuck in the error state. An assert has been added so that if we ever add DGFX support we'll know we need to handle the D3 case. v2: use "< 0" instead of "!= 0" in the FW state error check (Julia). Fixes: dd0e89e5edc2 ("drm/xe/gsc: GSC FW load") Signed-off-by: Daniele Ceraolo Spurio Cc: John Harrison Cc: Alan Previn Cc: # v6.8+ Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20240828215158.2743994-2-daniele.ceraolospurio@intel.com (cherry picked from commit 2160f6f6e3cf6893a83357c3b82ff8589bdc0f08) Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/xe/xe_gsc.c | 12 ++++++++++++ drivers/gpu/drm/xe/xe_uc_fw.h | 9 +++++++-- 2 files changed, 19 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_gsc.c b/drivers/gpu/drm/xe/xe_gsc.c index 2a612652bb13..29f96f409391 100644 --- a/drivers/gpu/drm/xe/xe_gsc.c +++ b/drivers/gpu/drm/xe/xe_gsc.c @@ -519,10 +519,22 @@ out_bo: void xe_gsc_load_start(struct xe_gsc *gsc) { struct xe_gt *gt = gsc_to_gt(gsc); + struct xe_device *xe = gt_to_xe(gt); if (!xe_uc_fw_is_loadable(&gsc->fw) || !gsc->q) return; + /* + * The GSC HW is only reset by driver FLR or D3cold entry. We don't + * support the former at runtime, while the latter is only supported on + * DGFX, for which we don't support GSC. Therefore, if GSC failed to + * load previously there is no need to try again because the HW is + * stuck in the error state. + */ + xe_assert(xe, !IS_DGFX(xe)); + if (xe_uc_fw_is_in_error_state(&gsc->fw)) + return; + /* GSC FW survives GT reset and D3Hot */ if (gsc_fw_is_loaded(gt)) { xe_uc_fw_change_status(&gsc->fw, XE_UC_FIRMWARE_TRANSFERRED); diff --git a/drivers/gpu/drm/xe/xe_uc_fw.h b/drivers/gpu/drm/xe/xe_uc_fw.h index c108e9d08e70..6195e353f269 100644 --- a/drivers/gpu/drm/xe/xe_uc_fw.h +++ b/drivers/gpu/drm/xe/xe_uc_fw.h @@ -65,7 +65,7 @@ const char *xe_uc_fw_status_repr(enum xe_uc_fw_status status) return ""; } -static inline int xe_uc_fw_status_to_error(enum xe_uc_fw_status status) +static inline int xe_uc_fw_status_to_error(const enum xe_uc_fw_status status) { switch (status) { case XE_UC_FIRMWARE_NOT_SUPPORTED: @@ -108,7 +108,7 @@ static inline const char *xe_uc_fw_type_repr(enum xe_uc_fw_type type) } static inline enum xe_uc_fw_status -__xe_uc_fw_status(struct xe_uc_fw *uc_fw) +__xe_uc_fw_status(const struct xe_uc_fw *uc_fw) { /* shouldn't call this before checking hw/blob availability */ XE_WARN_ON(uc_fw->status == XE_UC_FIRMWARE_UNINITIALIZED); @@ -156,6 +156,11 @@ static inline bool xe_uc_fw_is_overridden(const struct xe_uc_fw *uc_fw) return uc_fw->user_overridden; } +static inline bool xe_uc_fw_is_in_error_state(const struct xe_uc_fw *uc_fw) +{ + return xe_uc_fw_status_to_error(__xe_uc_fw_status(uc_fw)) < 0; +} + static inline void xe_uc_fw_sanitize(struct xe_uc_fw *uc_fw) { if (xe_uc_fw_is_loadable(uc_fw)) -- cgit v1.2.3 From fe13fd68337512da4b7aee6326c397e817a943c7 Mon Sep 17 00:00:00 2001 From: Matt Roper Date: Thu, 29 Aug 2024 15:06:21 -0700 Subject: drm/xe/pcode: Treat pcode as per-tile rather than per-GT There's only one instance of the pcode per tile, and for GT-related accesses both the primary and media GT share the same register interface. Since Xe was using per-GT locking, the pcode mutex wasn't actually protecting everything that it should since concurrent accesses related to a tile's primary GT and media GT were possible. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Signed-off-by: Matt Roper Reviewed-by: Lucas De Marchi Link: https://patchwork.freedesktop.org/patch/msgid/20240829220619.789159-5-matthew.d.roper@intel.com (cherry picked from commit 3034cc8107b8d0c7d1b56584394e215dab57f8a3) Signed-off-by: Rodrigo Vivi --- .../gpu/drm/xe/compat-i915-headers/intel_pcode.h | 8 +- .../gpu/drm/xe/compat-i915-headers/intel_uncore.h | 7 ++ drivers/gpu/drm/xe/xe_device_types.h | 6 ++ drivers/gpu/drm/xe/xe_gt.c | 2 - drivers/gpu/drm/xe/xe_gt_types.h | 6 -- drivers/gpu/drm/xe/xe_guc_pc.c | 2 +- drivers/gpu/drm/xe/xe_hwmon.c | 4 +- drivers/gpu/drm/xe/xe_pcode.c | 104 ++++++++++----------- drivers/gpu/drm/xe/xe_pcode.h | 16 ++-- drivers/gpu/drm/xe/xe_tile.c | 3 + drivers/gpu/drm/xe/xe_vram_freq.c | 6 +- 11 files changed, 85 insertions(+), 79 deletions(-) diff --git a/drivers/gpu/drm/xe/compat-i915-headers/intel_pcode.h b/drivers/gpu/drm/xe/compat-i915-headers/intel_pcode.h index 0c47661bdc6a..a473aa6697d0 100644 --- a/drivers/gpu/drm/xe/compat-i915-headers/intel_pcode.h +++ b/drivers/gpu/drm/xe/compat-i915-headers/intel_pcode.h @@ -13,7 +13,7 @@ static inline int snb_pcode_write_timeout(struct intel_uncore *uncore, u32 mbox, u32 val, int fast_timeout_us, int slow_timeout_ms) { - return xe_pcode_write_timeout(__compat_uncore_to_gt(uncore), mbox, val, + return xe_pcode_write_timeout(__compat_uncore_to_tile(uncore), mbox, val, slow_timeout_ms ?: 1); } @@ -21,13 +21,13 @@ static inline int snb_pcode_write(struct intel_uncore *uncore, u32 mbox, u32 val) { - return xe_pcode_write(__compat_uncore_to_gt(uncore), mbox, val); + return xe_pcode_write(__compat_uncore_to_tile(uncore), mbox, val); } static inline int snb_pcode_read(struct intel_uncore *uncore, u32 mbox, u32 *val, u32 *val1) { - return xe_pcode_read(__compat_uncore_to_gt(uncore), mbox, val, val1); + return xe_pcode_read(__compat_uncore_to_tile(uncore), mbox, val, val1); } static inline int @@ -35,7 +35,7 @@ skl_pcode_request(struct intel_uncore *uncore, u32 mbox, u32 request, u32 reply_mask, u32 reply, int timeout_base_ms) { - return xe_pcode_request(__compat_uncore_to_gt(uncore), mbox, request, reply_mask, reply, + return xe_pcode_request(__compat_uncore_to_tile(uncore), mbox, request, reply_mask, reply, timeout_base_ms); } diff --git a/drivers/gpu/drm/xe/compat-i915-headers/intel_uncore.h b/drivers/gpu/drm/xe/compat-i915-headers/intel_uncore.h index 083c4da2ea41..eb5b5f0e4bd9 100644 --- a/drivers/gpu/drm/xe/compat-i915-headers/intel_uncore.h +++ b/drivers/gpu/drm/xe/compat-i915-headers/intel_uncore.h @@ -17,6 +17,13 @@ static inline struct xe_gt *__compat_uncore_to_gt(struct intel_uncore *uncore) return xe_root_mmio_gt(xe); } +static inline struct xe_tile *__compat_uncore_to_tile(struct intel_uncore *uncore) +{ + struct xe_device *xe = container_of(uncore, struct xe_device, uncore); + + return xe_device_get_root_tile(xe); +} + static inline u32 intel_uncore_read(struct intel_uncore *uncore, i915_reg_t i915_reg) { diff --git a/drivers/gpu/drm/xe/xe_device_types.h b/drivers/gpu/drm/xe/xe_device_types.h index cbc582bcc90a..9e5fdf96750b 100644 --- a/drivers/gpu/drm/xe/xe_device_types.h +++ b/drivers/gpu/drm/xe/xe_device_types.h @@ -203,6 +203,12 @@ struct xe_tile { } vf; } sriov; + /** @pcode: tile's PCODE */ + struct { + /** @pcode.lock: protecting tile's PCODE mailbox data */ + struct mutex lock; + } pcode; + /** @migrate: Migration helper for vram blits and clearing */ struct xe_migrate *migrate; diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c index b9bcbbe27705..18cd3deb1498 100644 --- a/drivers/gpu/drm/xe/xe_gt.c +++ b/drivers/gpu/drm/xe/xe_gt.c @@ -47,7 +47,6 @@ #include "xe_migrate.h" #include "xe_mmio.h" #include "xe_pat.h" -#include "xe_pcode.h" #include "xe_pm.h" #include "xe_mocs.h" #include "xe_reg_sr.h" @@ -387,7 +386,6 @@ int xe_gt_init_early(struct xe_gt *gt) xe_tuning_process_gt(gt); xe_force_wake_init_gt(gt, gt_to_fw(gt)); - xe_pcode_init(gt); spin_lock_init(>->global_invl_lock); return 0; diff --git a/drivers/gpu/drm/xe/xe_gt_types.h b/drivers/gpu/drm/xe/xe_gt_types.h index 38a0d0e178c8..c582541970df 100644 --- a/drivers/gpu/drm/xe/xe_gt_types.h +++ b/drivers/gpu/drm/xe/xe_gt_types.h @@ -310,12 +310,6 @@ struct xe_gt { /** @eclass: per hardware engine class interface on the GT */ struct xe_hw_engine_class_intf eclass[XE_ENGINE_CLASS_MAX]; - /** @pcode: GT's PCODE */ - struct { - /** @pcode.lock: protecting GT's PCODE mailbox data */ - struct mutex lock; - } pcode; - /** @sysfs: sysfs' kobj used by xe_gt_sysfs */ struct kobject *sysfs; diff --git a/drivers/gpu/drm/xe/xe_guc_pc.c b/drivers/gpu/drm/xe/xe_guc_pc.c index 32e93a8127d4..ccd574e948aa 100644 --- a/drivers/gpu/drm/xe/xe_guc_pc.c +++ b/drivers/gpu/drm/xe/xe_guc_pc.c @@ -915,7 +915,7 @@ static void pc_init_pcode_freq(struct xe_guc_pc *pc) u32 min = DIV_ROUND_CLOSEST(pc->rpn_freq, GT_FREQUENCY_MULTIPLIER); u32 max = DIV_ROUND_CLOSEST(pc->rp0_freq, GT_FREQUENCY_MULTIPLIER); - XE_WARN_ON(xe_pcode_init_min_freq_table(pc_to_gt(pc), min, max)); + XE_WARN_ON(xe_pcode_init_min_freq_table(gt_to_tile(pc_to_gt(pc)), min, max)); } static int pc_init_freqs(struct xe_guc_pc *pc) diff --git a/drivers/gpu/drm/xe/xe_hwmon.c b/drivers/gpu/drm/xe/xe_hwmon.c index 1faeca70900e..98e3ec08279e 100644 --- a/drivers/gpu/drm/xe/xe_hwmon.c +++ b/drivers/gpu/drm/xe/xe_hwmon.c @@ -441,14 +441,14 @@ static int xe_hwmon_pcode_read_i1(struct xe_gt *gt, u32 *uval) if (gt_to_xe(gt)->info.platform == XE_DG2) return -ENXIO; - return xe_pcode_read(gt, PCODE_MBOX(PCODE_POWER_SETUP, + return xe_pcode_read(gt_to_tile(gt), PCODE_MBOX(PCODE_POWER_SETUP, POWER_SETUP_SUBCOMMAND_READ_I1, 0), uval, NULL); } static int xe_hwmon_pcode_write_i1(struct xe_gt *gt, u32 uval) { - return xe_pcode_write(gt, PCODE_MBOX(PCODE_POWER_SETUP, + return xe_pcode_write(gt_to_tile(gt), PCODE_MBOX(PCODE_POWER_SETUP, POWER_SETUP_SUBCOMMAND_WRITE_I1, 0), (uval & POWER_SETUP_I1_DATA_MASK)); } diff --git a/drivers/gpu/drm/xe/xe_pcode.c b/drivers/gpu/drm/xe/xe_pcode.c index 9c4eefdf6642..7397d556996a 100644 --- a/drivers/gpu/drm/xe/xe_pcode.c +++ b/drivers/gpu/drm/xe/xe_pcode.c @@ -12,7 +12,6 @@ #include "xe_assert.h" #include "xe_device.h" -#include "xe_gt.h" #include "xe_mmio.h" #include "xe_pcode_api.h" @@ -30,7 +29,7 @@ * - PCODE for display operations */ -static int pcode_mailbox_status(struct xe_gt *gt) +static int pcode_mailbox_status(struct xe_tile *tile) { u32 err; static const struct pcode_err_decode err_decode[] = { @@ -45,9 +44,9 @@ static int pcode_mailbox_status(struct xe_gt *gt) [PCODE_ERROR_MASK] = {-EPROTO, "Unknown"}, }; - err = xe_mmio_read32(gt, PCODE_MAILBOX) & PCODE_ERROR_MASK; + err = xe_mmio_read32(tile->primary_gt, PCODE_MAILBOX) & PCODE_ERROR_MASK; if (err) { - drm_err(>_to_xe(gt)->drm, "PCODE Mailbox failed: %d %s", err, + drm_err(&tile_to_xe(tile)->drm, "PCODE Mailbox failed: %d %s", err, err_decode[err].str ?: "Unknown"); return err_decode[err].errno ?: -EPROTO; } @@ -55,84 +54,85 @@ static int pcode_mailbox_status(struct xe_gt *gt) return 0; } -static int __pcode_mailbox_rw(struct xe_gt *gt, u32 mbox, u32 *data0, u32 *data1, +static int __pcode_mailbox_rw(struct xe_tile *tile, u32 mbox, u32 *data0, u32 *data1, unsigned int timeout_ms, bool return_data, bool atomic) { + struct xe_gt *mmio = tile->primary_gt; int err; - if (gt_to_xe(gt)->info.skip_pcode) + if (tile_to_xe(tile)->info.skip_pcode) return 0; - if ((xe_mmio_read32(gt, PCODE_MAILBOX) & PCODE_READY) != 0) + if ((xe_mmio_read32(mmio, PCODE_MAILBOX) & PCODE_READY) != 0) return -EAGAIN; - xe_mmio_write32(gt, PCODE_DATA0, *data0); - xe_mmio_write32(gt, PCODE_DATA1, data1 ? *data1 : 0); - xe_mmio_write32(gt, PCODE_MAILBOX, PCODE_READY | mbox); + xe_mmio_write32(mmio, PCODE_DATA0, *data0); + xe_mmio_write32(mmio, PCODE_DATA1, data1 ? *data1 : 0); + xe_mmio_write32(mmio, PCODE_MAILBOX, PCODE_READY | mbox); - err = xe_mmio_wait32(gt, PCODE_MAILBOX, PCODE_READY, 0, + err = xe_mmio_wait32(mmio, PCODE_MAILBOX, PCODE_READY, 0, timeout_ms * USEC_PER_MSEC, NULL, atomic); if (err) return err; if (return_data) { - *data0 = xe_mmio_read32(gt, PCODE_DATA0); + *data0 = xe_mmio_read32(mmio, PCODE_DATA0); if (data1) - *data1 = xe_mmio_read32(gt, PCODE_DATA1); + *data1 = xe_mmio_read32(mmio, PCODE_DATA1); } - return pcode_mailbox_status(gt); + return pcode_mailbox_status(tile); } -static int pcode_mailbox_rw(struct xe_gt *gt, u32 mbox, u32 *data0, u32 *data1, +static int pcode_mailbox_rw(struct xe_tile *tile, u32 mbox, u32 *data0, u32 *data1, unsigned int timeout_ms, bool return_data, bool atomic) { - if (gt_to_xe(gt)->info.skip_pcode) + if (tile_to_xe(tile)->info.skip_pcode) return 0; - lockdep_assert_held(>->pcode.lock); + lockdep_assert_held(&tile->pcode.lock); - return __pcode_mailbox_rw(gt, mbox, data0, data1, timeout_ms, return_data, atomic); + return __pcode_mailbox_rw(tile, mbox, data0, data1, timeout_ms, return_data, atomic); } -int xe_pcode_write_timeout(struct xe_gt *gt, u32 mbox, u32 data, int timeout) +int xe_pcode_write_timeout(struct xe_tile *tile, u32 mbox, u32 data, int timeout) { int err; - mutex_lock(>->pcode.lock); - err = pcode_mailbox_rw(gt, mbox, &data, NULL, timeout, false, false); - mutex_unlock(>->pcode.lock); + mutex_lock(&tile->pcode.lock); + err = pcode_mailbox_rw(tile, mbox, &data, NULL, timeout, false, false); + mutex_unlock(&tile->pcode.lock); return err; } -int xe_pcode_read(struct xe_gt *gt, u32 mbox, u32 *val, u32 *val1) +int xe_pcode_read(struct xe_tile *tile, u32 mbox, u32 *val, u32 *val1) { int err; - mutex_lock(>->pcode.lock); - err = pcode_mailbox_rw(gt, mbox, val, val1, 1, true, false); - mutex_unlock(>->pcode.lock); + mutex_lock(&tile->pcode.lock); + err = pcode_mailbox_rw(tile, mbox, val, val1, 1, true, false); + mutex_unlock(&tile->pcode.lock); return err; } -static int pcode_try_request(struct xe_gt *gt, u32 mbox, +static int pcode_try_request(struct xe_tile *tile, u32 mbox, u32 request, u32 reply_mask, u32 reply, u32 *status, bool atomic, int timeout_us, bool locked) { int slept, wait = 10; - xe_gt_assert(gt, timeout_us > 0); + xe_tile_assert(tile, timeout_us > 0); for (slept = 0; slept < timeout_us; slept += wait) { if (locked) - *status = pcode_mailbox_rw(gt, mbox, &request, NULL, 1, true, + *status = pcode_mailbox_rw(tile, mbox, &request, NULL, 1, true, atomic); else - *status = __pcode_mailbox_rw(gt, mbox, &request, NULL, 1, true, + *status = __pcode_mailbox_rw(tile, mbox, &request, NULL, 1, true, atomic); if ((*status == 0) && ((request & reply_mask) == reply)) return 0; @@ -149,7 +149,7 @@ static int pcode_try_request(struct xe_gt *gt, u32 mbox, /** * xe_pcode_request - send PCODE request until acknowledgment - * @gt: gt + * @tile: tile * @mbox: PCODE mailbox ID the request is targeted for * @request: request ID * @reply_mask: mask used to check for request acknowledgment @@ -166,17 +166,17 @@ static int pcode_try_request(struct xe_gt *gt, u32 mbox, * Returns 0 on success, %-ETIMEDOUT in case of a timeout, <0 in case of some * other error as reported by PCODE. */ -int xe_pcode_request(struct xe_gt *gt, u32 mbox, u32 request, - u32 reply_mask, u32 reply, int timeout_base_ms) +int xe_pcode_request(struct xe_tile *tile, u32 mbox, u32 request, + u32 reply_mask, u32 reply, int timeout_base_ms) { u32 status; int ret; - xe_gt_assert(gt, timeout_base_ms <= 3); + xe_tile_assert(tile, timeout_base_ms <= 3); - mutex_lock(>->pcode.lock); + mutex_lock(&tile->pcode.lock); - ret = pcode_try_request(gt, mbox, request, reply_mask, reply, &status, + ret = pcode_try_request(tile, mbox, request, reply_mask, reply, &status, false, timeout_base_ms * 1000, true); if (!ret) goto out; @@ -191,20 +191,20 @@ int xe_pcode_request(struct xe_gt *gt, u32 mbox, u32 request, * requests, and for any quirks of the PCODE firmware that delays * the request completion. */ - drm_err(>_to_xe(gt)->drm, + drm_err(&tile_to_xe(tile)->drm, "PCODE timeout, retrying with preemption disabled\n"); preempt_disable(); - ret = pcode_try_request(gt, mbox, request, reply_mask, reply, &status, + ret = pcode_try_request(tile, mbox, request, reply_mask, reply, &status, true, 50 * 1000, true); preempt_enable(); out: - mutex_unlock(>->pcode.lock); + mutex_unlock(&tile->pcode.lock); return status ? status : ret; } /** * xe_pcode_init_min_freq_table - Initialize PCODE's QOS frequency table - * @gt: gt instance + * @tile: tile instance * @min_gt_freq: Minimal (RPn) GT frequency in units of 50MHz. * @max_gt_freq: Maximal (RP0) GT frequency in units of 50MHz. * @@ -227,30 +227,30 @@ out: * - -EACCES, "PCODE Rejected" * - -EPROTO, "Unknown" */ -int xe_pcode_init_min_freq_table(struct xe_gt *gt, u32 min_gt_freq, +int xe_pcode_init_min_freq_table(struct xe_tile *tile, u32 min_gt_freq, u32 max_gt_freq) { int ret; u32 freq; - if (!gt_to_xe(gt)->info.has_llc) + if (!tile_to_xe(tile)->info.has_llc) return 0; if (max_gt_freq <= min_gt_freq) return -EINVAL; - mutex_lock(>->pcode.lock); + mutex_lock(&tile->pcode.lock); for (freq = min_gt_freq; freq <= max_gt_freq; freq++) { u32 data = freq << PCODE_FREQ_RING_RATIO_SHIFT | freq; - ret = pcode_mailbox_rw(gt, PCODE_WRITE_MIN_FREQ_TABLE, + ret = pcode_mailbox_rw(tile, PCODE_WRITE_MIN_FREQ_TABLE, &data, NULL, 1, false, false); if (ret) goto unlock; } unlock: - mutex_unlock(>->pcode.lock); + mutex_unlock(&tile->pcode.lock); return ret; } @@ -270,7 +270,7 @@ unlock: int xe_pcode_ready(struct xe_device *xe, bool locked) { u32 status, request = DGFX_GET_INIT_STATUS; - struct xe_gt *gt = xe_root_mmio_gt(xe); + struct xe_tile *tile = xe_device_get_root_tile(xe); int timeout_us = 180000000; /* 3 min */ int ret; @@ -281,15 +281,15 @@ int xe_pcode_ready(struct xe_device *xe, bool locked) return 0; if (locked) - mutex_lock(>->pcode.lock); + mutex_lock(&tile->pcode.lock); - ret = pcode_try_request(gt, DGFX_PCODE_STATUS, request, + ret = pcode_try_request(tile, DGFX_PCODE_STATUS, request, DGFX_INIT_STATUS_COMPLETE, DGFX_INIT_STATUS_COMPLETE, &status, false, timeout_us, locked); if (locked) - mutex_unlock(>->pcode.lock); + mutex_unlock(&tile->pcode.lock); if (ret) drm_err(&xe->drm, @@ -300,14 +300,14 @@ int xe_pcode_ready(struct xe_device *xe, bool locked) /** * xe_pcode_init: initialize components of PCODE - * @gt: gt instance + * @tile: tile instance * * This function initializes the xe_pcode component. * To be called once only during probe. */ -void xe_pcode_init(struct xe_gt *gt) +void xe_pcode_init(struct xe_tile *tile) { - drmm_mutex_init(>_to_xe(gt)->drm, >->pcode.lock); + drmm_mutex_init(&tile_to_xe(tile)->drm, &tile->pcode.lock); } /** diff --git a/drivers/gpu/drm/xe/xe_pcode.h b/drivers/gpu/drm/xe/xe_pcode.h index 3f54c6d2a57d..ba33991d72a7 100644 --- a/drivers/gpu/drm/xe/xe_pcode.h +++ b/drivers/gpu/drm/xe/xe_pcode.h @@ -7,21 +7,21 @@ #define _XE_PCODE_H_ #include -struct xe_gt; +struct xe_tile; struct xe_device; -void xe_pcode_init(struct xe_gt *gt); +void xe_pcode_init(struct xe_tile *tile); int xe_pcode_probe_early(struct xe_device *xe); int xe_pcode_ready(struct xe_device *xe, bool locked); -int xe_pcode_init_min_freq_table(struct xe_gt *gt, u32 min_gt_freq, +int xe_pcode_init_min_freq_table(struct xe_tile *tile, u32 min_gt_freq, u32 max_gt_freq); -int xe_pcode_read(struct xe_gt *gt, u32 mbox, u32 *val, u32 *val1); -int xe_pcode_write_timeout(struct xe_gt *gt, u32 mbox, u32 val, +int xe_pcode_read(struct xe_tile *tile, u32 mbox, u32 *val, u32 *val1); +int xe_pcode_write_timeout(struct xe_tile *tile, u32 mbox, u32 val, int timeout_ms); -#define xe_pcode_write(gt, mbox, val) \ - xe_pcode_write_timeout(gt, mbox, val, 1) +#define xe_pcode_write(tile, mbox, val) \ + xe_pcode_write_timeout(tile, mbox, val, 1) -int xe_pcode_request(struct xe_gt *gt, u32 mbox, u32 request, +int xe_pcode_request(struct xe_tile *tile, u32 mbox, u32 request, u32 reply_mask, u32 reply, int timeout_ms); #define PCODE_MBOX(mbcmd, param1, param2)\ diff --git a/drivers/gpu/drm/xe/xe_tile.c b/drivers/gpu/drm/xe/xe_tile.c index 15ea0a942f67..dda5268507d8 100644 --- a/drivers/gpu/drm/xe/xe_tile.c +++ b/drivers/gpu/drm/xe/xe_tile.c @@ -9,6 +9,7 @@ #include "xe_ggtt.h" #include "xe_gt.h" #include "xe_migrate.h" +#include "xe_pcode.h" #include "xe_sa.h" #include "xe_tile.h" #include "xe_tile_sysfs.h" @@ -124,6 +125,8 @@ int xe_tile_init_early(struct xe_tile *tile, struct xe_device *xe, u8 id) if (IS_ERR(tile->primary_gt)) return PTR_ERR(tile->primary_gt); + xe_pcode_init(tile); + return 0; } diff --git a/drivers/gpu/drm/xe/xe_vram_freq.c b/drivers/gpu/drm/xe/xe_vram_freq.c index 99ff95e408e0..b26e26d73dae 100644 --- a/drivers/gpu/drm/xe/xe_vram_freq.c +++ b/drivers/gpu/drm/xe/xe_vram_freq.c @@ -34,7 +34,6 @@ static ssize_t max_freq_show(struct device *dev, struct device_attribute *attr, char *buf) { struct xe_tile *tile = dev_to_tile(dev); - struct xe_gt *gt = tile->primary_gt; u32 val, mbox; int err; @@ -42,7 +41,7 @@ static ssize_t max_freq_show(struct device *dev, struct device_attribute *attr, | REG_FIELD_PREP(PCODE_MB_PARAM1, PCODE_MBOX_FC_SC_READ_FUSED_P0) | REG_FIELD_PREP(PCODE_MB_PARAM2, PCODE_MBOX_DOMAIN_HBM); - err = xe_pcode_read(gt, mbox, &val, NULL); + err = xe_pcode_read(tile, mbox, &val, NULL); if (err) return err; @@ -57,7 +56,6 @@ static ssize_t min_freq_show(struct device *dev, struct device_attribute *attr, char *buf) { struct xe_tile *tile = dev_to_tile(dev); - struct xe_gt *gt = tile->primary_gt; u32 val, mbox; int err; @@ -65,7 +63,7 @@ static ssize_t min_freq_show(struct device *dev, struct device_attribute *attr, | REG_FIELD_PREP(PCODE_MB_PARAM1, PCODE_MBOX_FC_SC_READ_FUSED_PN) | REG_FIELD_PREP(PCODE_MB_PARAM2, PCODE_MBOX_DOMAIN_HBM); - err = xe_pcode_read(gt, mbox, &val, NULL); + err = xe_pcode_read(tile, mbox, &val, NULL); if (err) return err; -- cgit v1.2.3 From 82122d1f54ec2244fb9c367ee2390b51ce08cf44 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 30 Aug 2024 14:35:06 -0400 Subject: drm/xe: Add missing runtime reference to wedged upon gt_reset Fixes this missed case: xe 0000:00:02.0: [drm] Missing outer runtime PM protection WARNING: CPU: 99 PID: 1455 at drivers/gpu/drm/xe/xe_pm.c:564 xe_pm_runtime_get_noresume+0x48/0x60 [xe] Call Trace: ? show_regs+0x67/0x70 ? __warn+0x94/0x1b0 ? xe_pm_runtime_get_noresume+0x48/0x60 [xe] ? report_bug+0x1b7/0x1d0 ? handle_bug+0x46/0x80 ? exc_invalid_op+0x19/0x70 ? asm_exc_invalid_op+0x1b/0x20 ? xe_pm_runtime_get_noresume+0x48/0x60 [xe] xe_device_declare_wedged+0x91/0x280 [xe] gt_reset_worker+0xa2/0x250 [xe] v2: Also move get and get the right Fixes tag (Himal, Brost) Fixes: fb74b205cdd2 ("drm/xe: Introduce a simple wedged state") Cc: Himal Prasad Ghimiray Cc: Matthew Brost Reviewed-by: Jonathan Cavitt Reviewed-by: Himal Prasad Ghimiray Link: https://patchwork.freedesktop.org/patch/msgid/20240830183507.298351-1-rodrigo.vivi@intel.com Signed-off-by: Rodrigo Vivi (cherry picked from commit bc947d9a8c3ebd207e52c0e35cfc88f3e1abe54f) Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/xe/xe_gt.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c index 18cd3deb1498..b8c73f69fbaf 100644 --- a/drivers/gpu/drm/xe/xe_gt.c +++ b/drivers/gpu/drm/xe/xe_gt.c @@ -753,12 +753,13 @@ static int gt_reset(struct xe_gt *gt) xe_gt_info(gt, "reset started\n"); + xe_pm_runtime_get(gt_to_xe(gt)); + if (xe_fault_inject_gt_reset()) { err = -ECANCELED; goto err_fail; } - xe_pm_runtime_get(gt_to_xe(gt)); xe_gt_sanitize(gt); err = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL); @@ -793,11 +794,11 @@ err_out: XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL)); err_msg: XE_WARN_ON(xe_uc_start(>->uc)); - xe_pm_runtime_put(gt_to_xe(gt)); err_fail: xe_gt_err(gt, "reset failed (%pe)\n", ERR_PTR(err)); xe_device_declare_wedged(gt_to_xe(gt)); + xe_pm_runtime_put(gt_to_xe(gt)); return err; } -- cgit v1.2.3 From 269a3f60843b9d2045e0c09d4bd34a91ed4222aa Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Tue, 6 Aug 2024 12:50:43 +0200 Subject: drm/xe/display: Match i915 driver suspend/resume sequences better Suspend fbdev sooner, and disable user access before suspending to prevent some races. I've noticed this when comparing xe suspend to i915's. Matches the following commits from i915: 24b412b1bfeb ("drm/i915: Disable intel HPD poll after DRM poll init/enable") 1ef28d86bea9 ("drm/i915: Suspend the framebuffer console earlier during system suspend") bd738d859e71 ("drm/i915: Prevent modesets during driver init/shutdown") Thanks to Imre for pointing me to those commits. Driver shutdown is currently missing, but I have some idea how to implement it next. Signed-off-by: Maarten Lankhorst Cc: Imre Deak Reviewed-by: Uma Shankar Link: https://patchwork.freedesktop.org/patch/msgid/20240806105044.596842-2-maarten.lankhorst@linux.intel.com Signed-off-by: Maarten Lankhorst,,, (cherry picked from commit 492be2a070f023c66aaef6ebd664567fda28c2a6) Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/xe/display/xe_display.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/xe/display/xe_display.c b/drivers/gpu/drm/xe/display/xe_display.c index 49de4e4f8a75..4b9ce1f34f4c 100644 --- a/drivers/gpu/drm/xe/display/xe_display.c +++ b/drivers/gpu/drm/xe/display/xe_display.c @@ -315,8 +315,11 @@ void xe_display_pm_suspend(struct xe_device *xe, bool runtime) * properly. */ intel_power_domains_disable(xe); - if (has_display(xe)) + intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true); + if (has_display(xe)) { drm_kms_helper_poll_disable(&xe->drm); + intel_display_driver_disable_user_access(xe); + } if (!runtime) intel_display_driver_suspend(xe); @@ -327,12 +330,13 @@ void xe_display_pm_suspend(struct xe_device *xe, bool runtime) intel_hpd_cancel_work(xe); + if (has_display(xe)) + intel_display_driver_suspend_access(xe); + intel_encoder_suspend_all(&xe->display); intel_opregion_suspend(xe, s2idle ? PCI_D1 : PCI_D3cold); - intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true); - intel_dmc_suspend(xe); } @@ -370,14 +374,19 @@ void xe_display_pm_resume(struct xe_device *xe, bool runtime) intel_display_driver_init_hw(xe); intel_hpd_init(xe); + if (has_display(xe)) + intel_display_driver_resume_access(xe); + /* MST sideband requires HPD interrupts enabled */ intel_dp_mst_resume(xe); if (!runtime) intel_display_driver_resume(xe); - intel_hpd_poll_disable(xe); - if (has_display(xe)) + if (has_display(xe)) { drm_kms_helper_poll_enable(&xe->drm); + intel_display_driver_enable_user_access(xe); + } + intel_hpd_poll_disable(xe); intel_opregion_resume(xe); -- cgit v1.2.3 From 880b398288487029ef5062ec5d7fcfc90672f782 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Fri, 23 Aug 2024 14:21:46 +0300 Subject: drm/xe: Suspend/resume user access only during system s/r Enable/Disable user access only during system suspend/resume. This should not happen during runtime s/r v2: rebased Reviewed-by: Arun R Murthy Signed-off-by: Imre Deak Signed-off-by: Vinod Govindapillai Link: https://patchwork.freedesktop.org/patch/msgid/20240823112148.327015-2-vinod.govindapillai@intel.com (cherry picked from commit a64e7e5b05e014dad9ae5858c9644d61400ec6ef) Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/xe/display/xe_display.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/xe/display/xe_display.c b/drivers/gpu/drm/xe/display/xe_display.c index 4b9ce1f34f4c..d83046c34f97 100644 --- a/drivers/gpu/drm/xe/display/xe_display.c +++ b/drivers/gpu/drm/xe/display/xe_display.c @@ -318,7 +318,8 @@ void xe_display_pm_suspend(struct xe_device *xe, bool runtime) intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true); if (has_display(xe)) { drm_kms_helper_poll_disable(&xe->drm); - intel_display_driver_disable_user_access(xe); + if (!runtime) + intel_display_driver_disable_user_access(xe); } if (!runtime) @@ -330,7 +331,7 @@ void xe_display_pm_suspend(struct xe_device *xe, bool runtime) intel_hpd_cancel_work(xe); - if (has_display(xe)) + if (!runtime && has_display(xe)) intel_display_driver_suspend_access(xe); intel_encoder_suspend_all(&xe->display); @@ -374,7 +375,7 @@ void xe_display_pm_resume(struct xe_device *xe, bool runtime) intel_display_driver_init_hw(xe); intel_hpd_init(xe); - if (has_display(xe)) + if (!runtime && has_display(xe)) intel_display_driver_resume_access(xe); /* MST sideband requires HPD interrupts enabled */ @@ -384,7 +385,8 @@ void xe_display_pm_resume(struct xe_device *xe, bool runtime) if (has_display(xe)) { drm_kms_helper_poll_enable(&xe->drm); - intel_display_driver_enable_user_access(xe); + if (!runtime) + intel_display_driver_enable_user_access(xe); } intel_hpd_poll_disable(xe); -- cgit v1.2.3 From 4bfc9c553f5ea8cbfe085824f3283885cf173c20 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 30 Aug 2024 14:35:07 -0400 Subject: drm/xe/display: Avoid encoder_suspend at runtime suspend Fix circular locking dependency on runtime suspend. <4> [74.952215] ====================================================== <4> [74.952217] WARNING: possible circular locking dependency detected <4> [74.952219] 6.10.0-rc7-xe #1 Not tainted <4> [74.952221] ------------------------------------------------------ <4> [74.952223] kworker/7:1/82 is trying to acquire lock: <4> [74.952226] ffff888120548488 (&dev->mode_config.mutex){+.+.}-{3:3}, at: drm_modeset_lock_all+0x40/0x1e0 [drm] <4> [74.952260] but task is already holding lock: <4> [74.952262] ffffffffa0ae59c0 (xe_pm_runtime_lockdep_map){+.+.}-{0:0}, at: xe_pm_runtime_suspend+0x2f/0x340 [xe] <4> [74.952322] which lock already depends on the new lock. The commit 'b1d90a86 ("drm/xe: Use the encoder suspend helper also used by the i915 driver")' didn't do anything wrong. It actually fixed a critical bug, because the encoder_suspend was never getting actually called because it was returning if (has_display(xe)) instead of if (!has_display(xe)). However, this ended up introducing the encoder suspend calls in the runtime routines as well, causing the circular locking dependency. Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2304 Fixes: b1d90a862c89 ("drm/xe: Use the encoder suspend helper also used by the i915 driver") Cc: Imre Deak Reviewed-by: Jonathan Cavitt Link: https://patchwork.freedesktop.org/patch/msgid/20240830183507.298351-2-rodrigo.vivi@intel.com Signed-off-by: Rodrigo Vivi (cherry picked from commit 8da19441d0a02b53e362df81843bb20db3a8006a) Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/xe/display/xe_display.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/xe/display/xe_display.c b/drivers/gpu/drm/xe/display/xe_display.c index d83046c34f97..c860fda410c8 100644 --- a/drivers/gpu/drm/xe/display/xe_display.c +++ b/drivers/gpu/drm/xe/display/xe_display.c @@ -331,10 +331,10 @@ void xe_display_pm_suspend(struct xe_device *xe, bool runtime) intel_hpd_cancel_work(xe); - if (!runtime && has_display(xe)) + if (!runtime && has_display(xe)) { intel_display_driver_suspend_access(xe); - - intel_encoder_suspend_all(&xe->display); + intel_encoder_suspend_all(&xe->display); + } intel_opregion_suspend(xe, s2idle ? PCI_D1 : PCI_D3cold); -- cgit v1.2.3