From 9d86d080cfb3ab935c842ac5525a90430a14c998 Mon Sep 17 00:00:00 2001 From: John Harrison Date: Wed, 2 Oct 2024 17:46:02 -0700 Subject: [PATCH 01/16] drm/xe/devcoredump: Use drm_puts and already cached local variables There are a bunch of calls to drm_printf with static strings. Switch them to drm_puts instead. There are also a bunch of 'coredump->snapshot.XXX' references when 'coredump->snapshot' has alread been cached locally as 'ss'. So use 'ss->XXX' instead. Signed-off-by: John Harrison Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-3-John.C.Harrison@Intel.com --- drivers/gpu/drm/xe/xe_devcoredump.c | 40 ++++++++++++++--------------- 1 file changed, 20 insertions(+), 20 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_devcoredump.c b/drivers/gpu/drm/xe/xe_devcoredump.c index bdb76e834e4c..d23719d5c2a3 100644 --- a/drivers/gpu/drm/xe/xe_devcoredump.c +++ b/drivers/gpu/drm/xe/xe_devcoredump.c @@ -85,9 +85,9 @@ static ssize_t __xe_devcoredump_read(char *buffer, size_t count, p = drm_coredump_printer(&iter); - drm_printf(&p, "**** Xe Device Coredump ****\n"); - drm_printf(&p, "kernel: " UTS_RELEASE "\n"); - drm_printf(&p, "module: " KBUILD_MODNAME "\n"); + drm_puts(&p, "**** Xe Device Coredump ****\n"); + drm_puts(&p, "kernel: " UTS_RELEASE "\n"); + drm_puts(&p, "module: " KBUILD_MODNAME "\n"); ts = ktime_to_timespec64(ss->snapshot_time); drm_printf(&p, "Snapshot time: %lld.%09ld\n", ts.tv_sec, ts.tv_nsec); @@ -96,20 +96,20 @@ static ssize_t __xe_devcoredump_read(char *buffer, size_t count, drm_printf(&p, "Process: %s\n", ss->process_name); xe_device_snapshot_print(xe, &p); - drm_printf(&p, "\n**** GuC CT ****\n"); - xe_guc_ct_snapshot_print(coredump->snapshot.ct, &p); - xe_guc_exec_queue_snapshot_print(coredump->snapshot.ge, &p); + drm_puts(&p, "\n**** GuC CT ****\n"); + xe_guc_ct_snapshot_print(ss->ct, &p); + xe_guc_exec_queue_snapshot_print(ss->ge, &p); - drm_printf(&p, "\n**** Job ****\n"); - xe_sched_job_snapshot_print(coredump->snapshot.job, &p); + drm_puts(&p, "\n**** Job ****\n"); + xe_sched_job_snapshot_print(ss->job, &p); - drm_printf(&p, "\n**** HW Engines ****\n"); + drm_puts(&p, "\n**** HW Engines ****\n"); for (i = 0; i < XE_NUM_HW_ENGINES; i++) - if (coredump->snapshot.hwe[i]) - xe_hw_engine_snapshot_print(coredump->snapshot.hwe[i], - &p); - drm_printf(&p, "\n**** VM state ****\n"); - xe_vm_snapshot_print(coredump->snapshot.vm, &p); + if (ss->hwe[i]) + xe_hw_engine_snapshot_print(ss->hwe[i], &p); + + drm_puts(&p, "\n**** VM state ****\n"); + xe_vm_snapshot_print(ss->vm, &p); return count - iter.remain; } @@ -247,18 +247,18 @@ static void devcoredump_snapshot(struct xe_devcoredump *coredump, if (xe_force_wake_get(gt_to_fw(q->gt), XE_FORCEWAKE_ALL)) xe_gt_info(ss->gt, "failed to get forcewake for coredump capture\n"); - coredump->snapshot.ct = xe_guc_ct_snapshot_capture(&guc->ct, true); - coredump->snapshot.ge = xe_guc_exec_queue_snapshot_capture(q); - coredump->snapshot.job = xe_sched_job_snapshot_capture(job); - coredump->snapshot.vm = xe_vm_snapshot_capture(q->vm); + ss->ct = xe_guc_ct_snapshot_capture(&guc->ct, true); + ss->ge = xe_guc_exec_queue_snapshot_capture(q); + ss->job = xe_sched_job_snapshot_capture(job); + ss->vm = xe_vm_snapshot_capture(q->vm); for_each_hw_engine(hwe, q->gt, id) { if (hwe->class != q->hwe->class || !(BIT(hwe->logical_instance) & adj_logical_mask)) { - coredump->snapshot.hwe[id] = NULL; + ss->hwe[id] = NULL; continue; } - coredump->snapshot.hwe[id] = xe_hw_engine_snapshot_capture(hwe); + ss->hwe[id] = xe_hw_engine_snapshot_capture(hwe); } queue_work(system_unbound_wq, &ss->work); -- 2.50.1 From c28fd6c358db44c87a1408f27ba412c94e25e6c2 Mon Sep 17 00:00:00 2001 From: John Harrison Date: Wed, 2 Oct 2024 17:46:03 -0700 Subject: [PATCH 02/16] drm/xe/devcoredump: Improve section headings and add tile info The xe_guc_exec_queue_snapshot is not really a GuC internal thing and is definitely not a GuC CT thing. So give it its own section heading. The snapshot itself is really a capture of the submission backend's internal state. Although all it currently prints out is the submission contexts. So label it as 'Contexts'. If more general state is added later then it could be change to 'Submission backend' or some such. Further, everything from the GuC CT section onwards is GT specific but there was no indication of which GT it was related to (and that is impossible to work out from the other fields that are given). So add a GT section heading. Also include the tile id of the GT, because again significant information. Lastly, drop a couple of unnecessary line feeds within sections. v2: Add GT section heading, add tile id to device section. Signed-off-by: John Harrison Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-4-John.C.Harrison@Intel.com --- drivers/gpu/drm/xe/xe_devcoredump.c | 5 +++++ drivers/gpu/drm/xe/xe_devcoredump_types.h | 3 ++- drivers/gpu/drm/xe/xe_device.c | 1 + drivers/gpu/drm/xe/xe_guc_submit.c | 2 +- drivers/gpu/drm/xe/xe_hw_engine.c | 1 - 5 files changed, 9 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_devcoredump.c b/drivers/gpu/drm/xe/xe_devcoredump.c index d23719d5c2a3..2690f1d1cde4 100644 --- a/drivers/gpu/drm/xe/xe_devcoredump.c +++ b/drivers/gpu/drm/xe/xe_devcoredump.c @@ -96,8 +96,13 @@ static ssize_t __xe_devcoredump_read(char *buffer, size_t count, drm_printf(&p, "Process: %s\n", ss->process_name); xe_device_snapshot_print(xe, &p); + drm_printf(&p, "\n**** GT #%d ****\n", ss->gt->info.id); + drm_printf(&p, "\tTile: %d\n", ss->gt->tile->id); + drm_puts(&p, "\n**** GuC CT ****\n"); xe_guc_ct_snapshot_print(ss->ct, &p); + + drm_puts(&p, "\n**** Contexts ****\n"); xe_guc_exec_queue_snapshot_print(ss->ge, &p); drm_puts(&p, "\n**** Job ****\n"); diff --git a/drivers/gpu/drm/xe/xe_devcoredump_types.h b/drivers/gpu/drm/xe/xe_devcoredump_types.h index 440d05d77a5a..3cc2f095fdfb 100644 --- a/drivers/gpu/drm/xe/xe_devcoredump_types.h +++ b/drivers/gpu/drm/xe/xe_devcoredump_types.h @@ -37,7 +37,8 @@ struct xe_devcoredump_snapshot { /* GuC snapshots */ /** @ct: GuC CT snapshot */ struct xe_guc_ct_snapshot *ct; - /** @ge: Guc Engine snapshot */ + + /** @ge: GuC Submission Engine snapshot */ struct xe_guc_submit_exec_queue_snapshot *ge; /** @hwe: HW Engine snapshot array */ diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c index 1940cd3229f4..cd241a8e1838 100644 --- a/drivers/gpu/drm/xe/xe_device.c +++ b/drivers/gpu/drm/xe/xe_device.c @@ -964,6 +964,7 @@ void xe_device_snapshot_print(struct xe_device *xe, struct drm_printer *p) for_each_gt(gt, xe, id) { drm_printf(p, "GT id: %u\n", id); + drm_printf(p, "\tTile: %u\n", gt->tile->id); drm_printf(p, "\tType: %s\n", gt->info.type == XE_GT_TYPE_MAIN ? "main" : "media"); drm_printf(p, "\tIP ver: %u.%u.%u\n", diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c index 8a5c21a87977..fc4008018d1f 100644 --- a/drivers/gpu/drm/xe/xe_guc_submit.c +++ b/drivers/gpu/drm/xe/xe_guc_submit.c @@ -2237,7 +2237,7 @@ xe_guc_exec_queue_snapshot_print(struct xe_guc_submit_exec_queue_snapshot *snaps if (!snapshot) return; - drm_printf(p, "\nGuC ID: %d\n", snapshot->guc.id); + drm_printf(p, "GuC ID: %d\n", snapshot->guc.id); drm_printf(p, "\tName: %s\n", snapshot->name); drm_printf(p, "\tClass: %d\n", snapshot->class); drm_printf(p, "\tLogical mask: 0x%x\n", snapshot->logical_mask); diff --git a/drivers/gpu/drm/xe/xe_hw_engine.c b/drivers/gpu/drm/xe/xe_hw_engine.c index ea6d9ef7fab6..6c9c27304cdc 100644 --- a/drivers/gpu/drm/xe/xe_hw_engine.c +++ b/drivers/gpu/drm/xe/xe_hw_engine.c @@ -1084,7 +1084,6 @@ void xe_hw_engine_snapshot_print(struct xe_hw_engine_snapshot *snapshot, if (snapshot->hwe->class == XE_ENGINE_CLASS_COMPUTE) drm_printf(p, "\tRCU_MODE: 0x%08x\n", snapshot->reg.rcu_mode); - drm_puts(p, "\n"); } /** -- 2.50.1 From ec1455ce7e35a31289d2dbc1070b980538698921 Mon Sep 17 00:00:00 2001 From: John Harrison Date: Wed, 2 Oct 2024 17:46:04 -0700 Subject: [PATCH 03/16] drm/xe/devcoredump: Add ASCII85 dump helper function There is a need to include the GuC log and other large binary objects in core dumps and via dmesg. So add a helper for dumping to a printer function via conversion to ASCII85 encoding. Another issue with dumping such a large buffer is that it can be slow, especially if dumping to dmesg over a serial port. So add a yield to prevent the 'task has been stuck for 120s' kernel hang check feature from firing. v2: Add a prefix to the output string. Fix memory allocation bug. v3: Correct a string size calculation and clean up a define (review feedback from Julia F). Signed-off-by: John Harrison Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-5-John.C.Harrison@Intel.com --- drivers/gpu/drm/xe/xe_devcoredump.c | 87 +++++++++++++++++++++++++++++ drivers/gpu/drm/xe/xe_devcoredump.h | 6 ++ 2 files changed, 93 insertions(+) diff --git a/drivers/gpu/drm/xe/xe_devcoredump.c b/drivers/gpu/drm/xe/xe_devcoredump.c index 2690f1d1cde4..0884c49942fe 100644 --- a/drivers/gpu/drm/xe/xe_devcoredump.c +++ b/drivers/gpu/drm/xe/xe_devcoredump.c @@ -6,6 +6,7 @@ #include "xe_devcoredump.h" #include "xe_devcoredump_types.h" +#include #include #include @@ -315,3 +316,89 @@ int xe_devcoredump_init(struct xe_device *xe) } #endif + +/** + * xe_print_blob_ascii85 - print a BLOB to some useful location in ASCII85 + * + * The output is split to multiple lines because some print targets, e.g. dmesg + * cannot handle arbitrarily long lines. Note also that printing to dmesg in + * piece-meal fashion is not possible, each separate call to drm_puts() has a + * line-feed automatically added! Therefore, the entire output line must be + * constructed in a local buffer first, then printed in one atomic output call. + * + * There is also a scheduler yield call to prevent the 'task has been stuck for + * 120s' kernel hang check feature from firing when printing to a slow target + * such as dmesg over a serial port. + * + * TODO: Add compression prior to the ASCII85 encoding to shrink huge buffers down. + * + * @p: the printer object to output to + * @prefix: optional prefix to add to output string + * @blob: the Binary Large OBject to dump out + * @offset: offset in bytes to skip from the front of the BLOB, must be a multiple of sizeof(u32) + * @size: the size in bytes of the BLOB, must be a multiple of sizeof(u32) + */ +void xe_print_blob_ascii85(struct drm_printer *p, const char *prefix, + const void *blob, size_t offset, size_t size) +{ + const u32 *blob32 = (const u32 *)blob; + char buff[ASCII85_BUFSZ], *line_buff; + size_t line_pos = 0; + +#define DMESG_MAX_LINE_LEN 800 +#define MIN_SPACE (ASCII85_BUFSZ + 2) /* 85 + "\n\0" */ + + if (size & 3) + drm_printf(p, "Size not word aligned: %zu", size); + if (offset & 3) + drm_printf(p, "Offset not word aligned: %zu", size); + + line_buff = kzalloc(DMESG_MAX_LINE_LEN, GFP_KERNEL); + if (IS_ERR_OR_NULL(line_buff)) { + drm_printf(p, "Failed to allocate line buffer: %pe", line_buff); + return; + } + + blob32 += offset / sizeof(*blob32); + size /= sizeof(*blob32); + + if (prefix) { + strscpy(line_buff, prefix, DMESG_MAX_LINE_LEN - MIN_SPACE - 2); + line_pos = strlen(line_buff); + + line_buff[line_pos++] = ':'; + line_buff[line_pos++] = ' '; + } + + while (size--) { + u32 val = *(blob32++); + + strscpy(line_buff + line_pos, ascii85_encode(val, buff), + DMESG_MAX_LINE_LEN - line_pos); + line_pos += strlen(line_buff + line_pos); + + if ((line_pos + MIN_SPACE) >= DMESG_MAX_LINE_LEN) { + line_buff[line_pos++] = '\n'; + line_buff[line_pos++] = 0; + + drm_puts(p, line_buff); + + line_pos = 0; + + /* Prevent 'stuck thread' time out errors */ + cond_resched(); + } + } + + if (line_pos) { + line_buff[line_pos++] = '\n'; + line_buff[line_pos++] = 0; + + drm_puts(p, line_buff); + } + + kfree(line_buff); + +#undef MIN_SPACE +#undef DMESG_MAX_LINE_LEN +} diff --git a/drivers/gpu/drm/xe/xe_devcoredump.h b/drivers/gpu/drm/xe/xe_devcoredump.h index e2fa65ce0932..a4eebc285fc8 100644 --- a/drivers/gpu/drm/xe/xe_devcoredump.h +++ b/drivers/gpu/drm/xe/xe_devcoredump.h @@ -6,6 +6,9 @@ #ifndef _XE_DEVCOREDUMP_H_ #define _XE_DEVCOREDUMP_H_ +#include + +struct drm_printer; struct xe_device; struct xe_sched_job; @@ -23,4 +26,7 @@ static inline int xe_devcoredump_init(struct xe_device *xe) } #endif +void xe_print_blob_ascii85(struct drm_printer *p, const char *prefix, + const void *blob, size_t offset, size_t size); + #endif -- 2.50.1 From a59a403419aa03d5e44c8cf014e415490395b17f Mon Sep 17 00:00:00 2001 From: John Harrison Date: Wed, 2 Oct 2024 17:46:05 -0700 Subject: [PATCH 04/16] drm/xe/guc: Copy GuC log prior to dumping Add an extra stage to the GuC log print to copy the log buffer into regular host memory first, rather than printing the live GPU buffer object directly. Doing so helps prevent inconsistencies due to the log being updated as it is being dumped. It also allows the use of the ASCII85 helper function for printing the log in a more compact form than a straight hex dump. v2: Use %zx instead of %lx for size_t prints. v3: Replace hexdump code with ascii85 call (review feedback from Matthew B). Move chunking code into next patch as that reduces the deltas of both. v4: Add a prefix to the ASCII85 output to aid tool parsing. Signed-off-by: John Harrison Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-6-John.C.Harrison@Intel.com --- drivers/gpu/drm/xe/xe_guc_log.c | 40 +++++++++++++++++++-------------- 1 file changed, 23 insertions(+), 17 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_guc_log.c b/drivers/gpu/drm/xe/xe_guc_log.c index 651543721ce5..bb9446b2d29d 100644 --- a/drivers/gpu/drm/xe/xe_guc_log.c +++ b/drivers/gpu/drm/xe/xe_guc_log.c @@ -8,9 +8,12 @@ #include #include +#include #include "xe_bo.h" +#include "xe_devcoredump.h" #include "xe_gt.h" +#include "xe_gt_printk.h" #include "xe_map.h" #include "xe_module.h" @@ -51,32 +54,35 @@ static size_t guc_log_size(void) CAPTURE_BUFFER_SIZE; } +/** + * xe_guc_log_print - dump a copy of the GuC log to some useful location + * @log: GuC log structure + * @p: the printer object to output to + */ void xe_guc_log_print(struct xe_guc_log *log, struct drm_printer *p) { struct xe_device *xe = log_to_xe(log); size_t size; - int i, j; + void *copy; - xe_assert(xe, log->bo); + if (!log->bo) { + drm_puts(p, "GuC log buffer not allocated"); + return; + } size = log->bo->size; -#define DW_PER_READ 128 - xe_assert(xe, !(size % (DW_PER_READ * sizeof(u32)))); - for (i = 0; i < size / sizeof(u32); i += DW_PER_READ) { - u32 read[DW_PER_READ]; - - xe_map_memcpy_from(xe, read, &log->bo->vmap, i * sizeof(u32), - DW_PER_READ * sizeof(u32)); -#define DW_PER_PRINT 4 - for (j = 0; j < DW_PER_READ / DW_PER_PRINT; ++j) { - u32 *print = read + j * DW_PER_PRINT; - - drm_printf(p, "0x%08x 0x%08x 0x%08x 0x%08x\n", - *(print + 0), *(print + 1), - *(print + 2), *(print + 3)); - } + copy = vmalloc(size); + if (!copy) { + drm_printf(p, "Failed to allocate %zu", size); + return; } + + xe_map_memcpy_from(xe, copy, &log->bo->vmap, 0, size); + + xe_print_blob_ascii85(p, "Log data", copy, 0, size); + + vfree(copy); } int xe_guc_log_init(struct xe_guc_log *log) -- 2.50.1 From d8ce1a97722617317b04eb9f19ab8d6d95379f7a Mon Sep 17 00:00:00 2001 From: John Harrison Date: Wed, 2 Oct 2024 17:46:06 -0700 Subject: [PATCH 05/16] drm/xe/guc: Use a two stage dump for GuC logs and add more info Split the GuC log dump into a two stage snapshot and print mechanism. This allows the log to be captured at the point of an error (which may be in a restricted context) and then dump it out later (from a regular context such as a worker function or a sysfs file handler). Also add a bunch of other useful pieces of information that can help (or are fundamentally required!) to decode and parse the log. v2: Add kerneldoc and fix a couple of comment typos - review feedback from Michal W. v3: Move chunking code to this patch as it makes the deltas simpler. Fix a bunch of kerneldoc issues. v4: Move the CS frequency out of the coredump snapshot function into the debugfs only code (as that info is already part of the main devcoredump). Add a header to the debugfs log to match the one in the devcoredump to aid processing by a unified tool. Add forcewake to the GuC timestamp read so it actually works. v6: Add colon to GuC version string (review feedback by Julia F). Signed-off-by: John Harrison Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-7-John.C.Harrison@Intel.com --- drivers/gpu/drm/xe/regs/xe_guc_regs.h | 1 + drivers/gpu/drm/xe/xe_guc_log.c | 178 +++++++++++++++++++++++--- drivers/gpu/drm/xe/xe_guc_log.h | 4 + drivers/gpu/drm/xe/xe_guc_log_types.h | 27 ++++ 4 files changed, 195 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/xe/regs/xe_guc_regs.h b/drivers/gpu/drm/xe/regs/xe_guc_regs.h index a5fd14307f94..b27b73680c12 100644 --- a/drivers/gpu/drm/xe/regs/xe_guc_regs.h +++ b/drivers/gpu/drm/xe/regs/xe_guc_regs.h @@ -84,6 +84,7 @@ #define HUC_LOADING_AGENT_GUC REG_BIT(1) #define GUC_WOPCM_OFFSET_VALID REG_BIT(0) #define GUC_MAX_IDLE_COUNT XE_REG(0xc3e4) +#define GUC_PMTIMESTAMP XE_REG(0xc3e8) #define GUC_SEND_INTERRUPT XE_REG(0xc4c8) #define GUC_SEND_TRIGGER REG_BIT(0) diff --git a/drivers/gpu/drm/xe/xe_guc_log.c b/drivers/gpu/drm/xe/xe_guc_log.c index bb9446b2d29d..79e01d724509 100644 --- a/drivers/gpu/drm/xe/xe_guc_log.c +++ b/drivers/gpu/drm/xe/xe_guc_log.c @@ -8,15 +8,23 @@ #include #include -#include +#include "regs/xe_guc_regs.h" #include "xe_bo.h" #include "xe_devcoredump.h" +#include "xe_force_wake.h" #include "xe_gt.h" #include "xe_gt_printk.h" #include "xe_map.h" +#include "xe_mmio.h" #include "xe_module.h" +static struct xe_guc * +log_to_guc(struct xe_guc_log *log) +{ + return container_of(log, struct xe_guc, log); +} + static struct xe_gt * log_to_gt(struct xe_guc_log *log) { @@ -54,35 +62,175 @@ static size_t guc_log_size(void) CAPTURE_BUFFER_SIZE; } +#define GUC_LOG_CHUNK_SIZE SZ_2M + +static struct xe_guc_log_snapshot *xe_guc_log_snapshot_alloc(struct xe_guc_log *log, bool atomic) +{ + struct xe_guc_log_snapshot *snapshot; + size_t remain; + int i; + + snapshot = kzalloc(sizeof(*snapshot), atomic ? GFP_ATOMIC : GFP_KERNEL); + if (!snapshot) + return NULL; + + /* + * NB: kmalloc has a hard limit well below the maximum GuC log buffer size. + * Also, can't use vmalloc as might be called from atomic context. So need + * to break the buffer up into smaller chunks that can be allocated. + */ + snapshot->size = log->bo->size; + snapshot->num_chunks = DIV_ROUND_UP(snapshot->size, GUC_LOG_CHUNK_SIZE); + + snapshot->copy = kcalloc(snapshot->num_chunks, sizeof(*snapshot->copy), + atomic ? GFP_ATOMIC : GFP_KERNEL); + if (!snapshot->copy) + goto fail_snap; + + remain = snapshot->size; + for (i = 0; i < snapshot->num_chunks; i++) { + size_t size = min(GUC_LOG_CHUNK_SIZE, remain); + + snapshot->copy[i] = kmalloc(size, atomic ? GFP_ATOMIC : GFP_KERNEL); + if (!snapshot->copy[i]) + goto fail_copy; + remain -= size; + } + + return snapshot; + +fail_copy: + for (i = 0; i < snapshot->num_chunks; i++) + kfree(snapshot->copy[i]); + kfree(snapshot->copy); +fail_snap: + kfree(snapshot); + return NULL; +} + /** - * xe_guc_log_print - dump a copy of the GuC log to some useful location + * xe_guc_log_snapshot_free - free a previously captured GuC log snapshot + * @snapshot: GuC log snapshot structure + * + * Return: pointer to a newly allocated snapshot object or null if out of memory. Caller is + * responsible for calling xe_guc_log_snapshot_free when done with the snapshot. + */ +void xe_guc_log_snapshot_free(struct xe_guc_log_snapshot *snapshot) +{ + int i; + + if (!snapshot) + return; + + if (!snapshot->copy) { + for (i = 0; i < snapshot->num_chunks; i++) + kfree(snapshot->copy[i]); + kfree(snapshot->copy); + } + + kfree(snapshot); +} + +/** + * xe_guc_log_snapshot_capture - create a new snapshot copy the GuC log for later dumping * @log: GuC log structure - * @p: the printer object to output to + * @atomic: is the call inside an atomic section of some kind? + * + * Return: pointer to a newly allocated snapshot object or null if out of memory. Caller is + * responsible for calling xe_guc_log_snapshot_free when done with the snapshot. */ -void xe_guc_log_print(struct xe_guc_log *log, struct drm_printer *p) +struct xe_guc_log_snapshot *xe_guc_log_snapshot_capture(struct xe_guc_log *log, bool atomic) { + struct xe_guc_log_snapshot *snapshot; struct xe_device *xe = log_to_xe(log); - size_t size; - void *copy; + struct xe_guc *guc = log_to_guc(log); + struct xe_gt *gt = log_to_gt(log); + size_t remain; + int i, err; if (!log->bo) { - drm_puts(p, "GuC log buffer not allocated"); - return; + xe_gt_err(gt, "GuC log buffer not allocated\n"); + return NULL; + } + + snapshot = xe_guc_log_snapshot_alloc(log, atomic); + if (!snapshot) { + xe_gt_err(gt, "GuC log snapshot not allocated\n"); + return NULL; } - size = log->bo->size; + remain = snapshot->size; + for (i = 0; i < snapshot->num_chunks; i++) { + size_t size = min(GUC_LOG_CHUNK_SIZE, remain); + + xe_map_memcpy_from(xe, snapshot->copy[i], &log->bo->vmap, + i * GUC_LOG_CHUNK_SIZE, size); + remain -= size; + } + + err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT); + if (err) { + snapshot->stamp = ~0; + } else { + snapshot->stamp = xe_mmio_read32(>->mmio, GUC_PMTIMESTAMP); + xe_force_wake_put(gt_to_fw(gt), XE_FW_GT); + } + snapshot->ktime = ktime_get_boottime_ns(); + snapshot->level = log->level; + snapshot->ver_found = guc->fw.versions.found[XE_UC_FW_VER_RELEASE]; + snapshot->ver_want = guc->fw.versions.wanted; + snapshot->path = guc->fw.path; + + return snapshot; +} + +/** + * xe_guc_log_snapshot_print - dump a previously saved copy of the GuC log to some useful location + * @snapshot: a snapshot of the GuC log + * @p: the printer object to output to + */ +void xe_guc_log_snapshot_print(struct xe_guc_log_snapshot *snapshot, struct drm_printer *p) +{ + size_t remain; + int i; - copy = vmalloc(size); - if (!copy) { - drm_printf(p, "Failed to allocate %zu", size); + if (!snapshot) { + drm_printf(p, "GuC log snapshot not allocated!\n"); return; } - xe_map_memcpy_from(xe, copy, &log->bo->vmap, 0, size); + drm_printf(p, "GuC firmware: %s\n", snapshot->path); + drm_printf(p, "GuC version: %u.%u.%u (wanted %u.%u.%u)\n", + snapshot->ver_found.major, snapshot->ver_found.minor, snapshot->ver_found.patch, + snapshot->ver_want.major, snapshot->ver_want.minor, snapshot->ver_want.patch); + drm_printf(p, "Kernel timestamp: 0x%08llX [%llu]\n", snapshot->ktime, snapshot->ktime); + drm_printf(p, "GuC timestamp: 0x%08X [%u]\n", snapshot->stamp, snapshot->stamp); + drm_printf(p, "Log level: %u\n", snapshot->level); + + remain = snapshot->size; + for (i = 0; i < snapshot->num_chunks; i++) { + size_t size = min(GUC_LOG_CHUNK_SIZE, remain); + + xe_print_blob_ascii85(p, i ? NULL : "Log data", snapshot->copy[i], 0, size); + remain -= size; + } +} + +/** + * xe_guc_log_print - dump a copy of the GuC log to some useful location + * @log: GuC log structure + * @p: the printer object to output to + */ +void xe_guc_log_print(struct xe_guc_log *log, struct drm_printer *p) +{ + struct xe_guc_log_snapshot *snapshot; - xe_print_blob_ascii85(p, "Log data", copy, 0, size); + drm_printf(p, "**** GuC Log ****\n"); - vfree(copy); + snapshot = xe_guc_log_snapshot_capture(log, false); + drm_printf(p, "CS reference clock: %u\n", log_to_gt(log)->info.reference_clock); + xe_guc_log_snapshot_print(snapshot, p); + xe_guc_log_snapshot_free(snapshot); } int xe_guc_log_init(struct xe_guc_log *log) diff --git a/drivers/gpu/drm/xe/xe_guc_log.h b/drivers/gpu/drm/xe/xe_guc_log.h index 2d25ab28b4b3..949d2c98343d 100644 --- a/drivers/gpu/drm/xe/xe_guc_log.h +++ b/drivers/gpu/drm/xe/xe_guc_log.h @@ -9,6 +9,7 @@ #include "xe_guc_log_types.h" struct drm_printer; +struct xe_device; #if IS_ENABLED(CONFIG_DRM_XE_LARGE_GUC_BUFFER) #define CRASH_BUFFER_SIZE SZ_1M @@ -38,6 +39,9 @@ struct drm_printer; int xe_guc_log_init(struct xe_guc_log *log); void xe_guc_log_print(struct xe_guc_log *log, struct drm_printer *p); +struct xe_guc_log_snapshot *xe_guc_log_snapshot_capture(struct xe_guc_log *log, bool atomic); +void xe_guc_log_snapshot_print(struct xe_guc_log_snapshot *snapshot, struct drm_printer *p); +void xe_guc_log_snapshot_free(struct xe_guc_log_snapshot *snapshot); static inline u32 xe_guc_log_get_level(struct xe_guc_log *log) diff --git a/drivers/gpu/drm/xe/xe_guc_log_types.h b/drivers/gpu/drm/xe/xe_guc_log_types.h index 125080d138a7..962b9edbd9eb 100644 --- a/drivers/gpu/drm/xe/xe_guc_log_types.h +++ b/drivers/gpu/drm/xe/xe_guc_log_types.h @@ -8,8 +8,35 @@ #include +#include "xe_uc_fw_types.h" + struct xe_bo; +/** + * struct xe_guc_log_snapshot: + * Capture of the GuC log plus various state useful for decoding the log + */ +struct xe_guc_log_snapshot { + /** @size: Size in bytes of the @copy allocation */ + size_t size; + /** @copy: Host memory copy of the log buffer for later dumping, split into chunks */ + void **copy; + /** @num_chunks: Number of chunks within @copy */ + int num_chunks; + /** @ktime: Kernel time the snapshot was taken */ + u64 ktime; + /** @stamp: GuC timestamp at which the snapshot was taken */ + u32 stamp; + /** @level: GuC log verbosity level */ + u32 level; + /** @ver_found: GuC firmware version */ + struct xe_uc_fw_version ver_found; + /** @ver_want: GuC firmware version that driver expected */ + struct xe_uc_fw_version ver_want; + /** @path: Path of GuC firmware blob */ + const char *path; +}; + /** * struct xe_guc_log - GuC log */ -- 2.50.1 From 754e707e20e47482384f6e64eb3af08273292010 Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Wed, 2 Oct 2024 17:46:07 -0700 Subject: [PATCH 06/16] drm/print: Introduce drm_line_printer This drm printer wrapper can be used to increase the robustness of the captured output generated by any other drm_printer to make sure we didn't lost any intermediate lines of the output by adding line numbers to each output line. Helpful for capturing some crash data. v2: Extended short int counters to full int (JohnH) Signed-off-by: Michal Wajdeczko Signed-off-by: John Harrison Cc: Jani Nikula Cc: dri-devel@lists.freedesktop.org Reviewed-by: Jani Nikula Acked-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-8-John.C.Harrison@Intel.com --- drivers/gpu/drm/drm_print.c | 14 ++++++++ include/drm/drm_print.h | 64 +++++++++++++++++++++++++++++++++++++ 2 files changed, 78 insertions(+) diff --git a/drivers/gpu/drm/drm_print.c b/drivers/gpu/drm/drm_print.c index 0081190201a7..08cfea04e22b 100644 --- a/drivers/gpu/drm/drm_print.c +++ b/drivers/gpu/drm/drm_print.c @@ -235,6 +235,20 @@ void __drm_printfn_err(struct drm_printer *p, struct va_format *vaf) } EXPORT_SYMBOL(__drm_printfn_err); +void __drm_printfn_line(struct drm_printer *p, struct va_format *vaf) +{ + unsigned int counter = ++p->line.counter; + const char *prefix = p->prefix ?: ""; + const char *pad = p->prefix ? " " : ""; + + if (p->line.series) + drm_printf(p->arg, "%s%s%u.%u: %pV", + prefix, pad, p->line.series, counter, vaf); + else + drm_printf(p->arg, "%s%s%u: %pV", prefix, pad, counter, vaf); +} +EXPORT_SYMBOL(__drm_printfn_line); + /** * drm_puts - print a const string to a &drm_printer stream * @p: the &drm printer diff --git a/include/drm/drm_print.h b/include/drm/drm_print.h index d2676831d765..b3906dc04388 100644 --- a/include/drm/drm_print.h +++ b/include/drm/drm_print.h @@ -177,6 +177,10 @@ struct drm_printer { void *arg; const void *origin; const char *prefix; + struct { + unsigned int series; + unsigned int counter; + } line; enum drm_debug_category category; }; @@ -187,6 +191,7 @@ void __drm_puts_seq_file(struct drm_printer *p, const char *str); void __drm_printfn_info(struct drm_printer *p, struct va_format *vaf); void __drm_printfn_dbg(struct drm_printer *p, struct va_format *vaf); void __drm_printfn_err(struct drm_printer *p, struct va_format *vaf); +void __drm_printfn_line(struct drm_printer *p, struct va_format *vaf); __printf(2, 3) void drm_printf(struct drm_printer *p, const char *f, ...); @@ -411,6 +416,65 @@ static inline struct drm_printer drm_err_printer(struct drm_device *drm, return p; } +/** + * drm_line_printer - construct a &drm_printer that prefixes outputs with line numbers + * @p: the &struct drm_printer which actually generates the output + * @prefix: optional output prefix, or NULL for no prefix + * @series: optional unique series identifier, or 0 to omit identifier in the output + * + * This printer can be used to increase the robustness of the captured output + * to make sure we didn't lost any intermediate lines of the output. Helpful + * while capturing some crash data. + * + * Example 1:: + * + * void crash_dump(struct drm_device *drm) + * { + * static unsigned int id; + * struct drm_printer p = drm_err_printer(drm, "crash"); + * struct drm_printer lp = drm_line_printer(&p, "dump", ++id); + * + * drm_printf(&lp, "foo"); + * drm_printf(&lp, "bar"); + * } + * + * Above code will print into the dmesg something like:: + * + * [ ] 0000:00:00.0: [drm] *ERROR* crash dump 1.1: foo + * [ ] 0000:00:00.0: [drm] *ERROR* crash dump 1.2: bar + * + * Example 2:: + * + * void line_dump(struct device *dev) + * { + * struct drm_printer p = drm_info_printer(dev); + * struct drm_printer lp = drm_line_printer(&p, NULL, 0); + * + * drm_printf(&lp, "foo"); + * drm_printf(&lp, "bar"); + * } + * + * Above code will print:: + * + * [ ] 0000:00:00.0: [drm] 1: foo + * [ ] 0000:00:00.0: [drm] 2: bar + * + * RETURNS: + * The &drm_printer object + */ +static inline struct drm_printer drm_line_printer(struct drm_printer *p, + const char *prefix, + unsigned int series) +{ + struct drm_printer lp = { + .printfn = __drm_printfn_line, + .arg = p, + .prefix = prefix, + .line = { .series = series, }, + }; + return lp; +} + /* * struct device based logging * -- 2.50.1 From d2c5a5a926f43b2e42c5c955f917bad8ad6dd68c Mon Sep 17 00:00:00 2001 From: John Harrison Date: Wed, 2 Oct 2024 17:46:08 -0700 Subject: [PATCH 07/16] drm/xe/guc: Dead CT helper Add a worker function helper for asynchronously dumping state when an internal/fatal error is detected in CT processing. Being asynchronous is required to avoid deadlocks and scheduling-while-atomic or process-stalled-for-too-long issues. Also check for a bunch more error conditions and improve the handling of some existing checks. v2: Use compile time CONFIG check for new (but not directly CT_DEAD related) checks and use unsigned int for a bitmask, rename CT_DEAD_RESET to CT_DEAD_REARM and add some explaining comments, rename 'hxg' macro parameter to 'ctb' - review feedback from Michal W. Drop CT_DEAD_ALIVE as no need for a bitfield define to just set the entire mask to zero. v3: Fix kerneldoc v4: Nullify some floating pointers after free. v5: Add section headings and device info to make the state dump look more like a devcoredump to allow parsing by the same tools (eventual aim is to just call the devcoredump code itself, but that currently requires an xe_sched_job, which is not available in the CT code). v6: Fix potential for leaking snapshots with concurrent error conditions (review feedback from Julia F). v7: Don't complain about unexpected G2H messages yet because there is a known issue causing them. Fix bit shift bug with v6 change. Add GT id to fake coredump headers and use puts instead of printf. v8: Disable the head mis-match check in g2h_read because it is failing on various discrete platforms due to unknown reasons. Signed-off-by: John Harrison Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-9-John.C.Harrison@Intel.com --- .../drm/xe/abi/guc_communication_ctb_abi.h | 1 + drivers/gpu/drm/xe/xe_guc.c | 2 +- drivers/gpu/drm/xe/xe_guc_ct.c | 336 ++++++++++++++++-- drivers/gpu/drm/xe/xe_guc_ct.h | 2 +- drivers/gpu/drm/xe/xe_guc_ct_types.h | 23 ++ 5 files changed, 335 insertions(+), 29 deletions(-) diff --git a/drivers/gpu/drm/xe/abi/guc_communication_ctb_abi.h b/drivers/gpu/drm/xe/abi/guc_communication_ctb_abi.h index 8f86a16dc577..f58198cf2cf6 100644 --- a/drivers/gpu/drm/xe/abi/guc_communication_ctb_abi.h +++ b/drivers/gpu/drm/xe/abi/guc_communication_ctb_abi.h @@ -52,6 +52,7 @@ struct guc_ct_buffer_desc { #define GUC_CTB_STATUS_OVERFLOW (1 << 0) #define GUC_CTB_STATUS_UNDERFLOW (1 << 1) #define GUC_CTB_STATUS_MISMATCH (1 << 2) +#define GUC_CTB_STATUS_DISABLED (1 << 3) u32 reserved[13]; } __packed; static_assert(sizeof(struct guc_ct_buffer_desc) == 64); diff --git a/drivers/gpu/drm/xe/xe_guc.c b/drivers/gpu/drm/xe/xe_guc.c index c2ddf883702b..4abc790ed1cd 100644 --- a/drivers/gpu/drm/xe/xe_guc.c +++ b/drivers/gpu/drm/xe/xe_guc.c @@ -1180,7 +1180,7 @@ void xe_guc_print_info(struct xe_guc *guc, struct drm_printer *p) xe_force_wake_put(gt_to_fw(gt), XE_FW_GT); - xe_guc_ct_print(&guc->ct, p, false); + xe_guc_ct_print(&guc->ct, p); xe_guc_submit_print(guc, p); } diff --git a/drivers/gpu/drm/xe/xe_guc_ct.c b/drivers/gpu/drm/xe/xe_guc_ct.c index c0dded21621f..79571c9d180e 100644 --- a/drivers/gpu/drm/xe/xe_guc_ct.c +++ b/drivers/gpu/drm/xe/xe_guc_ct.c @@ -26,12 +26,48 @@ #include "xe_gt_sriov_pf_monitor.h" #include "xe_gt_tlb_invalidation.h" #include "xe_guc.h" +#include "xe_guc_log.h" #include "xe_guc_relay.h" #include "xe_guc_submit.h" #include "xe_map.h" #include "xe_pm.h" #include "xe_trace_guc.h" +#if IS_ENABLED(CONFIG_DRM_XE_DEBUG) +enum { + /* Internal states, not error conditions */ + CT_DEAD_STATE_REARM, /* 0x0001 */ + CT_DEAD_STATE_CAPTURE, /* 0x0002 */ + + /* Error conditions */ + CT_DEAD_SETUP, /* 0x0004 */ + CT_DEAD_H2G_WRITE, /* 0x0008 */ + CT_DEAD_H2G_HAS_ROOM, /* 0x0010 */ + CT_DEAD_G2H_READ, /* 0x0020 */ + CT_DEAD_G2H_RECV, /* 0x0040 */ + CT_DEAD_G2H_RELEASE, /* 0x0080 */ + CT_DEAD_DEADLOCK, /* 0x0100 */ + CT_DEAD_PROCESS_FAILED, /* 0x0200 */ + CT_DEAD_FAST_G2H, /* 0x0400 */ + CT_DEAD_PARSE_G2H_RESPONSE, /* 0x0800 */ + CT_DEAD_PARSE_G2H_UNKNOWN, /* 0x1000 */ + CT_DEAD_PARSE_G2H_ORIGIN, /* 0x2000 */ + CT_DEAD_PARSE_G2H_TYPE, /* 0x4000 */ +}; + +static void ct_dead_worker_func(struct work_struct *w); +static void ct_dead_capture(struct xe_guc_ct *ct, struct guc_ctb *ctb, u32 reason_code); + +#define CT_DEAD(ct, ctb, reason_code) ct_dead_capture((ct), (ctb), CT_DEAD_##reason_code) +#else +#define CT_DEAD(ct, ctb, reason) \ + do { \ + struct guc_ctb *_ctb = (ctb); \ + if (_ctb) \ + _ctb->info.broken = true; \ + } while (0) +#endif + /* Used when a CT send wants to block and / or receive data */ struct g2h_fence { u32 *response_buffer; @@ -184,6 +220,10 @@ int xe_guc_ct_init(struct xe_guc_ct *ct) xa_init(&ct->fence_lookup); INIT_WORK(&ct->g2h_worker, g2h_worker_func); INIT_DELAYED_WORK(&ct->safe_mode_worker, safe_mode_worker_func); +#if IS_ENABLED(CONFIG_DRM_XE_DEBUG) + spin_lock_init(&ct->dead.lock); + INIT_WORK(&ct->dead.worker, ct_dead_worker_func); +#endif init_waitqueue_head(&ct->wq); init_waitqueue_head(&ct->g2h_fence_wq); @@ -421,10 +461,22 @@ int xe_guc_ct_enable(struct xe_guc_ct *ct) if (ct_needs_safe_mode(ct)) ct_enter_safe_mode(ct); +#if IS_ENABLED(CONFIG_DRM_XE_DEBUG) + /* + * The CT has now been reset so the dumper can be re-armed + * after any existing dead state has been dumped. + */ + spin_lock_irq(&ct->dead.lock); + if (ct->dead.reason) + ct->dead.reason |= (1 << CT_DEAD_STATE_REARM); + spin_unlock_irq(&ct->dead.lock); +#endif + return 0; err_out: xe_gt_err(gt, "Failed to enable GuC CT (%pe)\n", ERR_PTR(err)); + CT_DEAD(ct, NULL, SETUP); return err; } @@ -468,6 +520,19 @@ static bool h2g_has_room(struct xe_guc_ct *ct, u32 cmd_len) if (cmd_len > h2g->info.space) { h2g->info.head = desc_read(ct_to_xe(ct), h2g, head); + + if (h2g->info.head > h2g->info.size) { + struct xe_device *xe = ct_to_xe(ct); + u32 desc_status = desc_read(xe, h2g, status); + + desc_write(xe, h2g, status, desc_status | GUC_CTB_STATUS_OVERFLOW); + + xe_gt_err(ct_to_gt(ct), "CT: invalid head offset %u >= %u)\n", + h2g->info.head, h2g->info.size); + CT_DEAD(ct, h2g, H2G_HAS_ROOM); + return false; + } + h2g->info.space = CIRC_SPACE(h2g->info.tail, h2g->info.head, h2g->info.size) - h2g->info.resv_space; @@ -523,10 +588,24 @@ static void __g2h_reserve_space(struct xe_guc_ct *ct, u32 g2h_len, u32 num_g2h) static void __g2h_release_space(struct xe_guc_ct *ct, u32 g2h_len) { + bool bad = false; + lockdep_assert_held(&ct->fast_lock); - xe_gt_assert(ct_to_gt(ct), ct->ctbs.g2h.info.space + g2h_len <= - ct->ctbs.g2h.info.size - ct->ctbs.g2h.info.resv_space); - xe_gt_assert(ct_to_gt(ct), ct->g2h_outstanding); + + bad = ct->ctbs.g2h.info.space + g2h_len > + ct->ctbs.g2h.info.size - ct->ctbs.g2h.info.resv_space; + bad |= !ct->g2h_outstanding; + + if (bad) { + xe_gt_err(ct_to_gt(ct), "Invalid G2H release: %d + %d vs %d - %d -> %d vs %d, outstanding = %d!\n", + ct->ctbs.g2h.info.space, g2h_len, + ct->ctbs.g2h.info.size, ct->ctbs.g2h.info.resv_space, + ct->ctbs.g2h.info.space + g2h_len, + ct->ctbs.g2h.info.size - ct->ctbs.g2h.info.resv_space, + ct->g2h_outstanding); + CT_DEAD(ct, &ct->ctbs.g2h, G2H_RELEASE); + return; + } ct->ctbs.g2h.info.space += g2h_len; if (!--ct->g2h_outstanding) @@ -553,12 +632,43 @@ static int h2g_write(struct xe_guc_ct *ct, const u32 *action, u32 len, u32 full_len; struct iosys_map map = IOSYS_MAP_INIT_OFFSET(&h2g->cmds, tail * sizeof(u32)); + u32 desc_status; full_len = len + GUC_CTB_HDR_LEN; lockdep_assert_held(&ct->lock); xe_gt_assert(gt, full_len <= GUC_CTB_MSG_MAX_LEN); - xe_gt_assert(gt, tail <= h2g->info.size); + + desc_status = desc_read(xe, h2g, status); + if (desc_status) { + xe_gt_err(gt, "CT write: non-zero status: %u\n", desc_status); + goto corrupted; + } + + if (IS_ENABLED(CONFIG_DRM_XE_DEBUG)) { + u32 desc_tail = desc_read(xe, h2g, tail); + u32 desc_head = desc_read(xe, h2g, head); + + if (tail != desc_tail) { + desc_write(xe, h2g, status, desc_status | GUC_CTB_STATUS_MISMATCH); + xe_gt_err(gt, "CT write: tail was modified %u != %u\n", desc_tail, tail); + goto corrupted; + } + + if (tail > h2g->info.size) { + desc_write(xe, h2g, status, desc_status | GUC_CTB_STATUS_OVERFLOW); + xe_gt_err(gt, "CT write: tail out of range: %u vs %u\n", + tail, h2g->info.size); + goto corrupted; + } + + if (desc_head >= h2g->info.size) { + desc_write(xe, h2g, status, desc_status | GUC_CTB_STATUS_OVERFLOW); + xe_gt_err(gt, "CT write: invalid head offset %u >= %u)\n", + desc_head, h2g->info.size); + goto corrupted; + } + } /* Command will wrap, zero fill (NOPs), return and check credits again */ if (tail + full_len > h2g->info.size) { @@ -611,6 +721,10 @@ static int h2g_write(struct xe_guc_ct *ct, const u32 *action, u32 len, desc_read(xe, h2g, head), h2g->info.tail); return 0; + +corrupted: + CT_DEAD(ct, &ct->ctbs.h2g, H2G_WRITE); + return -EPIPE; } /* @@ -718,7 +832,6 @@ static int guc_ct_send_locked(struct xe_guc_ct *ct, const u32 *action, u32 len, { struct xe_device *xe = ct_to_xe(ct); struct xe_gt *gt = ct_to_gt(ct); - struct drm_printer p = xe_gt_info_printer(gt); unsigned int sleep_period_ms = 1; int ret; @@ -771,8 +884,13 @@ try_again: goto broken; #undef g2h_avail - if (dequeue_one_g2h(ct) < 0) + ret = dequeue_one_g2h(ct); + if (ret < 0) { + if (ret != -ECANCELED) + xe_gt_err(ct_to_gt(ct), "CTB receive failed (%pe)", + ERR_PTR(ret)); goto broken; + } goto try_again; } @@ -781,8 +899,7 @@ try_again: broken: xe_gt_err(gt, "No forward process on H2G, reset required\n"); - xe_guc_ct_print(ct, &p, true); - ct->ctbs.h2g.info.broken = true; + CT_DEAD(ct, &ct->ctbs.h2g, DEADLOCK); return -EDEADLK; } @@ -1021,6 +1138,7 @@ static int parse_g2h_response(struct xe_guc_ct *ct, u32 *msg, u32 len) else xe_gt_err(gt, "unexpected response %u for FAST_REQ H2G fence 0x%x!\n", type, fence); + CT_DEAD(ct, NULL, PARSE_G2H_RESPONSE); return -EPROTO; } @@ -1028,6 +1146,7 @@ static int parse_g2h_response(struct xe_guc_ct *ct, u32 *msg, u32 len) g2h_fence = xa_erase(&ct->fence_lookup, fence); if (unlikely(!g2h_fence)) { /* Don't tear down channel, as send could've timed out */ + /* CT_DEAD(ct, NULL, PARSE_G2H_UNKNOWN); */ xe_gt_warn(gt, "G2H fence (%u) not found!\n", fence); g2h_release_space(ct, GUC_CTB_HXG_MSG_MAX_LEN); return 0; @@ -1072,7 +1191,7 @@ static int parse_g2h_msg(struct xe_guc_ct *ct, u32 *msg, u32 len) if (unlikely(origin != GUC_HXG_ORIGIN_GUC)) { xe_gt_err(gt, "G2H channel broken on read, origin=%u, reset required\n", origin); - ct->ctbs.g2h.info.broken = true; + CT_DEAD(ct, &ct->ctbs.g2h, PARSE_G2H_ORIGIN); return -EPROTO; } @@ -1090,7 +1209,7 @@ static int parse_g2h_msg(struct xe_guc_ct *ct, u32 *msg, u32 len) default: xe_gt_err(gt, "G2H channel broken on read, type=%u, reset required\n", type); - ct->ctbs.g2h.info.broken = true; + CT_DEAD(ct, &ct->ctbs.g2h, PARSE_G2H_TYPE); ret = -EOPNOTSUPP; } @@ -1167,9 +1286,11 @@ static int process_g2h_msg(struct xe_guc_ct *ct, u32 *msg, u32 len) xe_gt_err(gt, "unexpected G2H action 0x%04x\n", action); } - if (ret) + if (ret) { xe_gt_err(gt, "G2H action 0x%04x failed (%pe)\n", action, ERR_PTR(ret)); + CT_DEAD(ct, NULL, PROCESS_FAILED); + } return 0; } @@ -1179,7 +1300,7 @@ static int g2h_read(struct xe_guc_ct *ct, u32 *msg, bool fast_path) struct xe_device *xe = ct_to_xe(ct); struct xe_gt *gt = ct_to_gt(ct); struct guc_ctb *g2h = &ct->ctbs.g2h; - u32 tail, head, len; + u32 tail, head, len, desc_status; s32 avail; u32 action; u32 *hxg; @@ -1198,6 +1319,63 @@ static int g2h_read(struct xe_guc_ct *ct, u32 *msg, bool fast_path) xe_gt_assert(gt, xe_guc_ct_enabled(ct)); + desc_status = desc_read(xe, g2h, status); + if (desc_status) { + if (desc_status & GUC_CTB_STATUS_DISABLED) { + /* + * Potentially valid if a CLIENT_RESET request resulted in + * contexts/engines being reset. But should never happen as + * no contexts should be active when CLIENT_RESET is sent. + */ + xe_gt_err(gt, "CT read: unexpected G2H after GuC has stopped!\n"); + desc_status &= ~GUC_CTB_STATUS_DISABLED; + } + + if (desc_status) { + xe_gt_err(gt, "CT read: non-zero status: %u\n", desc_status); + goto corrupted; + } + } + + if (IS_ENABLED(CONFIG_DRM_XE_DEBUG)) { + u32 desc_tail = desc_read(xe, g2h, tail); + /* + u32 desc_head = desc_read(xe, g2h, head); + + * info.head and desc_head are updated back-to-back at the end of + * this function and nowhere else. Hence, they cannot be different + * unless two g2h_read calls are running concurrently. Which is not + * possible because it is guarded by ct->fast_lock. And yet, some + * discrete platforms are reguarly hitting this error :(. + * + * desc_head rolling backwards shouldn't cause any noticeable + * problems - just a delay in GuC being allowed to proceed past that + * point in the queue. So for now, just disable the error until it + * can be root caused. + * + if (g2h->info.head != desc_head) { + desc_write(xe, g2h, status, desc_status | GUC_CTB_STATUS_MISMATCH); + xe_gt_err(gt, "CT read: head was modified %u != %u\n", + desc_head, g2h->info.head); + goto corrupted; + } + */ + + if (g2h->info.head > g2h->info.size) { + desc_write(xe, g2h, status, desc_status | GUC_CTB_STATUS_OVERFLOW); + xe_gt_err(gt, "CT read: head out of range: %u vs %u\n", + g2h->info.head, g2h->info.size); + goto corrupted; + } + + if (desc_tail >= g2h->info.size) { + desc_write(xe, g2h, status, desc_status | GUC_CTB_STATUS_OVERFLOW); + xe_gt_err(gt, "CT read: invalid tail offset %u >= %u)\n", + desc_tail, g2h->info.size); + goto corrupted; + } + } + /* Calculate DW available to read */ tail = desc_read(xe, g2h, tail); avail = tail - g2h->info.head; @@ -1214,9 +1392,7 @@ static int g2h_read(struct xe_guc_ct *ct, u32 *msg, bool fast_path) if (len > avail) { xe_gt_err(gt, "G2H channel broken on read, avail=%d, len=%d, reset required\n", avail, len); - g2h->info.broken = true; - - return -EPROTO; + goto corrupted; } head = (g2h->info.head + 1) % g2h->info.size; @@ -1262,6 +1438,10 @@ static int g2h_read(struct xe_guc_ct *ct, u32 *msg, bool fast_path) action, len, g2h->info.head, tail); return len; + +corrupted: + CT_DEAD(ct, &ct->ctbs.g2h, G2H_READ); + return -EPROTO; } static void g2h_fast_path(struct xe_guc_ct *ct, u32 *msg, u32 len) @@ -1288,9 +1468,11 @@ static void g2h_fast_path(struct xe_guc_ct *ct, u32 *msg, u32 len) xe_gt_warn(gt, "NOT_POSSIBLE"); } - if (ret) + if (ret) { xe_gt_err(gt, "G2H action 0x%04x failed (%pe)\n", action, ERR_PTR(ret)); + CT_DEAD(ct, NULL, FAST_G2H); + } } /** @@ -1350,7 +1532,6 @@ static int dequeue_one_g2h(struct xe_guc_ct *ct) static void receive_g2h(struct xe_guc_ct *ct) { - struct xe_gt *gt = ct_to_gt(ct); bool ongoing; int ret; @@ -1387,9 +1568,8 @@ static void receive_g2h(struct xe_guc_ct *ct) mutex_unlock(&ct->lock); if (unlikely(ret == -EPROTO || ret == -EOPNOTSUPP)) { - struct drm_printer p = xe_gt_info_printer(gt); - - xe_guc_ct_print(ct, &p, false); + xe_gt_err(ct_to_gt(ct), "CT dequeue failed: %d", ret); + CT_DEAD(ct, NULL, G2H_RECV); kick_reset(ct); } } while (ret == 1); @@ -1417,9 +1597,8 @@ static void guc_ctb_snapshot_capture(struct xe_device *xe, struct guc_ctb *ctb, snapshot->cmds = kmalloc_array(ctb->info.size, sizeof(u32), atomic ? GFP_ATOMIC : GFP_KERNEL); - if (!snapshot->cmds) { - drm_err(&xe->drm, "Skipping CTB commands snapshot. Only CTB info will be available.\n"); + drm_err(&xe->drm, "Skipping CTB commands snapshot. Only CT info will be available.\n"); return; } @@ -1500,7 +1679,7 @@ struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct, atomic ? GFP_ATOMIC : GFP_KERNEL); if (!snapshot) { - drm_err(&xe->drm, "Skipping CTB snapshot entirely.\n"); + xe_gt_err(ct_to_gt(ct), "Skipping CTB snapshot entirely.\n"); return NULL; } @@ -1564,16 +1743,119 @@ void xe_guc_ct_snapshot_free(struct xe_guc_ct_snapshot *snapshot) * xe_guc_ct_print - GuC CT Print. * @ct: GuC CT. * @p: drm_printer where it will be printed out. - * @atomic: Boolean to indicate if this is called from atomic context like - * reset or CTB handler or from some regular path like debugfs. * * This function quickly capture a snapshot and immediately print it out. */ -void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p, bool atomic) +void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p) { struct xe_guc_ct_snapshot *snapshot; - snapshot = xe_guc_ct_snapshot_capture(ct, atomic); + snapshot = xe_guc_ct_snapshot_capture(ct, false); xe_guc_ct_snapshot_print(snapshot, p); xe_guc_ct_snapshot_free(snapshot); } + +#if IS_ENABLED(CONFIG_DRM_XE_DEBUG) +static void ct_dead_capture(struct xe_guc_ct *ct, struct guc_ctb *ctb, u32 reason_code) +{ + struct xe_guc_log_snapshot *snapshot_log; + struct xe_guc_ct_snapshot *snapshot_ct; + struct xe_guc *guc = ct_to_guc(ct); + unsigned long flags; + bool have_capture; + + if (ctb) + ctb->info.broken = true; + + /* Ignore further errors after the first dump until a reset */ + if (ct->dead.reported) + return; + + spin_lock_irqsave(&ct->dead.lock, flags); + + /* And only capture one dump at a time */ + have_capture = ct->dead.reason & (1 << CT_DEAD_STATE_CAPTURE); + ct->dead.reason |= (1 << reason_code) | + (1 << CT_DEAD_STATE_CAPTURE); + + spin_unlock_irqrestore(&ct->dead.lock, flags); + + if (have_capture) + return; + + snapshot_log = xe_guc_log_snapshot_capture(&guc->log, true); + snapshot_ct = xe_guc_ct_snapshot_capture((ct), true); + + spin_lock_irqsave(&ct->dead.lock, flags); + + if (ct->dead.snapshot_log || ct->dead.snapshot_ct) { + xe_gt_err(ct_to_gt(ct), "Got unexpected dead CT capture!\n"); + xe_guc_log_snapshot_free(snapshot_log); + xe_guc_ct_snapshot_free(snapshot_ct); + } else { + ct->dead.snapshot_log = snapshot_log; + ct->dead.snapshot_ct = snapshot_ct; + } + + spin_unlock_irqrestore(&ct->dead.lock, flags); + + queue_work(system_unbound_wq, &(ct)->dead.worker); +} + +static void ct_dead_print(struct xe_dead_ct *dead) +{ + struct xe_guc_ct *ct = container_of(dead, struct xe_guc_ct, dead); + struct xe_device *xe = ct_to_xe(ct); + struct xe_gt *gt = ct_to_gt(ct); + static int g_count; + struct drm_printer ip = xe_gt_info_printer(gt); + struct drm_printer lp = drm_line_printer(&ip, "Capture", ++g_count); + + if (!dead->reason) { + xe_gt_err(gt, "CTB is dead for no reason!?\n"); + return; + } + + drm_printf(&lp, "CTB is dead - reason=0x%X\n", dead->reason); + + /* Can't generate a genuine core dump at this point, so just do the good bits */ + drm_puts(&lp, "**** Xe Device Coredump ****\n"); + xe_device_snapshot_print(xe, &lp); + + drm_printf(&lp, "**** GT #%d ****\n", gt->info.id); + drm_printf(&lp, "\tTile: %d\n", gt->tile->id); + + drm_puts(&lp, "**** GuC Log ****\n"); + xe_guc_log_snapshot_print(dead->snapshot_log, &lp); + + drm_puts(&lp, "**** GuC CT ****\n"); + xe_guc_ct_snapshot_print(dead->snapshot_ct, &lp); + + drm_puts(&lp, "Done.\n"); +} + +static void ct_dead_worker_func(struct work_struct *w) +{ + struct xe_guc_ct *ct = container_of(w, struct xe_guc_ct, dead.worker); + + if (!ct->dead.reported) { + ct->dead.reported = true; + ct_dead_print(&ct->dead); + } + + spin_lock_irq(&ct->dead.lock); + + xe_guc_log_snapshot_free(ct->dead.snapshot_log); + ct->dead.snapshot_log = NULL; + xe_guc_ct_snapshot_free(ct->dead.snapshot_ct); + ct->dead.snapshot_ct = NULL; + + if (ct->dead.reason & (1 << CT_DEAD_STATE_REARM)) { + /* A reset has occurred so re-arm the error reporting */ + ct->dead.reason = 0; + ct->dead.reported = false; + } + + spin_unlock_irq(&ct->dead.lock); +} +#endif diff --git a/drivers/gpu/drm/xe/xe_guc_ct.h b/drivers/gpu/drm/xe/xe_guc_ct.h index 190202fce2d0..293041bed7ed 100644 --- a/drivers/gpu/drm/xe/xe_guc_ct.h +++ b/drivers/gpu/drm/xe/xe_guc_ct.h @@ -21,7 +21,7 @@ xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct, bool atomic); void xe_guc_ct_snapshot_print(struct xe_guc_ct_snapshot *snapshot, struct drm_printer *p); void xe_guc_ct_snapshot_free(struct xe_guc_ct_snapshot *snapshot); -void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p, bool atomic); +void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p); static inline bool xe_guc_ct_enabled(struct xe_guc_ct *ct) { diff --git a/drivers/gpu/drm/xe/xe_guc_ct_types.h b/drivers/gpu/drm/xe/xe_guc_ct_types.h index 761cb9031298..85e127ec91d7 100644 --- a/drivers/gpu/drm/xe/xe_guc_ct_types.h +++ b/drivers/gpu/drm/xe/xe_guc_ct_types.h @@ -86,6 +86,24 @@ enum xe_guc_ct_state { XE_GUC_CT_STATE_ENABLED, }; +#if IS_ENABLED(CONFIG_DRM_XE_DEBUG) +/** struct xe_dead_ct - Information for debugging a dead CT */ +struct xe_dead_ct { + /** @lock: protects memory allocation/free operations, and @reason updates */ + spinlock_t lock; + /** @reason: bit mask of CT_DEAD_* reason codes */ + unsigned int reason; + /** @reported: for preventing multiple dumps per error sequence */ + bool reported; + /** @worker: worker thread to get out of interrupt context before dumping */ + struct work_struct worker; + /** snapshot_ct: copy of CT state and CTB content at point of error */ + struct xe_guc_ct_snapshot *snapshot_ct; + /** snapshot_log: copy of GuC log at point of error */ + struct xe_guc_log_snapshot *snapshot_log; +}; +#endif + /** * struct xe_guc_ct - GuC command transport (CT) layer * @@ -128,6 +146,11 @@ struct xe_guc_ct { u32 msg[GUC_CTB_MSG_MAX_LEN]; /** @fast_msg: Message buffer */ u32 fast_msg[GUC_CTB_MSG_MAX_LEN]; + +#if IS_ENABLED(CONFIG_DRM_XE_DEBUG) + /** @dead: information for debugging dead CTs */ + struct xe_dead_ct dead; +#endif }; #endif -- 2.50.1 From d7c925b2995f5cae3ad405f8988ad821d0dd0287 Mon Sep 17 00:00:00 2001 From: John Harrison Date: Wed, 2 Oct 2024 17:46:09 -0700 Subject: [PATCH 08/16] drm/xe/guc: Dump entire CTB on errors The dump of the CT buffers was only showing the unprocessed data which is not generally useful for saying why a hang occurred - because it was probably caused by the commands that were just processed. So save and dump the entire buffer but in a more compact dump format. Also zero fill it on allocation to avoid confusion over uninitialised data in the dump. v2: Add kerneldoc - review feedback from Michal W. v3: Fix kerneldoc. v4: Use ascii85 instead of hexdump (review feedback from Matthew B). v5: Dump the entire CTB object rather than separately dumping just the H2G and G2H sections. That way it includes the full header info. Signed-off-by: John Harrison Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-10-John.C.Harrison@Intel.com --- drivers/gpu/drm/xe/xe_guc_ct.c | 94 ++++++++++------------------ drivers/gpu/drm/xe/xe_guc_ct.h | 8 +-- drivers/gpu/drm/xe/xe_guc_ct_types.h | 6 +- 3 files changed, 41 insertions(+), 67 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_guc_ct.c b/drivers/gpu/drm/xe/xe_guc_ct.c index 79571c9d180e..b7f80449bd83 100644 --- a/drivers/gpu/drm/xe/xe_guc_ct.c +++ b/drivers/gpu/drm/xe/xe_guc_ct.c @@ -18,6 +18,7 @@ #include "abi/guc_actions_sriov_abi.h" #include "abi/guc_klvs_abi.h" #include "xe_bo.h" +#include "xe_devcoredump.h" #include "xe_device.h" #include "xe_gt.h" #include "xe_gt_pagefault.h" @@ -437,6 +438,7 @@ int xe_guc_ct_enable(struct xe_guc_ct *ct) xe_gt_assert(gt, !xe_guc_ct_enabled(ct)); + xe_map_memset(xe, &ct->bo->vmap, 0, 0, ct->bo->size); guc_ct_ctb_h2g_init(xe, &ct->ctbs.h2g, &ct->bo->vmap); guc_ct_ctb_g2h_init(xe, &ct->ctbs.g2h, &ct->bo->vmap); @@ -1585,48 +1587,33 @@ static void g2h_worker_func(struct work_struct *w) receive_g2h(ct); } -static void guc_ctb_snapshot_capture(struct xe_device *xe, struct guc_ctb *ctb, - struct guc_ctb_snapshot *snapshot, - bool atomic) +struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_alloc(struct xe_guc_ct *ct, bool atomic) { - u32 head, tail; + struct xe_guc_ct_snapshot *snapshot; - xe_map_memcpy_from(xe, &snapshot->desc, &ctb->desc, 0, - sizeof(struct guc_ct_buffer_desc)); - memcpy(&snapshot->info, &ctb->info, sizeof(struct guc_ctb_info)); + snapshot = kzalloc(sizeof(*snapshot), atomic ? GFP_ATOMIC : GFP_KERNEL); + if (!snapshot) + return NULL; - snapshot->cmds = kmalloc_array(ctb->info.size, sizeof(u32), - atomic ? GFP_ATOMIC : GFP_KERNEL); - if (!snapshot->cmds) { - drm_err(&xe->drm, "Skipping CTB commands snapshot. Only CT info will be available.\n"); - return; + if (ct->bo) { + snapshot->ctb_size = ct->bo->size; + snapshot->ctb = kmalloc(snapshot->ctb_size, atomic ? GFP_ATOMIC : GFP_KERNEL); } - head = snapshot->desc.head; - tail = snapshot->desc.tail; - - if (head != tail) { - struct iosys_map map = - IOSYS_MAP_INIT_OFFSET(&ctb->cmds, head * sizeof(u32)); - - while (head != tail) { - snapshot->cmds[head] = xe_map_rd(xe, &map, 0, u32); - ++head; - if (head == ctb->info.size) { - head = 0; - map = ctb->cmds; - } else { - iosys_map_incr(&map, sizeof(u32)); - } - } - } + return snapshot; +} + +static void guc_ctb_snapshot_capture(struct xe_device *xe, struct guc_ctb *ctb, + struct guc_ctb_snapshot *snapshot) +{ + xe_map_memcpy_from(xe, &snapshot->desc, &ctb->desc, 0, + sizeof(struct guc_ct_buffer_desc)); + memcpy(&snapshot->info, &ctb->info, sizeof(struct guc_ctb_info)); } static void guc_ctb_snapshot_print(struct guc_ctb_snapshot *snapshot, struct drm_printer *p) { - u32 head, tail; - drm_printf(p, "\tsize: %d\n", snapshot->info.size); drm_printf(p, "\tresv_space: %d\n", snapshot->info.resv_space); drm_printf(p, "\thead: %d\n", snapshot->info.head); @@ -1636,25 +1623,6 @@ static void guc_ctb_snapshot_print(struct guc_ctb_snapshot *snapshot, drm_printf(p, "\thead (memory): %d\n", snapshot->desc.head); drm_printf(p, "\ttail (memory): %d\n", snapshot->desc.tail); drm_printf(p, "\tstatus (memory): 0x%x\n", snapshot->desc.status); - - if (!snapshot->cmds) - return; - - head = snapshot->desc.head; - tail = snapshot->desc.tail; - - while (head != tail) { - drm_printf(p, "\tcmd[%d]: 0x%08x\n", head, - snapshot->cmds[head]); - ++head; - if (head == snapshot->info.size) - head = 0; - } -} - -static void guc_ctb_snapshot_free(struct guc_ctb_snapshot *snapshot) -{ - kfree(snapshot->cmds); } /** @@ -1675,9 +1643,7 @@ struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct, struct xe_device *xe = ct_to_xe(ct); struct xe_guc_ct_snapshot *snapshot; - snapshot = kzalloc(sizeof(*snapshot), - atomic ? GFP_ATOMIC : GFP_KERNEL); - + snapshot = xe_guc_ct_snapshot_alloc(ct, atomic); if (!snapshot) { xe_gt_err(ct_to_gt(ct), "Skipping CTB snapshot entirely.\n"); return NULL; @@ -1686,12 +1652,13 @@ struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct, if (xe_guc_ct_enabled(ct) || ct->state == XE_GUC_CT_STATE_STOPPED) { snapshot->ct_enabled = true; snapshot->g2h_outstanding = READ_ONCE(ct->g2h_outstanding); - guc_ctb_snapshot_capture(xe, &ct->ctbs.h2g, - &snapshot->h2g, atomic); - guc_ctb_snapshot_capture(xe, &ct->ctbs.g2h, - &snapshot->g2h, atomic); + guc_ctb_snapshot_capture(xe, &ct->ctbs.h2g, &snapshot->h2g); + guc_ctb_snapshot_capture(xe, &ct->ctbs.g2h, &snapshot->g2h); } + if (ct->bo && snapshot->ctb) + xe_map_memcpy_from(xe, snapshot->ctb, &ct->bo->vmap, 0, snapshot->ctb_size); + return snapshot; } @@ -1714,9 +1681,15 @@ void xe_guc_ct_snapshot_print(struct xe_guc_ct_snapshot *snapshot, drm_puts(p, "G2H CTB (all sizes in DW):\n"); guc_ctb_snapshot_print(&snapshot->g2h, p); - drm_printf(p, "\tg2h outstanding: %d\n", snapshot->g2h_outstanding); + + if (snapshot->ctb) { + xe_print_blob_ascii85(p, "CTB data", snapshot->ctb, 0, snapshot->ctb_size); + } else { + drm_printf(p, "CTB snapshot missing!\n"); + return; + } } else { drm_puts(p, "CT disabled\n"); } @@ -1734,8 +1707,7 @@ void xe_guc_ct_snapshot_free(struct xe_guc_ct_snapshot *snapshot) if (!snapshot) return; - guc_ctb_snapshot_free(&snapshot->h2g); - guc_ctb_snapshot_free(&snapshot->g2h); + kfree(snapshot->ctb); kfree(snapshot); } diff --git a/drivers/gpu/drm/xe/xe_guc_ct.h b/drivers/gpu/drm/xe/xe_guc_ct.h index 293041bed7ed..338f0b75d29f 100644 --- a/drivers/gpu/drm/xe/xe_guc_ct.h +++ b/drivers/gpu/drm/xe/xe_guc_ct.h @@ -9,6 +9,7 @@ #include "xe_guc_ct_types.h" struct drm_printer; +struct xe_device; int xe_guc_ct_init(struct xe_guc_ct *ct); int xe_guc_ct_enable(struct xe_guc_ct *ct); @@ -16,10 +17,9 @@ void xe_guc_ct_disable(struct xe_guc_ct *ct); void xe_guc_ct_stop(struct xe_guc_ct *ct); void xe_guc_ct_fast_path(struct xe_guc_ct *ct); -struct xe_guc_ct_snapshot * -xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct, bool atomic); -void xe_guc_ct_snapshot_print(struct xe_guc_ct_snapshot *snapshot, - struct drm_printer *p); +struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_alloc(struct xe_guc_ct *ct, bool atomic); +struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct, bool atomic); +void xe_guc_ct_snapshot_print(struct xe_guc_ct_snapshot *snapshot, struct drm_printer *p); void xe_guc_ct_snapshot_free(struct xe_guc_ct_snapshot *snapshot); void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p); diff --git a/drivers/gpu/drm/xe/xe_guc_ct_types.h b/drivers/gpu/drm/xe/xe_guc_ct_types.h index 85e127ec91d7..8e1b9d981d61 100644 --- a/drivers/gpu/drm/xe/xe_guc_ct_types.h +++ b/drivers/gpu/drm/xe/xe_guc_ct_types.h @@ -52,8 +52,6 @@ struct guc_ctb { struct guc_ctb_snapshot { /** @desc: snapshot of the CTB descriptor */ struct guc_ct_buffer_desc desc; - /** @cmds: snapshot of the CTB commands */ - u32 *cmds; /** @info: snapshot of the CTB info */ struct guc_ctb_info info; }; @@ -70,6 +68,10 @@ struct xe_guc_ct_snapshot { struct guc_ctb_snapshot g2h; /** @h2g: H2G CTB snapshot */ struct guc_ctb_snapshot h2g; + /** @ctb_size: size of the snapshot of the CTB */ + size_t ctb_size; + /** @ctb: snapshot of the entire CTB */ + u32 *ctb; }; /** -- 2.50.1 From 8b7dfb985532c69729d3b266a68fa379cd359f55 Mon Sep 17 00:00:00 2001 From: John Harrison Date: Wed, 2 Oct 2024 17:46:10 -0700 Subject: [PATCH 09/16] drm/xe/guc: Add GuC log to devcoredump captures Include the GuC log in devcoredump captures because they can be useful with debugging certain types of bug. v2: Fix kerneldoc v3: Drop module parameter as now using more compact ascii85 encoding rather than hexdump (although still not compressed) (review feedback from Matthew B). Rebase onto recent refactoring of devcoredump code. v4: Don't move the submission snapshot inside the GuC internals structure 'cos it really doesn't belong there. Signed-off-by: John Harrison Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-11-John.C.Harrison@Intel.com --- drivers/gpu/drm/xe/xe_devcoredump.c | 16 ++++++++++++---- drivers/gpu/drm/xe/xe_devcoredump_types.h | 10 +++++++--- 2 files changed, 19 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_devcoredump.c b/drivers/gpu/drm/xe/xe_devcoredump.c index 0884c49942fe..a2d816713489 100644 --- a/drivers/gpu/drm/xe/xe_devcoredump.c +++ b/drivers/gpu/drm/xe/xe_devcoredump.c @@ -18,8 +18,10 @@ #include "xe_gt.h" #include "xe_gt_printk.h" #include "xe_guc_ct.h" +#include "xe_guc_log.h" #include "xe_guc_submit.h" #include "xe_hw_engine.h" +#include "xe_module.h" #include "xe_sched_job.h" #include "xe_vm.h" @@ -100,8 +102,10 @@ static ssize_t __xe_devcoredump_read(char *buffer, size_t count, drm_printf(&p, "\n**** GT #%d ****\n", ss->gt->info.id); drm_printf(&p, "\tTile: %d\n", ss->gt->tile->id); + drm_puts(&p, "\n**** GuC Log ****\n"); + xe_guc_log_snapshot_print(ss->guc.log, &p); drm_puts(&p, "\n**** GuC CT ****\n"); - xe_guc_ct_snapshot_print(ss->ct, &p); + xe_guc_ct_snapshot_print(ss->guc.ct, &p); drm_puts(&p, "\n**** Contexts ****\n"); xe_guc_exec_queue_snapshot_print(ss->ge, &p); @@ -124,8 +128,11 @@ static void xe_devcoredump_snapshot_free(struct xe_devcoredump_snapshot *ss) { int i; - xe_guc_ct_snapshot_free(ss->ct); - ss->ct = NULL; + xe_guc_log_snapshot_free(ss->guc.log); + ss->guc.log = NULL; + + xe_guc_ct_snapshot_free(ss->guc.ct); + ss->guc.ct = NULL; xe_guc_exec_queue_snapshot_free(ss->ge); ss->ge = NULL; @@ -253,7 +260,8 @@ static void devcoredump_snapshot(struct xe_devcoredump *coredump, if (xe_force_wake_get(gt_to_fw(q->gt), XE_FORCEWAKE_ALL)) xe_gt_info(ss->gt, "failed to get forcewake for coredump capture\n"); - ss->ct = xe_guc_ct_snapshot_capture(&guc->ct, true); + ss->guc.log = xe_guc_log_snapshot_capture(&guc->log, true); + ss->guc.ct = xe_guc_ct_snapshot_capture(&guc->ct, true); ss->ge = xe_guc_exec_queue_snapshot_capture(q); ss->job = xe_sched_job_snapshot_capture(job); ss->vm = xe_vm_snapshot_capture(q->vm); diff --git a/drivers/gpu/drm/xe/xe_devcoredump_types.h b/drivers/gpu/drm/xe/xe_devcoredump_types.h index 3cc2f095fdfb..06ac75ce63dd 100644 --- a/drivers/gpu/drm/xe/xe_devcoredump_types.h +++ b/drivers/gpu/drm/xe/xe_devcoredump_types.h @@ -34,9 +34,13 @@ struct xe_devcoredump_snapshot { /** @work: Workqueue for deferred capture outside of signaling context */ struct work_struct work; - /* GuC snapshots */ - /** @ct: GuC CT snapshot */ - struct xe_guc_ct_snapshot *ct; + /** @guc: GuC snapshots */ + struct { + /** @guc.ct: GuC CT snapshot */ + struct xe_guc_ct_snapshot *ct; + /** @guc.log: GuC log snapshot */ + struct xe_guc_log_snapshot *log; + } guc; /** @ge: GuC Submission Engine snapshot */ struct xe_guc_submit_exec_queue_snapshot *ge; -- 2.50.1 From 691b5a6af36022648a0aa23899515a466eccbb61 Mon Sep 17 00:00:00 2001 From: John Harrison Date: Wed, 2 Oct 2024 17:46:11 -0700 Subject: [PATCH 10/16] drm/xe/guc: Add a helper function for dumping GuC log to dmesg Create a helper function that can be used to dump the GuC log to dmesg in a manner that is reliable for extraction and decode. The intention is that calls to this can be added by developers when debugging specific issues that require a GuC log but do not allow easy capture of the log - e.g. failures in selftests and failues that lead to kernel hangs. Also note that this is really a temporary stop-gap. The aim is to allow on demand creation and dumping of devcoredump captures (which includes the GuC log and much more). Currently this is not possible as much of the devcoredump code requires a 'struct xe_sched_job' and those are not available at many places that might want to do the dump. v2: Add kerneldoc - review feedback from Michal W. Signed-off-by: John Harrison Reviewed-by: Julia Filipchuk Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-12-John.C.Harrison@Intel.com --- drivers/gpu/drm/xe/xe_guc_log.c | 18 ++++++++++++++++++ drivers/gpu/drm/xe/xe_guc_log.h | 1 + 2 files changed, 19 insertions(+) diff --git a/drivers/gpu/drm/xe/xe_guc_log.c b/drivers/gpu/drm/xe/xe_guc_log.c index 79e01d724509..7fbc56cceaba 100644 --- a/drivers/gpu/drm/xe/xe_guc_log.c +++ b/drivers/gpu/drm/xe/xe_guc_log.c @@ -216,6 +216,24 @@ void xe_guc_log_snapshot_print(struct xe_guc_log_snapshot *snapshot, struct drm_ } } +/** + * xe_guc_log_print_dmesg - dump a copy of the GuC log to dmesg + * @log: GuC log structure + */ +void xe_guc_log_print_dmesg(struct xe_guc_log *log) +{ + struct xe_gt *gt = log_to_gt(log); + static int g_count; + struct drm_printer ip = xe_gt_info_printer(gt); + struct drm_printer lp = drm_line_printer(&ip, "Capture", ++g_count); + + drm_printf(&lp, "Dumping GuC log for %ps...\n", __builtin_return_address(0)); + + xe_guc_log_print(log, &lp); + + drm_printf(&lp, "Done.\n"); +} + /** * xe_guc_log_print - dump a copy of the GuC log to some useful location * @log: GuC log structure diff --git a/drivers/gpu/drm/xe/xe_guc_log.h b/drivers/gpu/drm/xe/xe_guc_log.h index 949d2c98343d..1fb2fae1f4e1 100644 --- a/drivers/gpu/drm/xe/xe_guc_log.h +++ b/drivers/gpu/drm/xe/xe_guc_log.h @@ -39,6 +39,7 @@ struct xe_device; int xe_guc_log_init(struct xe_guc_log *log); void xe_guc_log_print(struct xe_guc_log *log, struct drm_printer *p); +void xe_guc_log_print_dmesg(struct xe_guc_log *log); struct xe_guc_log_snapshot *xe_guc_log_snapshot_capture(struct xe_guc_log *log, bool atomic); void xe_guc_log_snapshot_print(struct xe_guc_log_snapshot *snapshot, struct drm_printer *p); void xe_guc_log_snapshot_free(struct xe_guc_log_snapshot *snapshot); -- 2.50.1 From 9ab440a9d0426cf7842240891cc457155db1a97e Mon Sep 17 00:00:00 2001 From: Shekhar Chauhan Date: Mon, 7 Oct 2024 08:41:44 -0700 Subject: [PATCH 11/16] drm/xe/ptl: L3bank mask is not available on the media GT On PTL platforms with media version 30.00, the fuse registers for reporting L3 bank availability to the GT just read out as ~0 and do not provide proper values. Xe does not use the L3 bank mask for anything internally; it only passes the mask through to userspace via the GT topology query. Since we don't have any way to get the real L3 bank mask, we don't want to pass garbage to userspace. Passing a zeroed mask or a copy of the primary GT's L3 bank mask would also be inaccurate and likely to cause confusion for userspace. The best approach is to simply not include L3 in the list of masks returned by the topology query in cases where we aren't able to provide a meaningful value. This won't change the behavior for any existing platforms (where we can always obtain L3 masks successfully for all GTs), it will only prevent us from mis-reporting bad information on upcoming platform(s). There's a good chance this will become a formal workaround in the future, but for now we don't have a lineage number so "no_media_l3" is used in place of a lineage as the OOB workaround descriptor. v2: - Re-calculate query size to properly match data returned. (Gustavo) - Update kerneldoc to clarify that the L3bank mask may not be included in the query results if the hardware doesn't make it available. (Gustavo) Cc: Matt Atwood Cc: Gustavo Sousa Signed-off-by: Shekhar Chauhan Co-developed-by: Matt Roper Signed-off-by: Matt Roper Reviewed-by: Jonathan Cavitt Reviewed-by: Gustavo Sousa Acked-by: Francois Dugast Link: https://patchwork.freedesktop.org/patch/msgid/20241007154143.2021124-2-matthew.d.roper@intel.com --- drivers/gpu/drm/xe/xe_gt_topology.c | 14 ++++++++++ drivers/gpu/drm/xe/xe_query.c | 42 +++++++++++++++++++++-------- drivers/gpu/drm/xe/xe_wa_oob.rules | 1 + include/uapi/drm/xe_drm.h | 4 ++- 4 files changed, 49 insertions(+), 12 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_gt_topology.c b/drivers/gpu/drm/xe/xe_gt_topology.c index 651ba53623e5..df2042db7ee6 100644 --- a/drivers/gpu/drm/xe/xe_gt_topology.c +++ b/drivers/gpu/drm/xe/xe_gt_topology.c @@ -5,6 +5,7 @@ #include "xe_gt_topology.h" +#include #include #include @@ -12,6 +13,7 @@ #include "xe_assert.h" #include "xe_gt.h" #include "xe_mmio.h" +#include "xe_wa.h" static void load_dss_mask(struct xe_gt *gt, xe_dss_mask_t mask, int numregs, ...) @@ -129,6 +131,18 @@ load_l3_bank_mask(struct xe_gt *gt, xe_l3_bank_mask_t l3_bank_mask) struct xe_device *xe = gt_to_xe(gt); u32 fuse3 = xe_mmio_read32(>->mmio, MIRROR_FUSE3); + /* + * PTL platforms with media version 30.00 do not provide proper values + * for the media GT's L3 bank registers. Skip the readout since we + * don't have any way to obtain real values. + * + * This may get re-described as an official workaround in the future, + * but there's no tracking number assigned yet so we use a custom + * OOB workaround descriptor. + */ + if (XE_WA(gt, no_media_l3)) + return; + if (GRAPHICS_VER(xe) >= 20) { xe_l3_bank_mask_t per_node = {}; u32 meml3_en = REG_FIELD_GET(XE2_NODE_ENABLE_MASK, fuse3); diff --git a/drivers/gpu/drm/xe/xe_query.c b/drivers/gpu/drm/xe/xe_query.c index 5246a4a2740e..158629971eab 100644 --- a/drivers/gpu/drm/xe/xe_query.c +++ b/drivers/gpu/drm/xe/xe_query.c @@ -9,6 +9,7 @@ #include #include +#include #include #include "regs/xe_engine_regs.h" @@ -23,6 +24,7 @@ #include "xe_macros.h" #include "xe_mmio.h" #include "xe_ttm_vram_mgr.h" +#include "xe_wa.h" static const u16 xe_to_user_engine_class[] = { [XE_ENGINE_CLASS_RENDER] = DRM_XE_ENGINE_CLASS_RENDER, @@ -455,12 +457,23 @@ static int query_hwconfig(struct xe_device *xe, static size_t calc_topo_query_size(struct xe_device *xe) { - return xe->info.gt_count * - (4 * sizeof(struct drm_xe_query_topology_mask) + - sizeof_field(struct xe_gt, fuse_topo.g_dss_mask) + - sizeof_field(struct xe_gt, fuse_topo.c_dss_mask) + - sizeof_field(struct xe_gt, fuse_topo.l3_bank_mask) + - sizeof_field(struct xe_gt, fuse_topo.eu_mask_per_dss)); + struct xe_gt *gt; + size_t query_size = 0; + int id; + + for_each_gt(gt, xe, id) { + query_size += 3 * sizeof(struct drm_xe_query_topology_mask) + + sizeof_field(struct xe_gt, fuse_topo.g_dss_mask) + + sizeof_field(struct xe_gt, fuse_topo.c_dss_mask) + + sizeof_field(struct xe_gt, fuse_topo.eu_mask_per_dss); + + /* L3bank mask may not be available for some GTs */ + if (!XE_WA(gt, no_media_l3)) + query_size += sizeof(struct drm_xe_query_topology_mask) + + sizeof_field(struct xe_gt, fuse_topo.l3_bank_mask); + } + + return query_size; } static int copy_mask(void __user **ptr, @@ -513,11 +526,18 @@ static int query_gt_topology(struct xe_device *xe, if (err) return err; - topo.type = DRM_XE_TOPO_L3_BANK; - err = copy_mask(&query_ptr, &topo, gt->fuse_topo.l3_bank_mask, - sizeof(gt->fuse_topo.l3_bank_mask)); - if (err) - return err; + /* + * If the kernel doesn't have a way to obtain a correct L3bank + * mask, then it's better to omit L3 from the query rather than + * reporting bogus or zeroed information to userspace. + */ + if (!XE_WA(gt, no_media_l3)) { + topo.type = DRM_XE_TOPO_L3_BANK; + err = copy_mask(&query_ptr, &topo, gt->fuse_topo.l3_bank_mask, + sizeof(gt->fuse_topo.l3_bank_mask)); + if (err) + return err; + } topo.type = gt->fuse_topo.eu_type == XE_GT_EU_TYPE_SIMD16 ? DRM_XE_TOPO_SIMD16_EU_PER_DSS : diff --git a/drivers/gpu/drm/xe/xe_wa_oob.rules b/drivers/gpu/drm/xe/xe_wa_oob.rules index 920ca5060146..0154fbe154e9 100644 --- a/drivers/gpu/drm/xe/xe_wa_oob.rules +++ b/drivers/gpu/drm/xe/xe_wa_oob.rules @@ -37,3 +37,4 @@ 16023588340 GRAPHICS_VERSION(2001) 14019789679 GRAPHICS_VERSION(1255) GRAPHICS_VERSION_RANGE(1270, 2004) +no_media_l3 MEDIA_VERSION(3000) diff --git a/include/uapi/drm/xe_drm.h b/include/uapi/drm/xe_drm.h index b6fbe4988f2e..c4182e95a619 100644 --- a/include/uapi/drm/xe_drm.h +++ b/include/uapi/drm/xe_drm.h @@ -512,7 +512,9 @@ struct drm_xe_query_gt_list { * containing the following in mask: * ``DSS_COMPUTE ff ff ff ff 00 00 00 00`` * means 32 DSS are available for compute. - * - %DRM_XE_TOPO_L3_BANK - To query the mask of enabled L3 banks + * - %DRM_XE_TOPO_L3_BANK - To query the mask of enabled L3 banks. This type + * may be omitted if the driver is unable to query the mask from the + * hardware. * - %DRM_XE_TOPO_EU_PER_DSS - To query the mask of Execution Units (EU) * available per Dual Sub Slices (DSS). For example a query response * containing the following in mask: -- 2.50.1 From 317d81085c87f12118754df0986b66b3f8581f2d Mon Sep 17 00:00:00 2001 From: Matt Roper Date: Mon, 7 Oct 2024 18:35:04 -0700 Subject: [PATCH 12/16] drm/xe/xe3: Xe3 uses the same PAT settings as Xe2 Xe3 platforms use the same PAT tables as Xe2. Bspec: 71582 Signed-off-by: Matt Roper Signed-off-by: Matt Atwood Reviewed-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20241008013509.61233-2-matthew.s.atwood@intel.com --- drivers/gpu/drm/xe/xe_pat.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/xe/xe_pat.c b/drivers/gpu/drm/xe/xe_pat.c index 6bd6adfdfc74..b16473818173 100644 --- a/drivers/gpu/drm/xe/xe_pat.c +++ b/drivers/gpu/drm/xe/xe_pat.c @@ -387,7 +387,7 @@ static const struct xe_pat_ops xe2_pat_ops = { void xe_pat_init_early(struct xe_device *xe) { - if (GRAPHICS_VER(xe) == 20) { + if (GRAPHICS_VER(xe) == 30 || GRAPHICS_VER(xe) == 20) { xe->pat.ops = &xe2_pat_ops; xe->pat.table = xe2_pat_table; -- 2.50.1 From 800d75bf20ae429698ecf7a8a392b638f86f9642 Mon Sep 17 00:00:00 2001 From: Haridhar Kalvala Date: Mon, 7 Oct 2024 18:35:05 -0700 Subject: [PATCH 13/16] drm/xe/xe3: Define Xe3 feature flags Define a common set of Xe3 feature flags and definitions that will be used for all platforms in this family. The feature flags are inherited unchanged from the Xe2 (XE2_FEATURES) platform. Following B-spec details inherited from Xe2 feature flag definition commit. v2: reuse graphics_xe2 definition Bspec: 58695 - dma_mask_size remains 46 (not documented in bspec) - supports_usm=1 (Bspec 59651) - has_flatccs=1 (Bspec 58797) - has_4tile=1 (Bspec 58788) - has_asid=1 (Bspec 59654, 59265, 60288) - has_range_tlb_invalidate=1 (Bspec 71126) - five-level page table (Bspec 59505) - 1 VD + 1 VE + 1 SFC (Bspec 67103, 70819) - platform engine mask (Bspec 60149) Cc: Matt Roper Signed-off-by: Haridhar Kalvala Signed-off-by: Matt Atwood Reviewed-by: Shekhar Chauhan Signed-off-by: Matt Roper Link: https://patchwork.freedesktop.org/patch/msgid/20241008013509.61233-3-matthew.s.atwood@intel.com --- drivers/gpu/drm/xe/xe_pci.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/xe/xe_pci.c b/drivers/gpu/drm/xe/xe_pci.c index af7b2f2ff13a..d99f5ddbc70d 100644 --- a/drivers/gpu/drm/xe/xe_pci.c +++ b/drivers/gpu/drm/xe/xe_pci.c @@ -208,7 +208,7 @@ static const struct xe_media_desc media_xelpmp = { }; static const struct xe_media_desc media_xe2 = { - .name = "Xe2_LPM / Xe2_HPM", + .name = "Xe2_LPM / Xe2_HPM / Xe3_LPM", .hw_engine_mask = GENMASK(XE_HW_ENGINE_VCS7, XE_HW_ENGINE_VCS0) | GENMASK(XE_HW_ENGINE_VECS3, XE_HW_ENGINE_VECS0) | @@ -356,6 +356,8 @@ static const struct gmdid_map graphics_ip_map[] = { { 1274, &graphics_xelpg }, /* Xe_LPG+ */ { 2001, &graphics_xe2 }, { 2004, &graphics_xe2 }, + { 3000, &graphics_xe2 }, + { 3001, &graphics_xe2 }, }; /* Map of GMD_ID values to media IP */ @@ -363,6 +365,7 @@ static const struct gmdid_map media_ip_map[] = { { 1300, &media_xelpmp }, { 1301, &media_xe2 }, { 2000, &media_xe2 }, + { 3000, &media_xe2 }, }; #define INTEL_VGA_DEVICE(id, info) { \ -- 2.50.1 From 37466119ff12c40ecf42b916f755ba4b0a5f8b1a Mon Sep 17 00:00:00 2001 From: Haridhar Kalvala Date: Mon, 7 Oct 2024 18:35:07 -0700 Subject: [PATCH 14/16] drm/xe/ptl: PTL re-uses Xe2 MOCS table PTL is Xe3 architecture but there is no difference between LNL and PTL in MOCS table. So, PTL uses the same MOCS table as LNL. Bspec: 71582 Cc: Matt Roper Cc: Shekhar Chauhan Signed-off-by: Haridhar Kalvala Signed-off-by: Matt Atwood Reviewed-by: Shekhar Chauhan Signed-off-by: Matt Roper Link: https://patchwork.freedesktop.org/patch/msgid/20241008013509.61233-5-matthew.s.atwood@intel.com --- drivers/gpu/drm/xe/xe_mocs.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/xe/xe_mocs.c b/drivers/gpu/drm/xe/xe_mocs.c index 8df41cd12d51..231d0e86ed83 100644 --- a/drivers/gpu/drm/xe/xe_mocs.c +++ b/drivers/gpu/drm/xe/xe_mocs.c @@ -576,6 +576,7 @@ static unsigned int get_mocs_settings(struct xe_device *xe, memset(info, 0, sizeof(struct xe_mocs_info)); switch (xe->info.platform) { + case XE_PANTHERLAKE: case XE_LUNARLAKE: case XE_BATTLEMAGE: info->ops = &xe2_mocs_ops; -- 2.50.1 From 2298d8a81f2dc6987448e5ddd823f4892194f5b6 Mon Sep 17 00:00:00 2001 From: Haridhar Kalvala Date: Mon, 7 Oct 2024 18:35:08 -0700 Subject: [PATCH 15/16] drm/xe/ptl: Add PTL platform definition PTL is an integrated GPU based on the Xe3 architecture. v2: explicitly turn off display until display patches land. Bspec: 72574 Cc: Matt Roper Signed-off-by: Haridhar Kalvala Signed-off-by: Matt Atwood Reviewed-by: Shekhar Chauhan Signed-off-by: Matt Roper Link: https://patchwork.freedesktop.org/patch/msgid/20241008013509.61233-6-matthew.s.atwood@intel.com --- drivers/gpu/drm/xe/xe_pci.c | 7 +++++++ drivers/gpu/drm/xe/xe_platform_types.h | 1 + include/drm/intel/xe_pciids.h | 11 +++++++++++ 3 files changed, 19 insertions(+) diff --git a/drivers/gpu/drm/xe/xe_pci.c b/drivers/gpu/drm/xe/xe_pci.c index d99f5ddbc70d..64a8336ca437 100644 --- a/drivers/gpu/drm/xe/xe_pci.c +++ b/drivers/gpu/drm/xe/xe_pci.c @@ -346,6 +346,12 @@ static const struct xe_device_desc bmg_desc = { .has_heci_cscfi = 1, }; +static const struct xe_device_desc ptl_desc = { + PLATFORM(PANTHERLAKE), + .has_display = false, + .require_force_probe = true, +}; + #undef PLATFORM __diag_pop(); @@ -395,6 +401,7 @@ static const struct pci_device_id pciidlist[] = { XE_MTL_IDS(INTEL_VGA_DEVICE, &mtl_desc), XE_LNL_IDS(INTEL_VGA_DEVICE, &lnl_desc), XE_BMG_IDS(INTEL_VGA_DEVICE, &bmg_desc), + XE_PTL_IDS(INTEL_VGA_DEVICE, &ptl_desc), { } }; MODULE_DEVICE_TABLE(pci, pciidlist); diff --git a/drivers/gpu/drm/xe/xe_platform_types.h b/drivers/gpu/drm/xe/xe_platform_types.h index 79b7042c4534..d08574c4cdb8 100644 --- a/drivers/gpu/drm/xe/xe_platform_types.h +++ b/drivers/gpu/drm/xe/xe_platform_types.h @@ -23,6 +23,7 @@ enum xe_platform { XE_METEORLAKE, XE_LUNARLAKE, XE_BATTLEMAGE, + XE_PANTHERLAKE, }; enum xe_subplatform { diff --git a/include/drm/intel/xe_pciids.h b/include/drm/intel/xe_pciids.h index 65520a90c17c..6d8d013f74e0 100644 --- a/include/drm/intel/xe_pciids.h +++ b/include/drm/intel/xe_pciids.h @@ -220,4 +220,15 @@ MACRO__(0xE20D, ## __VA_ARGS__), \ MACRO__(0xE212, ## __VA_ARGS__) +#define XE_PTL_IDS(MACRO__, ...) \ + MACRO__(0xB080, ## __VA_ARGS__), \ + MACRO__(0xB081, ## __VA_ARGS__), \ + MACRO__(0xB082, ## __VA_ARGS__), \ + MACRO__(0xB090, ## __VA_ARGS__), \ + MACRO__(0xB091, ## __VA_ARGS__), \ + MACRO__(0xB092, ## __VA_ARGS__), \ + MACRO__(0xB0A0, ## __VA_ARGS__), \ + MACRO__(0xB0A1, ## __VA_ARGS__), \ + MACRO__(0xB0A2, ## __VA_ARGS__) + #endif -- 2.50.1 From d6d87a10d9e45f7b5ae142dedb7ff76d1cf9e475 Mon Sep 17 00:00:00 2001 From: Matt Roper Date: Mon, 7 Oct 2024 18:35:09 -0700 Subject: [PATCH 16/16] drm/xe/xe3lpm: Add new "instance0" steering table MCR steering on Xe3 media IP is almost the same as it was on Xe2, except for one new range (0x38D0D0 - 0x38D0FF) which has changed to an MCR "MEDIAINF" range on Xe3. Since we can always steer to grpid / instanceid 0 for MEDIAINF ranges, define a new "INSTANCE0" steering table for Xe3 media. Xe3 can continue to use the same OADDRM/GPMXMT table as Xe2. v2: Merge continuous entries 38D0D0 - 38F0FF Bspec: 74298 Signed-off-by: Matt Roper Signed-off-by: Matt Atwood Reviewed-by: Gustavo Sousa Link: https://patchwork.freedesktop.org/patch/msgid/20241008013509.61233-7-matthew.s.atwood@intel.com --- drivers/gpu/drm/xe/xe_gt_mcr.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/xe/xe_gt_mcr.c b/drivers/gpu/drm/xe/xe_gt_mcr.c index 4c0767403881..e21eb6c0fc20 100644 --- a/drivers/gpu/drm/xe/xe_gt_mcr.c +++ b/drivers/gpu/drm/xe/xe_gt_mcr.c @@ -237,6 +237,17 @@ static const struct xe_mmio_range xe2lpm_instance0_steering_table[] = { {}, }; +static const struct xe_mmio_range xe3lpm_instance0_steering_table[] = { + { 0x384000, 0x3847DF }, /* GAM, rsvd, GAM */ + { 0x384900, 0x384AFF }, /* GAM */ + { 0x389560, 0x3895FF }, /* MEDIAINF */ + { 0x38B600, 0x38B8FF }, /* L3BANK */ + { 0x38C800, 0x38D07F }, /* GAM, MEDIAINF */ + { 0x38D0D0, 0x38F0FF }, /* MEDIAINF, GAM */ + { 0x393C00, 0x393C7F }, /* MEDIAINF */ + {}, +}; + static void init_steering_l3bank(struct xe_gt *gt) { struct xe_mmio *mmio = >->mmio; @@ -441,7 +452,10 @@ void xe_gt_mcr_init(struct xe_gt *gt) if (gt->info.type == XE_GT_TYPE_MEDIA) { drm_WARN_ON(&xe->drm, MEDIA_VER(xe) < 13); - if (MEDIA_VERx100(xe) >= 1301) { + if (MEDIA_VER(xe) >= 30) { + gt->steering[OADDRM].ranges = xe2lpm_gpmxmt_steering_table; + gt->steering[INSTANCE0].ranges = xe3lpm_instance0_steering_table; + } else if (MEDIA_VERx100(xe) >= 1301) { gt->steering[OADDRM].ranges = xe2lpm_gpmxmt_steering_table; gt->steering[INSTANCE0].ranges = xe2lpm_instance0_steering_table; } else { -- 2.50.1