* [PATCH 0/3] drm/xe/oa: Wa_14026633728
@ 2026-04-07 3:02 Ashutosh Dixit
2026-04-07 3:02 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit
` (5 more replies)
0 siblings, 6 replies; 30+ messages in thread
From: Ashutosh Dixit @ 2026-04-07 3:02 UTC (permalink / raw)
To: intel-xe; +Cc: Umesh Nerlige Ramappa
Ashutosh Dixit (3):
drm/xe/oa: Use xe_map layer
drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap
drm/xe/oa: Implement Wa_14026633728
drivers/gpu/drm/xe/xe_oa.c | 117 +++++++++++++++--------------
drivers/gpu/drm/xe/xe_oa_types.h | 3 -
drivers/gpu/drm/xe/xe_wa_oob.rules | 1 +
3 files changed, 62 insertions(+), 59 deletions(-)
--
2.48.1
^ permalink raw reply [flat|nested] 30+ messages in thread* [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-07 3:02 [PATCH 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit @ 2026-04-07 3:02 ` Ashutosh Dixit 2026-04-07 22:49 ` Umesh Nerlige Ramappa 2026-04-07 3:02 ` [PATCH 2/3] drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap Ashutosh Dixit ` (4 subsequent siblings) 5 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-07 3:02 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa OA code should have used xe_map layer to begin with. In CRI, the OA buffer can be located both in system and device memory. For these reasons, move OA code to use the xe_map layer when accessing the OA buffer. Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> --- drivers/gpu/drm/xe/xe_oa.c | 88 ++++++++++++++++++-------------- drivers/gpu/drm/xe/xe_oa_types.h | 3 -- 2 files changed, 49 insertions(+), 42 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index 6337e671c97ae..dfd3b6a789cc3 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) #define oa_report_header_64bit(__s) \ ((__s)->oa_buffer.format->header == HDR_64_BIT) -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) +static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) { - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + return oa_report_header_64bit(stream) ? + xe_map_rd(stream->oa->xe, map, tail, u64) : + xe_map_rd(stream->oa->xe, map, tail, u32); } -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)report = 0; - else - *report = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head, u64, 0) : + xe_map_wr(stream->oa->xe, map, head, u32, 0); } -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + return oa_report_header_64bit(stream) ? - *((u64 *)report + 1) : - *((u32 *)report + 1); + xe_map_rd(stream->oa->xe, map, tail + 2, u64) : + xe_map_rd(stream->oa->xe, map, tail + 1, u32); } -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)&report[2] = 0; - else - report[1] = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : + xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); } static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) * they were written. If not : (╯°□°）╯︵ ┻━┻ */ while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { - void *report = stream->oa_buffer.vaddr + tail; - - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) break; tail = xe_oa_circ_diff(stream, tail, report_size); @@ -311,30 +317,37 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) return HRTIMER_RESTART; } +static inline unsigned long +xe_oa_copy_to_user(void __user *dst, const struct iosys_map *src, size_t src_offset, size_t len) +{ + if (src->is_iomem) + return copy_to_user(dst, src->vaddr_iomem + src_offset, len); + else + return copy_to_user(dst, src->vaddr + src_offset, len); +} + static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, - size_t count, size_t *offset, const u8 *report) + size_t count, size_t *offset, u32 head) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; int report_size = stream->oa_buffer.format->size; int report_size_partial; - u8 *oa_buf_end; if ((count - *offset) < report_size) return -ENOSPC; buf += *offset; - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - report_size_partial = oa_buf_end - report; + report_size_partial = stream->oa_buffer.circ_size - head; if (report_size_partial < report_size) { - if (copy_to_user(buf, report, report_size_partial)) + if (xe_oa_copy_to_user(buf, map, head, report_size_partial)) return -EFAULT; buf += report_size_partial; - if (copy_to_user(buf, stream->oa_buffer.vaddr, - report_size - report_size_partial)) + if (xe_oa_copy_to_user(buf, map, 0, report_size - report_size_partial)) return -EFAULT; - } else if (copy_to_user(buf, report, report_size)) { + } else if (xe_oa_copy_to_user(buf, map, head, report_size)) { return -EFAULT; } @@ -347,7 +360,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, size_t count, size_t *offset) { int report_size = stream->oa_buffer.format->size; - u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); size_t start_offset = *offset; unsigned long flags; @@ -364,26 +376,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, for (; xe_oa_circ_diff(stream, tail, head); head = xe_oa_circ_incr(stream, head, report_size)) { - u8 *report = oa_buf_base + head; - - ret = xe_oa_append_report(stream, buf, count, offset, report); + ret = xe_oa_append_report(stream, buf, count, offset, head); if (ret) break; if (!(stream->oa_buffer.circ_size % report_size)) { /* Clear out report id and timestamp to detect unlanded reports */ - oa_report_id_clear(stream, (void *)report); - oa_timestamp_clear(stream, (void *)report); + oa_report_id_clear(stream, head); + oa_timestamp_clear(stream, head); } else { - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - u32 part = oa_buf_end - report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + u32 part = stream->oa_buffer.circ_size - head; /* Zero out the entire report */ if (report_size <= part) { - memset(report, 0, report_size); + xe_map_memset(stream->oa->xe, map, head, 0, report_size); } else { - memset(report, 0, part); - memset(oa_buf_base, 0, report_size - part); + xe_map_memset(stream->oa->xe, map, head, 0, part); + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); } } } @@ -436,7 +446,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, + xe_bo_size(stream->oa_buffer.bo)); } static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) @@ -891,7 +902,6 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) stream->oa_buffer.bo = bo; /* mmap implementation requires OA buffer to be in system memory */ xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); - stream->oa_buffer.vaddr = bo->vmap.vaddr; return 0; } diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h index b03ffd5134834..2dd550c93b8ac 100644 --- a/drivers/gpu/drm/xe/xe_oa_types.h +++ b/drivers/gpu/drm/xe/xe_oa_types.h @@ -162,9 +162,6 @@ struct xe_oa_buffer { /** @format: xe_bo backing the OA buffer */ struct xe_bo *bo; - /** @vaddr: mapped vaddr of the OA buffer */ - u8 *vaddr; - /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ spinlock_t ptr_lock; -- 2.48.1 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-07 3:02 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit @ 2026-04-07 22:49 ` Umesh Nerlige Ramappa 2026-04-08 15:32 ` Dixit, Ashutosh 0 siblings, 1 reply; 30+ messages in thread From: Umesh Nerlige Ramappa @ 2026-04-07 22:49 UTC (permalink / raw) To: Ashutosh Dixit; +Cc: intel-xe On Mon, Apr 06, 2026 at 08:02:17PM -0700, Ashutosh Dixit wrote: >OA code should have used xe_map layer to begin with. In CRI, the OA buffer >can be located both in system and device memory. For these reasons, move OA >code to use the xe_map layer when accessing the OA buffer. > >Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> >--- > drivers/gpu/drm/xe/xe_oa.c | 88 ++++++++++++++++++-------------- > drivers/gpu/drm/xe/xe_oa_types.h | 3 -- > 2 files changed, 49 insertions(+), 42 deletions(-) > >diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c >index 6337e671c97ae..dfd3b6a789cc3 100644 >--- a/drivers/gpu/drm/xe/xe_oa.c >+++ b/drivers/gpu/drm/xe/xe_oa.c >@@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) > #define oa_report_header_64bit(__s) \ > ((__s)->oa_buffer.format->header == HDR_64_BIT) > >-static u64 oa_report_id(struct xe_oa_stream *stream, void *report) >+static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) > { >- return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ return oa_report_header_64bit(stream) ? >+ xe_map_rd(stream->oa->xe, map, tail, u64) : >+ xe_map_rd(stream->oa->xe, map, tail, u32); > } > >-static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) >+static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) > { >- if (oa_report_header_64bit(stream)) >- *(u64 *)report = 0; >- else >- *report = 0; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ oa_report_header_64bit(stream) ? >+ xe_map_wr(stream->oa->xe, map, head, u64, 0) : >+ xe_map_wr(stream->oa->xe, map, head, u32, 0); > } > >-static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) >+static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) > { >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ > return oa_report_header_64bit(stream) ? >- *((u64 *)report + 1) : >- *((u32 *)report + 1); >+ xe_map_rd(stream->oa->xe, map, tail + 2, u64) : >+ xe_map_rd(stream->oa->xe, map, tail + 1, u32); > } > >-static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) >+static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) > { >- if (oa_report_header_64bit(stream)) >- *(u64 *)&report[2] = 0; >- else >- report[1] = 0; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ oa_report_header_64bit(stream) ? >+ xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : >+ xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); > } > > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) >@@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > * they were written. If not : (╯°□°）╯︵ ┻━┻ > */ > while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { >- void *report = stream->oa_buffer.vaddr + tail; >- >- if (oa_report_id(stream, report) || oa_timestamp(stream, report)) >+ if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) > break; > > tail = xe_oa_circ_diff(stream, tail, report_size); >@@ -311,30 +317,37 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) > return HRTIMER_RESTART; > } > >+static inline unsigned long >+xe_oa_copy_to_user(void __user *dst, const struct iosys_map *src, size_t src_offset, size_t len) >+{ >+ if (src->is_iomem) >+ return copy_to_user(dst, src->vaddr_iomem + src_offset, len); I think this requires xe_map_memcpy_from for the iomem case. Thanks, Umesh >+ else >+ return copy_to_user(dst, src->vaddr + src_offset, len); >+} >+ > static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, >- size_t count, size_t *offset, const u8 *report) >+ size_t count, size_t *offset, u32 head) > { >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; > int report_size = stream->oa_buffer.format->size; > int report_size_partial; >- u8 *oa_buf_end; > > if ((count - *offset) < report_size) > return -ENOSPC; > > buf += *offset; > >- oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >- report_size_partial = oa_buf_end - report; >+ report_size_partial = stream->oa_buffer.circ_size - head; > > if (report_size_partial < report_size) { >- if (copy_to_user(buf, report, report_size_partial)) >+ if (xe_oa_copy_to_user(buf, map, head, report_size_partial)) > return -EFAULT; > buf += report_size_partial; > >- if (copy_to_user(buf, stream->oa_buffer.vaddr, >- report_size - report_size_partial)) >+ if (xe_oa_copy_to_user(buf, map, 0, report_size - report_size_partial)) > return -EFAULT; >- } else if (copy_to_user(buf, report, report_size)) { >+ } else if (xe_oa_copy_to_user(buf, map, head, report_size)) { > return -EFAULT; > } > >@@ -347,7 +360,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > size_t count, size_t *offset) > { > int report_size = stream->oa_buffer.format->size; >- u8 *oa_buf_base = stream->oa_buffer.vaddr; > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); > size_t start_offset = *offset; > unsigned long flags; >@@ -364,26 +376,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > for (; xe_oa_circ_diff(stream, tail, head); > head = xe_oa_circ_incr(stream, head, report_size)) { >- u8 *report = oa_buf_base + head; >- >- ret = xe_oa_append_report(stream, buf, count, offset, report); >+ ret = xe_oa_append_report(stream, buf, count, offset, head); > if (ret) > break; > > if (!(stream->oa_buffer.circ_size % report_size)) { > /* Clear out report id and timestamp to detect unlanded reports */ >- oa_report_id_clear(stream, (void *)report); >- oa_timestamp_clear(stream, (void *)report); >+ oa_report_id_clear(stream, head); >+ oa_timestamp_clear(stream, head); > } else { >- u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >- u32 part = oa_buf_end - report; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ u32 part = stream->oa_buffer.circ_size - head; > > /* Zero out the entire report */ > if (report_size <= part) { >- memset(report, 0, report_size); >+ xe_map_memset(stream->oa->xe, map, head, 0, report_size); > } else { >- memset(report, 0, part); >- memset(oa_buf_base, 0, report_size - part); >+ xe_map_memset(stream->oa->xe, map, head, 0, part); >+ xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); > } > } > } >@@ -436,7 +446,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) > spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); > > /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ >- memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); >+ xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, >+ xe_bo_size(stream->oa_buffer.bo)); > } > > static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) >@@ -891,7 +902,6 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > stream->oa_buffer.bo = bo; > /* mmap implementation requires OA buffer to be in system memory */ > xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); >- stream->oa_buffer.vaddr = bo->vmap.vaddr; > return 0; > } > >diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h >index b03ffd5134834..2dd550c93b8ac 100644 >--- a/drivers/gpu/drm/xe/xe_oa_types.h >+++ b/drivers/gpu/drm/xe/xe_oa_types.h >@@ -162,9 +162,6 @@ struct xe_oa_buffer { > /** @format: xe_bo backing the OA buffer */ > struct xe_bo *bo; > >- /** @vaddr: mapped vaddr of the OA buffer */ >- u8 *vaddr; >- > /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ > spinlock_t ptr_lock; > >-- >2.48.1 > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-07 22:49 ` Umesh Nerlige Ramappa @ 2026-04-08 15:32 ` Dixit, Ashutosh 2026-04-08 18:21 ` Umesh Nerlige Ramappa 0 siblings, 1 reply; 30+ messages in thread From: Dixit, Ashutosh @ 2026-04-08 15:32 UTC (permalink / raw) To: Umesh Nerlige Ramappa; +Cc: intel-xe On Tue, 07 Apr 2026 15:49:27 -0700, Umesh Nerlige Ramappa wrote: > Hi Umesh, > On Mon, Apr 06, 2026 at 08:02:17PM -0700, Ashutosh Dixit wrote: > > @@ -311,30 +317,37 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) > > return HRTIMER_RESTART; > > } > > > > +static inline unsigned long > > +xe_oa_copy_to_user(void __user *dst, const struct iosys_map *src, size_t src_offset, size_t len) > > +{ > > + if (src->is_iomem) > > + return copy_to_user(dst, src->vaddr_iomem + src_offset, len); > > I think this requires xe_map_memcpy_from for the iomem case. > > > + else > > + return copy_to_user(dst, src->vaddr + src_offset, len); > > +} > > + Yes, you have certainly spotted the iffy part of this patch. However, this situation is not that simple: * xe_map_memcpy_from or iosys_map_memcpy_from copies from iomem to kernel memory (potentially using special accessors for iomem), not to user memory * copy_to_user copies from kernel memory to user memory (using special checks), not from iomem So in the kernel we don't have functions which copy directly from iomem to user memory. That is, functions like iosys_map_copy_to_user() or copy_to_user_from_iomem() do not exist, because they would need to combine iomem access and copy_to_user (likely copy_to_user will need to be rewritten with iomem accessors, if such a function were to be implemented). There may be a reason why such functions don't exist in the kernel. Another way might be to copy from iomem to kernel memory and from kernel to user memory (that is have both iosys_map_memcpy_from and copy_to_user calls), but this would entail an extra copy into a bounce buffer in kernel memory. Something like this: static inline unsigned long xe_oa_copy_to_user(void __user *dst, const struct iosys_map *src, size_t src_offset, size_t len) { void *tmp = kmalloc(len, GFP_KERNEL); // kmalloc or statically allocated tmp buffer xe_map_memcpy_from(xe, tmp, src, src_offset, len); // extra copy return copy_to_user(dst, tmp, len); } In our case, xe_map_memcpy_from or iosys_map_memcpy_from are basically a memcpy (because vram bar is ioremapped, special accessors for iomem are not needed). Therefore, to avoid the extra copy, I decided to drop the iomem access part (since the copy_to_user part cannot be dropped). Hence I end up with the xe_oa_copy_to_user() above. The function was tested on BMG with OA buffer in vram and works as expected. I did think about renaming the above function xe_map_copy_to_user() and moving it to xe_map.h, so that is possible. But since there are no other users (and none are expected) at this point of xe_map_copy_to_user() apart from OA because of this workaround (other observation stream buffers are in system memory, not vram), I left the function as specific to OA. Let me know what you think. Thanks. ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-08 15:32 ` Dixit, Ashutosh @ 2026-04-08 18:21 ` Umesh Nerlige Ramappa 0 siblings, 0 replies; 30+ messages in thread From: Umesh Nerlige Ramappa @ 2026-04-08 18:21 UTC (permalink / raw) To: Dixit, Ashutosh; +Cc: intel-xe On Wed, Apr 08, 2026 at 08:32:50AM -0700, Dixit, Ashutosh wrote: >On Tue, 07 Apr 2026 15:49:27 -0700, Umesh Nerlige Ramappa wrote: >> > >Hi Umesh, > >> On Mon, Apr 06, 2026 at 08:02:17PM -0700, Ashutosh Dixit wrote: >> > @@ -311,30 +317,37 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) >> > return HRTIMER_RESTART; >> > } >> > >> > +static inline unsigned long >> > +xe_oa_copy_to_user(void __user *dst, const struct iosys_map *src, size_t src_offset, size_t len) >> > +{ >> > + if (src->is_iomem) >> > + return copy_to_user(dst, src->vaddr_iomem + src_offset, len); >> >> I think this requires xe_map_memcpy_from for the iomem case. >> >> > + else >> > + return copy_to_user(dst, src->vaddr + src_offset, len); >> > +} >> > + > >Yes, you have certainly spotted the iffy part of this patch. However, this >situation is not that simple: > >* xe_map_memcpy_from or iosys_map_memcpy_from copies from iomem to kernel > memory (potentially using special accessors for iomem), not to user > memory > >* copy_to_user copies from kernel memory to user memory (using special > checks), not from iomem > >So in the kernel we don't have functions which copy directly from iomem to >user memory. That is, functions like iosys_map_copy_to_user() or >copy_to_user_from_iomem() do not exist, because they would need to combine >iomem access and copy_to_user (likely copy_to_user will need to be >rewritten with iomem accessors, if such a function were to be >implemented). There may be a reason why such functions don't exist in the >kernel. > >Another way might be to copy from iomem to kernel memory and from kernel to >user memory (that is have both iosys_map_memcpy_from and copy_to_user >calls), but this would entail an extra copy into a bounce buffer in kernel >memory. Something like this: > >static inline unsigned long >xe_oa_copy_to_user(void __user *dst, const struct iosys_map *src, size_t src_offset, size_t len) >{ > void *tmp = kmalloc(len, GFP_KERNEL); // kmalloc or statically allocated tmp buffer > > xe_map_memcpy_from(xe, tmp, src, src_offset, len); // extra copy > > return copy_to_user(dst, tmp, len); >} I did notice that there wasn't anything like copy_from_iomem_to_user. I was imagining you could copy over 32-bit iomem words into the user buffer using the iosys_map accessors. That should be possible and would avoid an intermediate buffer. > >In our case, xe_map_memcpy_from or iosys_map_memcpy_from are basically a >memcpy (because vram bar is ioremapped, special accessors for iomem are not >needed). Hmm, then why even have this patch :). fwiu, you are accessing the OA buffer using iosys_map accessors in this patch. It should work without it because of the above statement. If we go the iosys_map route, then let's do it consistently. If not, we should just drop this patch. I would lean into reading 32bit words into the user buffer. Regards, Umesh >Therefore, to avoid the extra copy, I decided to drop the iomem >access part (since the copy_to_user part cannot be dropped). Hence I end up >with the xe_oa_copy_to_user() above. The function was tested on BMG with OA >buffer in vram and works as expected. > >I did think about renaming the above function xe_map_copy_to_user() and >moving it to xe_map.h, so that is possible. But since there are no other >users (and none are expected) at this point of xe_map_copy_to_user() apart >from OA because of this workaround (other observation stream buffers are in >system memory, not vram), I left the function as specific to OA. > >Let me know what you think. Thanks. ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 2/3] drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap 2026-04-07 3:02 [PATCH 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit 2026-04-07 3:02 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit @ 2026-04-07 3:02 ` Ashutosh Dixit 2026-04-07 22:52 ` Umesh Nerlige Ramappa 2026-04-07 3:02 ` [PATCH 3/3] drm/xe/oa: Implement Wa_14026633728 Ashutosh Dixit ` (3 subsequent siblings) 5 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-07 3:02 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa OA buffer mmap can currently only mmap OA buffer in system memory. CRI MERTOA buffer can be located in device memory. Switch OA buffer mmap to using drm_gem_mmap_obj, which can handle mmap's of both system and device memory buffers. Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> --- drivers/gpu/drm/xe/xe_oa.c | 19 +++---------------- 1 file changed, 3 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index dfd3b6a789cc3..0e65141e57ee8 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -9,6 +9,7 @@ #include <linux/poll.h> #include <drm/drm_drv.h> +#include <drm/drm_gem.h> #include <drm/drm_managed.h> #include <drm/drm_syncobj.h> #include <uapi/drm/xe_drm.h> @@ -900,8 +901,6 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) return PTR_ERR(bo); stream->oa_buffer.bo = bo; - /* mmap implementation requires OA buffer to be in system memory */ - xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); return 0; } @@ -1683,8 +1682,6 @@ static int xe_oa_mmap(struct file *file, struct vm_area_struct *vma) { struct xe_oa_stream *stream = file->private_data; struct xe_bo *bo = stream->oa_buffer.bo; - unsigned long start = vma->vm_start; - int i, ret; if (xe_observation_paranoid && !perfmon_capable()) { drm_dbg(&stream->oa->xe->drm, "Insufficient privilege to map OA buffer\n"); @@ -1692,7 +1689,7 @@ static int xe_oa_mmap(struct file *file, struct vm_area_struct *vma) } /* Can mmap the entire OA buffer or nothing (no partial OA buffer mmaps) */ - if (vma->vm_end - vma->vm_start != xe_bo_size(stream->oa_buffer.bo)) { + if (vma->vm_end - vma->vm_start != xe_bo_size(bo)) { drm_dbg(&stream->oa->xe->drm, "Wrong mmap size, must be OA buffer size\n"); return -EINVAL; } @@ -1708,17 +1705,7 @@ static int xe_oa_mmap(struct file *file, struct vm_area_struct *vma) vm_flags_mod(vma, VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP | VM_DONTCOPY, VM_MAYWRITE | VM_MAYEXEC); - xe_assert(stream->oa->xe, bo->ttm.ttm->num_pages == vma_pages(vma)); - for (i = 0; i < bo->ttm.ttm->num_pages; i++) { - ret = remap_pfn_range(vma, start, page_to_pfn(bo->ttm.ttm->pages[i]), - PAGE_SIZE, vma->vm_page_prot); - if (ret) - break; - - start += PAGE_SIZE; - } - - return ret; + return drm_gem_mmap_obj(&bo->ttm.base, xe_bo_size(bo), vma); } static const struct file_operations xe_oa_fops = { -- 2.48.1 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 2/3] drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap 2026-04-07 3:02 ` [PATCH 2/3] drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap Ashutosh Dixit @ 2026-04-07 22:52 ` Umesh Nerlige Ramappa 0 siblings, 0 replies; 30+ messages in thread From: Umesh Nerlige Ramappa @ 2026-04-07 22:52 UTC (permalink / raw) To: Ashutosh Dixit; +Cc: intel-xe On Mon, Apr 06, 2026 at 08:02:18PM -0700, Ashutosh Dixit wrote: >OA buffer mmap can currently only mmap OA buffer in system memory. CRI >MERTOA buffer can be located in device memory. Switch OA buffer mmap to >using drm_gem_mmap_obj, which can handle mmap's of both system and device >memory buffers. > >Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> If it works for both types of memory, then this looks better than before, Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com> Thanks, Umesh >--- > drivers/gpu/drm/xe/xe_oa.c | 19 +++---------------- > 1 file changed, 3 insertions(+), 16 deletions(-) > >diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c >index dfd3b6a789cc3..0e65141e57ee8 100644 >--- a/drivers/gpu/drm/xe/xe_oa.c >+++ b/drivers/gpu/drm/xe/xe_oa.c >@@ -9,6 +9,7 @@ > #include <linux/poll.h> > > #include <drm/drm_drv.h> >+#include <drm/drm_gem.h> > #include <drm/drm_managed.h> > #include <drm/drm_syncobj.h> > #include <uapi/drm/xe_drm.h> >@@ -900,8 +901,6 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > return PTR_ERR(bo); > > stream->oa_buffer.bo = bo; >- /* mmap implementation requires OA buffer to be in system memory */ >- xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); > return 0; > } > >@@ -1683,8 +1682,6 @@ static int xe_oa_mmap(struct file *file, struct vm_area_struct *vma) > { > struct xe_oa_stream *stream = file->private_data; > struct xe_bo *bo = stream->oa_buffer.bo; >- unsigned long start = vma->vm_start; >- int i, ret; > > if (xe_observation_paranoid && !perfmon_capable()) { > drm_dbg(&stream->oa->xe->drm, "Insufficient privilege to map OA buffer\n"); >@@ -1692,7 +1689,7 @@ static int xe_oa_mmap(struct file *file, struct vm_area_struct *vma) > } > > /* Can mmap the entire OA buffer or nothing (no partial OA buffer mmaps) */ >- if (vma->vm_end - vma->vm_start != xe_bo_size(stream->oa_buffer.bo)) { >+ if (vma->vm_end - vma->vm_start != xe_bo_size(bo)) { > drm_dbg(&stream->oa->xe->drm, "Wrong mmap size, must be OA buffer size\n"); > return -EINVAL; > } >@@ -1708,17 +1705,7 @@ static int xe_oa_mmap(struct file *file, struct vm_area_struct *vma) > vm_flags_mod(vma, VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP | VM_DONTCOPY, > VM_MAYWRITE | VM_MAYEXEC); > >- xe_assert(stream->oa->xe, bo->ttm.ttm->num_pages == vma_pages(vma)); >- for (i = 0; i < bo->ttm.ttm->num_pages; i++) { >- ret = remap_pfn_range(vma, start, page_to_pfn(bo->ttm.ttm->pages[i]), >- PAGE_SIZE, vma->vm_page_prot); >- if (ret) >- break; >- >- start += PAGE_SIZE; >- } >- >- return ret; >+ return drm_gem_mmap_obj(&bo->ttm.base, xe_bo_size(bo), vma); > } > > static const struct file_operations xe_oa_fops = { >-- >2.48.1 > ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 3/3] drm/xe/oa: Implement Wa_14026633728 2026-04-07 3:02 [PATCH 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit 2026-04-07 3:02 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 2026-04-07 3:02 ` [PATCH 2/3] drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap Ashutosh Dixit @ 2026-04-07 3:02 ` Ashutosh Dixit 2026-04-07 23:17 ` Umesh Nerlige Ramappa 2026-04-07 3:10 ` ✓ CI.KUnit: success for drm/xe/oa: Wa_14026633728 Patchwork ` (2 subsequent siblings) 5 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-07 3:02 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa CRI Wa_14026633728 requires MERTOA buffer to be allocated in device memory, not system memory (which is the default for OA buffers). Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> --- drivers/gpu/drm/xe/xe_oa.c | 10 +++++++++- drivers/gpu/drm/xe/xe_wa_oob.rules | 1 + 2 files changed, 10 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index 0e65141e57ee8..f65cec62a0d2d 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -250,6 +250,11 @@ static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); } +static bool mert_wa_14026633728(struct xe_oa_stream *s) +{ + return s->oa_unit->type == DRM_XE_OA_UNIT_TYPE_MERT && XE_GT_WA(s->gt, 14026633728); +} + static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) { u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); @@ -892,11 +897,14 @@ static void xe_oa_stream_destroy(struct xe_oa_stream *stream) static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) { + u32 vram = mert_wa_14026633728(stream) ? + XE_BO_FLAG_VRAM_IF_DGFX(xe_device_get_root_tile(stream->oa->xe)) : + XE_BO_FLAG_SYSTEM; struct xe_bo *bo; bo = xe_bo_create_pin_map_novm(stream->oa->xe, stream->gt->tile, size, ttm_bo_type_kernel, - XE_BO_FLAG_SYSTEM | XE_BO_FLAG_GGTT, false); + vram | XE_BO_FLAG_GGTT, false); if (IS_ERR(bo)) return PTR_ERR(bo); diff --git a/drivers/gpu/drm/xe/xe_wa_oob.rules b/drivers/gpu/drm/xe/xe_wa_oob.rules index f8a185103b805..a7c1bd9bcb943 100644 --- a/drivers/gpu/drm/xe/xe_wa_oob.rules +++ b/drivers/gpu/drm/xe/xe_wa_oob.rules @@ -65,3 +65,4 @@ 14025883347 MEDIA_VERSION_RANGE(1301, 3503) GRAPHICS_VERSION_RANGE(2004, 3005) +14026633728 PLATFORM(CRESCENTISLAND) -- 2.48.1 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 3/3] drm/xe/oa: Implement Wa_14026633728 2026-04-07 3:02 ` [PATCH 3/3] drm/xe/oa: Implement Wa_14026633728 Ashutosh Dixit @ 2026-04-07 23:17 ` Umesh Nerlige Ramappa 0 siblings, 0 replies; 30+ messages in thread From: Umesh Nerlige Ramappa @ 2026-04-07 23:17 UTC (permalink / raw) To: Ashutosh Dixit; +Cc: intel-xe On Mon, Apr 06, 2026 at 08:02:19PM -0700, Ashutosh Dixit wrote: >CRI Wa_14026633728 requires MERTOA buffer to be allocated in device memory, >not system memory (which is the default for OA buffers). > >Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> LGTM, Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com> >--- > drivers/gpu/drm/xe/xe_oa.c | 10 +++++++++- > drivers/gpu/drm/xe/xe_wa_oob.rules | 1 + > 2 files changed, 10 insertions(+), 1 deletion(-) > >diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c >index 0e65141e57ee8..f65cec62a0d2d 100644 >--- a/drivers/gpu/drm/xe/xe_oa.c >+++ b/drivers/gpu/drm/xe/xe_oa.c >@@ -250,6 +250,11 @@ static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) > xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); > } > >+static bool mert_wa_14026633728(struct xe_oa_stream *s) >+{ >+ return s->oa_unit->type == DRM_XE_OA_UNIT_TYPE_MERT && XE_GT_WA(s->gt, 14026633728); >+} >+ > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > { > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); >@@ -892,11 +897,14 @@ static void xe_oa_stream_destroy(struct xe_oa_stream *stream) > > static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > { >+ u32 vram = mert_wa_14026633728(stream) ? >+ XE_BO_FLAG_VRAM_IF_DGFX(xe_device_get_root_tile(stream->oa->xe)) : >+ XE_BO_FLAG_SYSTEM; > struct xe_bo *bo; > > bo = xe_bo_create_pin_map_novm(stream->oa->xe, stream->gt->tile, > size, ttm_bo_type_kernel, >- XE_BO_FLAG_SYSTEM | XE_BO_FLAG_GGTT, false); >+ vram | XE_BO_FLAG_GGTT, false); > if (IS_ERR(bo)) > return PTR_ERR(bo); > >diff --git a/drivers/gpu/drm/xe/xe_wa_oob.rules b/drivers/gpu/drm/xe/xe_wa_oob.rules >index f8a185103b805..a7c1bd9bcb943 100644 >--- a/drivers/gpu/drm/xe/xe_wa_oob.rules >+++ b/drivers/gpu/drm/xe/xe_wa_oob.rules >@@ -65,3 +65,4 @@ > > 14025883347 MEDIA_VERSION_RANGE(1301, 3503) > GRAPHICS_VERSION_RANGE(2004, 3005) >+14026633728 PLATFORM(CRESCENTISLAND) >-- >2.48.1 > ^ permalink raw reply [flat|nested] 30+ messages in thread
* ✓ CI.KUnit: success for drm/xe/oa: Wa_14026633728 2026-04-07 3:02 [PATCH 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit ` (2 preceding siblings ...) 2026-04-07 3:02 ` [PATCH 3/3] drm/xe/oa: Implement Wa_14026633728 Ashutosh Dixit @ 2026-04-07 3:10 ` Patchwork 2026-04-07 3:51 ` ✓ Xe.CI.BAT: " Patchwork 2026-04-07 5:02 ` ✓ Xe.CI.FULL: " Patchwork 5 siblings, 0 replies; 30+ messages in thread From: Patchwork @ 2026-04-07 3:10 UTC (permalink / raw) To: Ashutosh Dixit; +Cc: intel-xe == Series Details == Series: drm/xe/oa: Wa_14026633728 URL : https://patchwork.freedesktop.org/series/164412/ State : success == Summary == + trap cleanup EXIT + /kernel/tools/testing/kunit/kunit.py run --kunitconfig /kernel/drivers/gpu/drm/xe/.kunitconfig [03:09:08] Configuring KUnit Kernel ... Generating .config ... Populating config with: $ make ARCH=um O=.kunit olddefconfig [03:09:13] Building KUnit Kernel ... Populating config with: $ make ARCH=um O=.kunit olddefconfig Building with: $ make all compile_commands.json scripts_gdb ARCH=um O=.kunit --jobs=48 [03:09:44] Starting KUnit Kernel (1/1)... [03:09:44] ============================================================ Running tests with: $ .kunit/linux kunit.enable=1 mem=1G console=tty kunit_shutdown=halt [03:09:44] ================== guc_buf (11 subtests) =================== [03:09:44] [PASSED] test_smallest [03:09:44] [PASSED] test_largest [03:09:44] [PASSED] test_granular [03:09:44] [PASSED] test_unique [03:09:44] [PASSED] test_overlap [03:09:44] [PASSED] test_reusable [03:09:44] [PASSED] test_too_big [03:09:44] [PASSED] test_flush [03:09:44] [PASSED] test_lookup [03:09:44] [PASSED] test_data [03:09:44] [PASSED] test_class [03:09:44] ===================== [PASSED] guc_buf ===================== [03:09:44] =================== guc_dbm (7 subtests) =================== [03:09:44] [PASSED] test_empty [03:09:44] [PASSED] test_default [03:09:44] ======================== test_size ======================== [03:09:44] [PASSED] 4 [03:09:44] [PASSED] 8 [03:09:44] [PASSED] 32 [03:09:44] [PASSED] 256 [03:09:44] ==================== [PASSED] test_size ==================== [03:09:44] ======================= test_reuse ======================== [03:09:44] [PASSED] 4 [03:09:44] [PASSED] 8 [03:09:44] [PASSED] 32 [03:09:44] [PASSED] 256 [03:09:44] =================== [PASSED] test_reuse ==================== [03:09:44] =================== test_range_overlap ==================== [03:09:44] [PASSED] 4 [03:09:44] [PASSED] 8 [03:09:44] [PASSED] 32 [03:09:44] [PASSED] 256 [03:09:44] =============== [PASSED] test_range_overlap ================ [03:09:44] =================== test_range_compact ==================== [03:09:44] [PASSED] 4 [03:09:44] [PASSED] 8 [03:09:44] [PASSED] 32 [03:09:44] [PASSED] 256 [03:09:44] =============== [PASSED] test_range_compact ================ [03:09:44] ==================== test_range_spare ===================== [03:09:44] [PASSED] 4 [03:09:44] [PASSED] 8 [03:09:44] [PASSED] 32 [03:09:44] [PASSED] 256 [03:09:44] ================ [PASSED] test_range_spare ================= [03:09:44] ===================== [PASSED] guc_dbm ===================== [03:09:44] =================== guc_idm (6 subtests) =================== [03:09:44] [PASSED] bad_init [03:09:44] [PASSED] no_init [03:09:44] [PASSED] init_fini [03:09:44] [PASSED] check_used [03:09:44] [PASSED] check_quota [03:09:44] [PASSED] check_all [03:09:44] ===================== [PASSED] guc_idm ===================== [03:09:44] ================== no_relay (3 subtests) =================== [03:09:44] [PASSED] xe_drops_guc2pf_if_not_ready [03:09:44] [PASSED] xe_drops_guc2vf_if_not_ready [03:09:44] [PASSED] xe_rejects_send_if_not_ready [03:09:44] ==================== [PASSED] no_relay ===================== [03:09:44] ================== pf_relay (14 subtests) ================== [03:09:44] [PASSED] pf_rejects_guc2pf_too_short [03:09:44] [PASSED] pf_rejects_guc2pf_too_long [03:09:44] [PASSED] pf_rejects_guc2pf_no_payload [03:09:44] [PASSED] pf_fails_no_payload [03:09:44] [PASSED] pf_fails_bad_origin [03:09:44] [PASSED] pf_fails_bad_type [03:09:44] [PASSED] pf_txn_reports_error [03:09:44] [PASSED] pf_txn_sends_pf2guc [03:09:44] [PASSED] pf_sends_pf2guc [03:09:44] [SKIPPED] pf_loopback_nop [03:09:44] [SKIPPED] pf_loopback_echo [03:09:44] [SKIPPED] pf_loopback_fail [03:09:44] [SKIPPED] pf_loopback_busy [03:09:44] [SKIPPED] pf_loopback_retry [03:09:44] ==================== [PASSED] pf_relay ===================== [03:09:44] ================== vf_relay (3 subtests) =================== [03:09:44] [PASSED] vf_rejects_guc2vf_too_short [03:09:44] [PASSED] vf_rejects_guc2vf_too_long [03:09:44] [PASSED] vf_rejects_guc2vf_no_payload [03:09:44] ==================== [PASSED] vf_relay ===================== [03:09:44] ================ pf_gt_config (9 subtests) ================= [03:09:44] [PASSED] fair_contexts_1vf [03:09:44] [PASSED] fair_doorbells_1vf [03:09:44] [PASSED] fair_ggtt_1vf [03:09:44] ====================== fair_vram_1vf ====================== [03:09:44] [PASSED] 3.50 GiB [03:09:44] [PASSED] 11.5 GiB [03:09:44] [PASSED] 15.5 GiB [03:09:44] [PASSED] 31.5 GiB [03:09:44] [PASSED] 63.5 GiB [03:09:44] [PASSED] 1.91 GiB [03:09:44] ================== [PASSED] fair_vram_1vf ================== [03:09:44] ================ fair_vram_1vf_admin_only ================= [03:09:44] [PASSED] 3.50 GiB [03:09:44] [PASSED] 11.5 GiB [03:09:44] [PASSED] 15.5 GiB [03:09:44] [PASSED] 31.5 GiB [03:09:44] [PASSED] 63.5 GiB [03:09:44] [PASSED] 1.91 GiB [03:09:44] ============ [PASSED] fair_vram_1vf_admin_only ============= [03:09:44] ====================== fair_contexts ====================== [03:09:44] [PASSED] 1 VF [03:09:44] [PASSED] 2 VFs [03:09:44] [PASSED] 3 VFs [03:09:44] [PASSED] 4 VFs [03:09:44] [PASSED] 5 VFs [03:09:44] [PASSED] 6 VFs [03:09:44] [PASSED] 7 VFs [03:09:44] [PASSED] 8 VFs [03:09:44] [PASSED] 9 VFs [03:09:44] [PASSED] 10 VFs [03:09:44] [PASSED] 11 VFs [03:09:44] [PASSED] 12 VFs [03:09:44] [PASSED] 13 VFs [03:09:44] [PASSED] 14 VFs [03:09:44] [PASSED] 15 VFs [03:09:44] [PASSED] 16 VFs [03:09:44] [PASSED] 17 VFs [03:09:44] [PASSED] 18 VFs [03:09:44] [PASSED] 19 VFs [03:09:44] [PASSED] 20 VFs [03:09:44] [PASSED] 21 VFs [03:09:44] [PASSED] 22 VFs [03:09:44] [PASSED] 23 VFs [03:09:44] [PASSED] 24 VFs [03:09:44] [PASSED] 25 VFs [03:09:44] [PASSED] 26 VFs [03:09:44] [PASSED] 27 VFs [03:09:44] [PASSED] 28 VFs [03:09:44] [PASSED] 29 VFs [03:09:44] [PASSED] 30 VFs [03:09:44] [PASSED] 31 VFs [03:09:44] [PASSED] 32 VFs [03:09:44] [PASSED] 33 VFs [03:09:44] [PASSED] 34 VFs [03:09:44] [PASSED] 35 VFs [03:09:44] [PASSED] 36 VFs [03:09:44] [PASSED] 37 VFs [03:09:44] [PASSED] 38 VFs [03:09:44] [PASSED] 39 VFs [03:09:44] [PASSED] 40 VFs [03:09:44] [PASSED] 41 VFs [03:09:44] [PASSED] 42 VFs [03:09:44] [PASSED] 43 VFs [03:09:44] [PASSED] 44 VFs [03:09:44] [PASSED] 45 VFs [03:09:44] [PASSED] 46 VFs [03:09:44] [PASSED] 47 VFs [03:09:44] [PASSED] 48 VFs [03:09:44] [PASSED] 49 VFs [03:09:44] [PASSED] 50 VFs [03:09:44] [PASSED] 51 VFs [03:09:44] [PASSED] 52 VFs [03:09:44] [PASSED] 53 VFs [03:09:44] [PASSED] 54 VFs [03:09:44] [PASSED] 55 VFs [03:09:44] [PASSED] 56 VFs [03:09:44] [PASSED] 57 VFs [03:09:44] [PASSED] 58 VFs [03:09:44] [PASSED] 59 VFs [03:09:44] [PASSED] 60 VFs [03:09:44] [PASSED] 61 VFs [03:09:44] [PASSED] 62 VFs [03:09:44] [PASSED] 63 VFs [03:09:44] ================== [PASSED] fair_contexts ================== [03:09:44] ===================== fair_doorbells ====================== [03:09:44] [PASSED] 1 VF [03:09:44] [PASSED] 2 VFs [03:09:44] [PASSED] 3 VFs [03:09:44] [PASSED] 4 VFs [03:09:44] [PASSED] 5 VFs [03:09:44] [PASSED] 6 VFs [03:09:44] [PASSED] 7 VFs [03:09:44] [PASSED] 8 VFs [03:09:44] [PASSED] 9 VFs [03:09:44] [PASSED] 10 VFs [03:09:44] [PASSED] 11 VFs [03:09:44] [PASSED] 12 VFs [03:09:44] [PASSED] 13 VFs [03:09:44] [PASSED] 14 VFs [03:09:44] [PASSED] 15 VFs [03:09:44] [PASSED] 16 VFs [03:09:44] [PASSED] 17 VFs [03:09:44] [PASSED] 18 VFs [03:09:44] [PASSED] 19 VFs [03:09:44] [PASSED] 20 VFs [03:09:44] [PASSED] 21 VFs [03:09:44] [PASSED] 22 VFs [03:09:44] [PASSED] 23 VFs [03:09:44] [PASSED] 24 VFs [03:09:44] [PASSED] 25 VFs [03:09:44] [PASSED] 26 VFs [03:09:44] [PASSED] 27 VFs [03:09:44] [PASSED] 28 VFs [03:09:44] [PASSED] 29 VFs [03:09:44] [PASSED] 30 VFs [03:09:44] [PASSED] 31 VFs [03:09:44] [PASSED] 32 VFs [03:09:44] [PASSED] 33 VFs [03:09:44] [PASSED] 34 VFs [03:09:44] [PASSED] 35 VFs [03:09:44] [PASSED] 36 VFs [03:09:44] [PASSED] 37 VFs [03:09:44] [PASSED] 38 VFs [03:09:44] [PASSED] 39 VFs [03:09:44] [PASSED] 40 VFs [03:09:44] [PASSED] 41 VFs [03:09:44] [PASSED] 42 VFs [03:09:44] [PASSED] 43 VFs [03:09:44] [PASSED] 44 VFs [03:09:44] [PASSED] 45 VFs [03:09:44] [PASSED] 46 VFs [03:09:44] [PASSED] 47 VFs [03:09:44] [PASSED] 48 VFs [03:09:44] [PASSED] 49 VFs [03:09:44] [PASSED] 50 VFs [03:09:44] [PASSED] 51 VFs [03:09:44] [PASSED] 52 VFs [03:09:44] [PASSED] 53 VFs [03:09:44] [PASSED] 54 VFs [03:09:44] [PASSED] 55 VFs [03:09:44] [PASSED] 56 VFs [03:09:44] [PASSED] 57 VFs [03:09:44] [PASSED] 58 VFs [03:09:44] [PASSED] 59 VFs [03:09:44] [PASSED] 60 VFs [03:09:44] [PASSED] 61 VFs [03:09:44] [PASSED] 62 VFs [03:09:44] [PASSED] 63 VFs [03:09:44] ================= [PASSED] fair_doorbells ================== [03:09:44] ======================== fair_ggtt ======================== [03:09:44] [PASSED] 1 VF [03:09:44] [PASSED] 2 VFs [03:09:44] [PASSED] 3 VFs [03:09:44] [PASSED] 4 VFs [03:09:44] [PASSED] 5 VFs [03:09:44] [PASSED] 6 VFs [03:09:44] [PASSED] 7 VFs [03:09:44] [PASSED] 8 VFs [03:09:44] [PASSED] 9 VFs [03:09:44] [PASSED] 10 VFs [03:09:44] [PASSED] 11 VFs [03:09:44] [PASSED] 12 VFs [03:09:44] [PASSED] 13 VFs [03:09:44] [PASSED] 14 VFs [03:09:44] [PASSED] 15 VFs [03:09:44] [PASSED] 16 VFs [03:09:44] [PASSED] 17 VFs [03:09:44] [PASSED] 18 VFs [03:09:44] [PASSED] 19 VFs [03:09:44] [PASSED] 20 VFs [03:09:44] [PASSED] 21 VFs [03:09:44] [PASSED] 22 VFs [03:09:44] [PASSED] 23 VFs [03:09:44] [PASSED] 24 VFs [03:09:44] [PASSED] 25 VFs [03:09:44] [PASSED] 26 VFs [03:09:44] [PASSED] 27 VFs [03:09:44] [PASSED] 28 VFs [03:09:44] [PASSED] 29 VFs [03:09:44] [PASSED] 30 VFs [03:09:44] [PASSED] 31 VFs [03:09:44] [PASSED] 32 VFs [03:09:44] [PASSED] 33 VFs [03:09:44] [PASSED] 34 VFs [03:09:44] [PASSED] 35 VFs [03:09:44] [PASSED] 36 VFs [03:09:44] [PASSED] 37 VFs [03:09:44] [PASSED] 38 VFs [03:09:44] [PASSED] 39 VFs [03:09:44] [PASSED] 40 VFs [03:09:44] [PASSED] 41 VFs [03:09:44] [PASSED] 42 VFs [03:09:44] [PASSED] 43 VFs [03:09:44] [PASSED] 44 VFs [03:09:44] [PASSED] 45 VFs [03:09:44] [PASSED] 46 VFs [03:09:44] [PASSED] 47 VFs [03:09:44] [PASSED] 48 VFs [03:09:44] [PASSED] 49 VFs [03:09:44] [PASSED] 50 VFs [03:09:44] [PASSED] 51 VFs [03:09:44] [PASSED] 52 VFs [03:09:44] [PASSED] 53 VFs [03:09:44] [PASSED] 54 VFs [03:09:44] [PASSED] 55 VFs [03:09:44] [PASSED] 56 VFs [03:09:44] [PASSED] 57 VFs [03:09:44] [PASSED] 58 VFs [03:09:44] [PASSED] 59 VFs [03:09:44] [PASSED] 60 VFs [03:09:44] [PASSED] 61 VFs [03:09:44] [PASSED] 62 VFs [03:09:44] [PASSED] 63 VFs [03:09:44] ==================== [PASSED] fair_ggtt ==================== [03:09:45] ======================== fair_vram ======================== [03:09:45] [PASSED] 1 VF [03:09:45] [PASSED] 2 VFs [03:09:45] [PASSED] 3 VFs [03:09:45] [PASSED] 4 VFs [03:09:45] [PASSED] 5 VFs [03:09:45] [PASSED] 6 VFs [03:09:45] [PASSED] 7 VFs [03:09:45] [PASSED] 8 VFs [03:09:45] [PASSED] 9 VFs [03:09:45] [PASSED] 10 VFs [03:09:45] [PASSED] 11 VFs [03:09:45] [PASSED] 12 VFs [03:09:45] [PASSED] 13 VFs [03:09:45] [PASSED] 14 VFs [03:09:45] [PASSED] 15 VFs [03:09:45] [PASSED] 16 VFs [03:09:45] [PASSED] 17 VFs [03:09:45] [PASSED] 18 VFs [03:09:45] [PASSED] 19 VFs [03:09:45] [PASSED] 20 VFs [03:09:45] [PASSED] 21 VFs [03:09:45] [PASSED] 22 VFs [03:09:45] [PASSED] 23 VFs [03:09:45] [PASSED] 24 VFs [03:09:45] [PASSED] 25 VFs [03:09:45] [PASSED] 26 VFs [03:09:45] [PASSED] 27 VFs [03:09:45] [PASSED] 28 VFs [03:09:45] [PASSED] 29 VFs [03:09:45] [PASSED] 30 VFs [03:09:45] [PASSED] 31 VFs [03:09:45] [PASSED] 32 VFs [03:09:45] [PASSED] 33 VFs [03:09:45] [PASSED] 34 VFs [03:09:45] [PASSED] 35 VFs [03:09:45] [PASSED] 36 VFs [03:09:45] [PASSED] 37 VFs [03:09:45] [PASSED] 38 VFs [03:09:45] [PASSED] 39 VFs [03:09:45] [PASSED] 40 VFs [03:09:45] [PASSED] 41 VFs [03:09:45] [PASSED] 42 VFs [03:09:45] [PASSED] 43 VFs [03:09:45] [PASSED] 44 VFs [03:09:45] [PASSED] 45 VFs [03:09:45] [PASSED] 46 VFs [03:09:45] [PASSED] 47 VFs [03:09:45] [PASSED] 48 VFs [03:09:45] [PASSED] 49 VFs [03:09:45] [PASSED] 50 VFs [03:09:45] [PASSED] 51 VFs [03:09:45] [PASSED] 52 VFs [03:09:45] [PASSED] 53 VFs [03:09:45] [PASSED] 54 VFs [03:09:45] [PASSED] 55 VFs [03:09:45] [PASSED] 56 VFs [03:09:45] [PASSED] 57 VFs [03:09:45] [PASSED] 58 VFs [03:09:45] [PASSED] 59 VFs [03:09:45] [PASSED] 60 VFs [03:09:45] [PASSED] 61 VFs [03:09:45] [PASSED] 62 VFs [03:09:45] [PASSED] 63 VFs [03:09:45] ==================== [PASSED] fair_vram ==================== [03:09:45] ================== [PASSED] pf_gt_config =================== [03:09:45] ===================== lmtt (1 subtest) ===================== [03:09:45] ======================== test_ops ========================= [03:09:45] [PASSED] 2-level [03:09:45] [PASSED] multi-level [03:09:45] ==================== [PASSED] test_ops ===================== [03:09:45] ====================== [PASSED] lmtt ======================= [03:09:45] ================= pf_service (11 subtests) ================= [03:09:45] [PASSED] pf_negotiate_any [03:09:45] [PASSED] pf_negotiate_base_match [03:09:45] [PASSED] pf_negotiate_base_newer [03:09:45] [PASSED] pf_negotiate_base_next [03:09:45] [SKIPPED] pf_negotiate_base_older [03:09:45] [PASSED] pf_negotiate_base_prev [03:09:45] [PASSED] pf_negotiate_latest_match [03:09:45] [PASSED] pf_negotiate_latest_newer [03:09:45] [PASSED] pf_negotiate_latest_next [03:09:45] [SKIPPED] pf_negotiate_latest_older [03:09:45] [SKIPPED] pf_negotiate_latest_prev [03:09:45] =================== [PASSED] pf_service ==================== [03:09:45] ================= xe_guc_g2g (2 subtests) ================== [03:09:45] ============== xe_live_guc_g2g_kunit_default ============== [03:09:45] ========= [SKIPPED] xe_live_guc_g2g_kunit_default ========== [03:09:45] ============== xe_live_guc_g2g_kunit_allmem =============== [03:09:45] ========== [SKIPPED] xe_live_guc_g2g_kunit_allmem ========== [03:09:45] =================== [SKIPPED] xe_guc_g2g =================== [03:09:45] =================== xe_mocs (2 subtests) =================== [03:09:45] ================ xe_live_mocs_kernel_kunit ================ [03:09:45] =========== [SKIPPED] xe_live_mocs_kernel_kunit ============ [03:09:45] ================ xe_live_mocs_reset_kunit ================= [03:09:45] ============ [SKIPPED] xe_live_mocs_reset_kunit ============ [03:09:45] ==================== [SKIPPED] xe_mocs ===================== [03:09:45] ================= xe_migrate (2 subtests) ================== [03:09:45] ================= xe_migrate_sanity_kunit ================= [03:09:45] ============ [SKIPPED] xe_migrate_sanity_kunit ============= [03:09:45] ================== xe_validate_ccs_kunit ================== [03:09:45] ============= [SKIPPED] xe_validate_ccs_kunit ============== [03:09:45] =================== [SKIPPED] xe_migrate =================== [03:09:45] ================== xe_dma_buf (1 subtest) ================== [03:09:45] ==================== xe_dma_buf_kunit ===================== [03:09:45] ================ [SKIPPED] xe_dma_buf_kunit ================ [03:09:45] =================== [SKIPPED] xe_dma_buf =================== [03:09:45] ================= xe_bo_shrink (1 subtest) ================= [03:09:45] =================== xe_bo_shrink_kunit ==================== [03:09:45] =============== [SKIPPED] xe_bo_shrink_kunit =============== [03:09:45] ================== [SKIPPED] xe_bo_shrink ================== [03:09:45] ==================== xe_bo (2 subtests) ==================== [03:09:45] ================== xe_ccs_migrate_kunit =================== [03:09:45] ============== [SKIPPED] xe_ccs_migrate_kunit ============== [03:09:45] ==================== xe_bo_evict_kunit ==================== [03:09:45] =============== [SKIPPED] xe_bo_evict_kunit ================ [03:09:45] ===================== [SKIPPED] xe_bo ====================== [03:09:45] ==================== args (13 subtests) ==================== [03:09:45] [PASSED] count_args_test [03:09:45] [PASSED] call_args_example [03:09:45] [PASSED] call_args_test [03:09:45] [PASSED] drop_first_arg_example [03:09:45] [PASSED] drop_first_arg_test [03:09:45] [PASSED] first_arg_example [03:09:45] [PASSED] first_arg_test [03:09:45] [PASSED] last_arg_example [03:09:45] [PASSED] last_arg_test [03:09:45] [PASSED] pick_arg_example [03:09:45] [PASSED] if_args_example [03:09:45] [PASSED] if_args_test [03:09:45] [PASSED] sep_comma_example [03:09:45] ====================== [PASSED] args ======================= [03:09:45] =================== xe_pci (3 subtests) ==================== [03:09:45] ==================== check_graphics_ip ==================== [03:09:45] [PASSED] 12.00 Xe_LP [03:09:45] [PASSED] 12.10 Xe_LP+ [03:09:45] [PASSED] 12.55 Xe_HPG [03:09:45] [PASSED] 12.60 Xe_HPC [03:09:45] [PASSED] 12.70 Xe_LPG [03:09:45] [PASSED] 12.71 Xe_LPG [03:09:45] [PASSED] 12.74 Xe_LPG+ [03:09:45] [PASSED] 20.01 Xe2_HPG [03:09:45] [PASSED] 20.02 Xe2_HPG [03:09:45] [PASSED] 20.04 Xe2_LPG [03:09:45] [PASSED] 30.00 Xe3_LPG [03:09:45] [PASSED] 30.01 Xe3_LPG [03:09:45] [PASSED] 30.03 Xe3_LPG [03:09:45] [PASSED] 30.04 Xe3_LPG [03:09:45] [PASSED] 30.05 Xe3_LPG [03:09:45] [PASSED] 35.10 Xe3p_LPG [03:09:45] [PASSED] 35.11 Xe3p_XPC [03:09:45] ================ [PASSED] check_graphics_ip ================ [03:09:45] ===================== check_media_ip ====================== [03:09:45] [PASSED] 12.00 Xe_M [03:09:45] [PASSED] 12.55 Xe_HPM [03:09:45] [PASSED] 13.00 Xe_LPM+ [03:09:45] [PASSED] 13.01 Xe2_HPM [03:09:45] [PASSED] 20.00 Xe2_LPM [03:09:45] [PASSED] 30.00 Xe3_LPM [03:09:45] [PASSED] 30.02 Xe3_LPM [03:09:45] [PASSED] 35.00 Xe3p_LPM [03:09:45] [PASSED] 35.03 Xe3p_HPM [03:09:45] ================= [PASSED] check_media_ip ================== [03:09:45] =================== check_platform_desc =================== [03:09:45] [PASSED] 0x9A60 (TIGERLAKE) [03:09:45] [PASSED] 0x9A68 (TIGERLAKE) [03:09:45] [PASSED] 0x9A70 (TIGERLAKE) [03:09:45] [PASSED] 0x9A40 (TIGERLAKE) [03:09:45] [PASSED] 0x9A49 (TIGERLAKE) [03:09:45] [PASSED] 0x9A59 (TIGERLAKE) [03:09:45] [PASSED] 0x9A78 (TIGERLAKE) [03:09:45] [PASSED] 0x9AC0 (TIGERLAKE) [03:09:45] [PASSED] 0x9AC9 (TIGERLAKE) [03:09:45] [PASSED] 0x9AD9 (TIGERLAKE) [03:09:45] [PASSED] 0x9AF8 (TIGERLAKE) [03:09:45] [PASSED] 0x4C80 (ROCKETLAKE) [03:09:45] [PASSED] 0x4C8A (ROCKETLAKE) [03:09:45] [PASSED] 0x4C8B (ROCKETLAKE) [03:09:45] [PASSED] 0x4C8C (ROCKETLAKE) [03:09:45] [PASSED] 0x4C90 (ROCKETLAKE) [03:09:45] [PASSED] 0x4C9A (ROCKETLAKE) [03:09:45] [PASSED] 0x4680 (ALDERLAKE_S) [03:09:45] [PASSED] 0x4682 (ALDERLAKE_S) [03:09:45] [PASSED] 0x4688 (ALDERLAKE_S) [03:09:45] [PASSED] 0x468A (ALDERLAKE_S) [03:09:45] [PASSED] 0x468B (ALDERLAKE_S) [03:09:45] [PASSED] 0x4690 (ALDERLAKE_S) [03:09:45] [PASSED] 0x4692 (ALDERLAKE_S) [03:09:45] [PASSED] 0x4693 (ALDERLAKE_S) [03:09:45] [PASSED] 0x46A0 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46A1 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46A2 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46A3 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46A6 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46A8 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46AA (ALDERLAKE_P) [03:09:45] [PASSED] 0x462A (ALDERLAKE_P) [03:09:45] [PASSED] 0x4626 (ALDERLAKE_P) [03:09:45] [PASSED] 0x4628 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46B0 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46B1 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46B2 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46B3 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46C0 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46C1 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46C2 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46C3 (ALDERLAKE_P) [03:09:45] [PASSED] 0x46D0 (ALDERLAKE_N) [03:09:45] [PASSED] 0x46D1 (ALDERLAKE_N) [03:09:45] [PASSED] 0x46D2 (ALDERLAKE_N) [03:09:45] [PASSED] 0x46D3 (ALDERLAKE_N) [03:09:45] [PASSED] 0x46D4 (ALDERLAKE_N) [03:09:45] [PASSED] 0xA721 (ALDERLAKE_P) [03:09:45] [PASSED] 0xA7A1 (ALDERLAKE_P) [03:09:45] [PASSED] 0xA7A9 (ALDERLAKE_P) [03:09:45] [PASSED] 0xA7AC (ALDERLAKE_P) [03:09:45] [PASSED] 0xA7AD (ALDERLAKE_P) [03:09:45] [PASSED] 0xA720 (ALDERLAKE_P) [03:09:45] [PASSED] 0xA7A0 (ALDERLAKE_P) [03:09:45] [PASSED] 0xA7A8 (ALDERLAKE_P) [03:09:45] [PASSED] 0xA7AA (ALDERLAKE_P) [03:09:45] [PASSED] 0xA7AB (ALDERLAKE_P) [03:09:45] [PASSED] 0xA780 (ALDERLAKE_S) [03:09:45] [PASSED] 0xA781 (ALDERLAKE_S) [03:09:45] [PASSED] 0xA782 (ALDERLAKE_S) [03:09:45] [PASSED] 0xA783 (ALDERLAKE_S) [03:09:45] [PASSED] 0xA788 (ALDERLAKE_S) [03:09:45] [PASSED] 0xA789 (ALDERLAKE_S) [03:09:45] [PASSED] 0xA78A (ALDERLAKE_S) [03:09:45] [PASSED] 0xA78B (ALDERLAKE_S) [03:09:45] [PASSED] 0x4905 (DG1) [03:09:45] [PASSED] 0x4906 (DG1) [03:09:45] [PASSED] 0x4907 (DG1) [03:09:45] [PASSED] 0x4908 (DG1) [03:09:45] [PASSED] 0x4909 (DG1) [03:09:45] [PASSED] 0x56C0 (DG2) [03:09:45] [PASSED] 0x56C2 (DG2) [03:09:45] [PASSED] 0x56C1 (DG2) [03:09:45] [PASSED] 0x7D51 (METEORLAKE) [03:09:45] [PASSED] 0x7DD1 (METEORLAKE) [03:09:45] [PASSED] 0x7D41 (METEORLAKE) [03:09:45] [PASSED] 0x7D67 (METEORLAKE) [03:09:45] [PASSED] 0xB640 (METEORLAKE) [03:09:45] [PASSED] 0x56A0 (DG2) [03:09:45] [PASSED] 0x56A1 (DG2) [03:09:45] [PASSED] 0x56A2 (DG2) [03:09:45] [PASSED] 0x56BE (DG2) [03:09:45] [PASSED] 0x56BF (DG2) [03:09:45] [PASSED] 0x5690 (DG2) [03:09:45] [PASSED] 0x5691 (DG2) [03:09:45] [PASSED] 0x5692 (DG2) [03:09:45] [PASSED] 0x56A5 (DG2) [03:09:45] [PASSED] 0x56A6 (DG2) [03:09:45] [PASSED] 0x56B0 (DG2) [03:09:45] [PASSED] 0x56B1 (DG2) [03:09:45] [PASSED] 0x56BA (DG2) [03:09:45] [PASSED] 0x56BB (DG2) [03:09:45] [PASSED] 0x56BC (DG2) [03:09:45] [PASSED] 0x56BD (DG2) [03:09:45] [PASSED] 0x5693 (DG2) [03:09:45] [PASSED] 0x5694 (DG2) [03:09:45] [PASSED] 0x5695 (DG2) [03:09:45] [PASSED] 0x56A3 (DG2) [03:09:45] [PASSED] 0x56A4 (DG2) [03:09:45] [PASSED] 0x56B2 (DG2) [03:09:45] [PASSED] 0x56B3 (DG2) [03:09:45] [PASSED] 0x5696 (DG2) [03:09:45] [PASSED] 0x5697 (DG2) [03:09:45] [PASSED] 0xB69 (PVC) [03:09:45] [PASSED] 0xB6E (PVC) [03:09:45] [PASSED] 0xBD4 (PVC) [03:09:45] [PASSED] 0xBD5 (PVC) [03:09:45] [PASSED] 0xBD6 (PVC) [03:09:45] [PASSED] 0xBD7 (PVC) [03:09:45] [PASSED] 0xBD8 (PVC) [03:09:45] [PASSED] 0xBD9 (PVC) [03:09:45] [PASSED] 0xBDA (PVC) [03:09:45] [PASSED] 0xBDB (PVC) [03:09:45] [PASSED] 0xBE0 (PVC) [03:09:45] [PASSED] 0xBE1 (PVC) [03:09:45] [PASSED] 0xBE5 (PVC) [03:09:45] [PASSED] 0x7D40 (METEORLAKE) [03:09:45] [PASSED] 0x7D45 (METEORLAKE) [03:09:45] [PASSED] 0x7D55 (METEORLAKE) [03:09:45] [PASSED] 0x7D60 (METEORLAKE) [03:09:45] [PASSED] 0x7DD5 (METEORLAKE) [03:09:45] [PASSED] 0x6420 (LUNARLAKE) [03:09:45] [PASSED] 0x64A0 (LUNARLAKE) [03:09:45] [PASSED] 0x64B0 (LUNARLAKE) [03:09:45] [PASSED] 0xE202 (BATTLEMAGE) [03:09:45] [PASSED] 0xE209 (BATTLEMAGE) [03:09:45] [PASSED] 0xE20B (BATTLEMAGE) [03:09:45] [PASSED] 0xE20C (BATTLEMAGE) [03:09:45] [PASSED] 0xE20D (BATTLEMAGE) [03:09:45] [PASSED] 0xE210 (BATTLEMAGE) [03:09:45] [PASSED] 0xE211 (BATTLEMAGE) [03:09:45] [PASSED] 0xE212 (BATTLEMAGE) [03:09:45] [PASSED] 0xE216 (BATTLEMAGE) [03:09:45] [PASSED] 0xE220 (BATTLEMAGE) [03:09:45] [PASSED] 0xE221 (BATTLEMAGE) [03:09:45] [PASSED] 0xE222 (BATTLEMAGE) [03:09:45] [PASSED] 0xE223 (BATTLEMAGE) [03:09:45] [PASSED] 0xB080 (PANTHERLAKE) [03:09:45] [PASSED] 0xB081 (PANTHERLAKE) [03:09:45] [PASSED] 0xB082 (PANTHERLAKE) [03:09:45] [PASSED] 0xB083 (PANTHERLAKE) [03:09:45] [PASSED] 0xB084 (PANTHERLAKE) [03:09:45] [PASSED] 0xB085 (PANTHERLAKE) [03:09:45] [PASSED] 0xB086 (PANTHERLAKE) [03:09:45] [PASSED] 0xB087 (PANTHERLAKE) [03:09:45] [PASSED] 0xB08F (PANTHERLAKE) [03:09:45] [PASSED] 0xB090 (PANTHERLAKE) [03:09:45] [PASSED] 0xB0A0 (PANTHERLAKE) [03:09:45] [PASSED] 0xB0B0 (PANTHERLAKE) [03:09:45] [PASSED] 0xFD80 (PANTHERLAKE) [03:09:45] [PASSED] 0xFD81 (PANTHERLAKE) [03:09:45] [PASSED] 0xD740 (NOVALAKE_S) [03:09:45] [PASSED] 0xD741 (NOVALAKE_S) [03:09:45] [PASSED] 0xD742 (NOVALAKE_S) [03:09:45] [PASSED] 0xD743 (NOVALAKE_S) [03:09:45] [PASSED] 0xD744 (NOVALAKE_S) [03:09:45] [PASSED] 0xD745 (NOVALAKE_S) [03:09:45] [PASSED] 0x674C (CRESCENTISLAND) [03:09:45] [PASSED] 0xD750 (NOVALAKE_P) [03:09:45] [PASSED] 0xD751 (NOVALAKE_P) [03:09:45] [PASSED] 0xD752 (NOVALAKE_P) [03:09:45] [PASSED] 0xD753 (NOVALAKE_P) [03:09:45] [PASSED] 0xD754 (NOVALAKE_P) [03:09:45] [PASSED] 0xD755 (NOVALAKE_P) [03:09:45] [PASSED] 0xD756 (NOVALAKE_P) [03:09:45] [PASSED] 0xD757 (NOVALAKE_P) [03:09:45] [PASSED] 0xD75F (NOVALAKE_P) [03:09:45] =============== [PASSED] check_platform_desc =============== [03:09:45] ===================== [PASSED] xe_pci ====================== [03:09:45] =================== xe_rtp (2 subtests) ==================== [03:09:45] =============== xe_rtp_process_to_sr_tests ================ [03:09:45] [PASSED] coalesce-same-reg [03:09:45] [PASSED] no-match-no-add [03:09:45] [PASSED] match-or [03:09:45] [PASSED] match-or-xfail [03:09:45] [PASSED] no-match-no-add-multiple-rules [03:09:45] [PASSED] two-regs-two-entries [03:09:45] [PASSED] clr-one-set-other [03:09:45] [PASSED] set-field [03:09:45] [PASSED] conflict-duplicate stty: 'standard input': Inappropriate ioctl for device [03:09:45] [PASSED] conflict-not-disjoint [03:09:45] [PASSED] conflict-reg-type [03:09:45] =========== [PASSED] xe_rtp_process_to_sr_tests ============ [03:09:45] ================== xe_rtp_process_tests =================== [03:09:45] [PASSED] active1 [03:09:45] [PASSED] active2 [03:09:45] [PASSED] active-inactive [03:09:45] [PASSED] inactive-active [03:09:45] [PASSED] inactive-1st_or_active-inactive [03:09:45] [PASSED] inactive-2nd_or_active-inactive [03:09:45] [PASSED] inactive-last_or_active-inactive [03:09:45] [PASSED] inactive-no_or_active-inactive [03:09:45] ============== [PASSED] xe_rtp_process_tests =============== [03:09:45] ===================== [PASSED] xe_rtp ====================== [03:09:45] ==================== xe_wa (1 subtest) ===================== [03:09:45] ======================== xe_wa_gt ========================= [03:09:45] [PASSED] TIGERLAKE B0 [03:09:45] [PASSED] DG1 A0 [03:09:45] [PASSED] DG1 B0 [03:09:45] [PASSED] ALDERLAKE_S A0 [03:09:45] [PASSED] ALDERLAKE_S B0 [03:09:45] [PASSED] ALDERLAKE_S C0 [03:09:45] [PASSED] ALDERLAKE_S D0 [03:09:45] [PASSED] ALDERLAKE_P A0 [03:09:45] [PASSED] ALDERLAKE_P B0 [03:09:45] [PASSED] ALDERLAKE_P C0 [03:09:45] [PASSED] ALDERLAKE_S RPLS D0 [03:09:45] [PASSED] ALDERLAKE_P RPLU E0 [03:09:45] [PASSED] DG2 G10 C0 [03:09:45] [PASSED] DG2 G11 B1 [03:09:45] [PASSED] DG2 G12 A1 [03:09:45] [PASSED] METEORLAKE 12.70(Xe_LPG) A0 13.00(Xe_LPM+) A0 [03:09:45] [PASSED] METEORLAKE 12.71(Xe_LPG) A0 13.00(Xe_LPM+) A0 [03:09:45] [PASSED] METEORLAKE 12.74(Xe_LPG+) A0 13.00(Xe_LPM+) A0 [03:09:45] [PASSED] LUNARLAKE 20.04(Xe2_LPG) A0 20.00(Xe2_LPM) A0 [03:09:45] [PASSED] LUNARLAKE 20.04(Xe2_LPG) B0 20.00(Xe2_LPM) A0 [03:09:45] [PASSED] BATTLEMAGE 20.01(Xe2_HPG) A0 13.01(Xe2_HPM) A1 [03:09:45] [PASSED] PANTHERLAKE 30.00(Xe3_LPG) A0 30.00(Xe3_LPM) A0 [03:09:45] ==================== [PASSED] xe_wa_gt ===================== [03:09:45] ====================== [PASSED] xe_wa ====================== [03:09:45] ============================================================ [03:09:45] Testing complete. Ran 597 tests: passed: 579, skipped: 18 [03:09:45] Elapsed time: 36.240s total, 4.194s configuring, 31.429s building, 0.591s running + /kernel/tools/testing/kunit/kunit.py run --kunitconfig /kernel/drivers/gpu/drm/tests/.kunitconfig [03:09:45] Configuring KUnit Kernel ... Regenerating .config ... Populating config with: $ make ARCH=um O=.kunit olddefconfig [03:09:46] Building KUnit Kernel ... Populating config with: $ make ARCH=um O=.kunit olddefconfig Building with: $ make all compile_commands.json scripts_gdb ARCH=um O=.kunit --jobs=48 [03:10:11] Starting KUnit Kernel (1/1)... [03:10:11] ============================================================ Running tests with: $ .kunit/linux kunit.enable=1 mem=1G console=tty kunit_shutdown=halt [03:10:11] ============ drm_test_pick_cmdline (2 subtests) ============ [03:10:11] [PASSED] drm_test_pick_cmdline_res_1920_1080_60 [03:10:11] =============== drm_test_pick_cmdline_named =============== [03:10:11] [PASSED] NTSC [03:10:11] [PASSED] NTSC-J [03:10:11] [PASSED] PAL [03:10:11] [PASSED] PAL-M [03:10:11] =========== [PASSED] drm_test_pick_cmdline_named =========== [03:10:11] ============== [PASSED] drm_test_pick_cmdline ============== [03:10:11] == drm_test_atomic_get_connector_for_encoder (1 subtest) === [03:10:11] [PASSED] drm_test_drm_atomic_get_connector_for_encoder [03:10:11] ==== [PASSED] drm_test_atomic_get_connector_for_encoder ==== [03:10:11] =========== drm_validate_clone_mode (2 subtests) =========== [03:10:11] ============== drm_test_check_in_clone_mode =============== [03:10:11] [PASSED] in_clone_mode [03:10:11] [PASSED] not_in_clone_mode [03:10:11] ========== [PASSED] drm_test_check_in_clone_mode =========== [03:10:11] =============== drm_test_check_valid_clones =============== [03:10:11] [PASSED] not_in_clone_mode [03:10:11] [PASSED] valid_clone [03:10:11] [PASSED] invalid_clone [03:10:11] =========== [PASSED] drm_test_check_valid_clones =========== [03:10:11] ============= [PASSED] drm_validate_clone_mode ============= [03:10:11] ============= drm_validate_modeset (1 subtest) ============= [03:10:11] [PASSED] drm_test_check_connector_changed_modeset [03:10:11] ============== [PASSED] drm_validate_modeset =============== [03:10:11] ====== drm_test_bridge_get_current_state (2 subtests) ====== [03:10:11] [PASSED] drm_test_drm_bridge_get_current_state_atomic [03:10:11] [PASSED] drm_test_drm_bridge_get_current_state_legacy [03:10:11] ======== [PASSED] drm_test_bridge_get_current_state ======== [03:10:11] ====== drm_test_bridge_helper_reset_crtc (3 subtests) ====== [03:10:11] [PASSED] drm_test_drm_bridge_helper_reset_crtc_atomic [03:10:11] [PASSED] drm_test_drm_bridge_helper_reset_crtc_atomic_disabled [03:10:11] [PASSED] drm_test_drm_bridge_helper_reset_crtc_legacy [03:10:11] ======== [PASSED] drm_test_bridge_helper_reset_crtc ======== [03:10:11] ============== drm_bridge_alloc (2 subtests) =============== [03:10:11] [PASSED] drm_test_drm_bridge_alloc_basic [03:10:11] [PASSED] drm_test_drm_bridge_alloc_get_put [03:10:11] ================ [PASSED] drm_bridge_alloc ================= [03:10:11] ============= drm_cmdline_parser (40 subtests) ============= [03:10:11] [PASSED] drm_test_cmdline_force_d_only [03:10:11] [PASSED] drm_test_cmdline_force_D_only_dvi [03:10:11] [PASSED] drm_test_cmdline_force_D_only_hdmi [03:10:11] [PASSED] drm_test_cmdline_force_D_only_not_digital [03:10:11] [PASSED] drm_test_cmdline_force_e_only [03:10:11] [PASSED] drm_test_cmdline_res [03:10:11] [PASSED] drm_test_cmdline_res_vesa [03:10:11] [PASSED] drm_test_cmdline_res_vesa_rblank [03:10:11] [PASSED] drm_test_cmdline_res_rblank [03:10:11] [PASSED] drm_test_cmdline_res_bpp [03:10:11] [PASSED] drm_test_cmdline_res_refresh [03:10:11] [PASSED] drm_test_cmdline_res_bpp_refresh [03:10:11] [PASSED] drm_test_cmdline_res_bpp_refresh_interlaced [03:10:11] [PASSED] drm_test_cmdline_res_bpp_refresh_margins [03:10:11] [PASSED] drm_test_cmdline_res_bpp_refresh_force_off [03:10:11] [PASSED] drm_test_cmdline_res_bpp_refresh_force_on [03:10:11] [PASSED] drm_test_cmdline_res_bpp_refresh_force_on_analog [03:10:11] [PASSED] drm_test_cmdline_res_bpp_refresh_force_on_digital [03:10:11] [PASSED] drm_test_cmdline_res_bpp_refresh_interlaced_margins_force_on [03:10:11] [PASSED] drm_test_cmdline_res_margins_force_on [03:10:11] [PASSED] drm_test_cmdline_res_vesa_margins [03:10:11] [PASSED] drm_test_cmdline_name [03:10:11] [PASSED] drm_test_cmdline_name_bpp [03:10:11] [PASSED] drm_test_cmdline_name_option [03:10:11] [PASSED] drm_test_cmdline_name_bpp_option [03:10:11] [PASSED] drm_test_cmdline_rotate_0 [03:10:11] [PASSED] drm_test_cmdline_rotate_90 [03:10:11] [PASSED] drm_test_cmdline_rotate_180 [03:10:11] [PASSED] drm_test_cmdline_rotate_270 [03:10:11] [PASSED] drm_test_cmdline_hmirror [03:10:11] [PASSED] drm_test_cmdline_vmirror [03:10:11] [PASSED] drm_test_cmdline_margin_options [03:10:11] [PASSED] drm_test_cmdline_multiple_options [03:10:11] [PASSED] drm_test_cmdline_bpp_extra_and_option [03:10:11] [PASSED] drm_test_cmdline_extra_and_option [03:10:11] [PASSED] drm_test_cmdline_freestanding_options [03:10:11] [PASSED] drm_test_cmdline_freestanding_force_e_and_options [03:10:11] [PASSED] drm_test_cmdline_panel_orientation [03:10:11] ================ drm_test_cmdline_invalid ================= [03:10:11] [PASSED] margin_only [03:10:11] [PASSED] interlace_only [03:10:11] [PASSED] res_missing_x [03:10:11] [PASSED] res_missing_y [03:10:11] [PASSED] res_bad_y [03:10:11] [PASSED] res_missing_y_bpp [03:10:11] [PASSED] res_bad_bpp [03:10:11] [PASSED] res_bad_refresh [03:10:11] [PASSED] res_bpp_refresh_force_on_off [03:10:11] [PASSED] res_invalid_mode [03:10:11] [PASSED] res_bpp_wrong_place_mode [03:10:11] [PASSED] name_bpp_refresh [03:10:11] [PASSED] name_refresh [03:10:11] [PASSED] name_refresh_wrong_mode [03:10:11] [PASSED] name_refresh_invalid_mode [03:10:11] [PASSED] rotate_multiple [03:10:11] [PASSED] rotate_invalid_val [03:10:11] [PASSED] rotate_truncated [03:10:11] [PASSED] invalid_option [03:10:11] [PASSED] invalid_tv_option [03:10:11] [PASSED] truncated_tv_option [03:10:11] ============ [PASSED] drm_test_cmdline_invalid ============= [03:10:11] =============== drm_test_cmdline_tv_options =============== [03:10:11] [PASSED] NTSC [03:10:11] [PASSED] NTSC_443 [03:10:11] [PASSED] NTSC_J [03:10:11] [PASSED] PAL [03:10:11] [PASSED] PAL_M [03:10:11] [PASSED] PAL_N [03:10:11] [PASSED] SECAM [03:10:11] [PASSED] MONO_525 [03:10:11] [PASSED] MONO_625 [03:10:11] =========== [PASSED] drm_test_cmdline_tv_options =========== [03:10:11] =============== [PASSED] drm_cmdline_parser ================ [03:10:11] ========== drmm_connector_hdmi_init (20 subtests) ========== [03:10:11] [PASSED] drm_test_connector_hdmi_init_valid [03:10:11] [PASSED] drm_test_connector_hdmi_init_bpc_8 [03:10:11] [PASSED] drm_test_connector_hdmi_init_bpc_10 [03:10:11] [PASSED] drm_test_connector_hdmi_init_bpc_12 [03:10:11] [PASSED] drm_test_connector_hdmi_init_bpc_invalid [03:10:11] [PASSED] drm_test_connector_hdmi_init_bpc_null [03:10:11] [PASSED] drm_test_connector_hdmi_init_formats_empty [03:10:11] [PASSED] drm_test_connector_hdmi_init_formats_no_rgb [03:10:11] === drm_test_connector_hdmi_init_formats_yuv420_allowed === [03:10:11] [PASSED] supported_formats=0x9 yuv420_allowed=1 [03:10:11] [PASSED] supported_formats=0x9 yuv420_allowed=0 [03:10:11] [PASSED] supported_formats=0x5 yuv420_allowed=1 [03:10:11] [PASSED] supported_formats=0x5 yuv420_allowed=0 [03:10:11] === [PASSED] drm_test_connector_hdmi_init_formats_yuv420_allowed === [03:10:11] [PASSED] drm_test_connector_hdmi_init_null_ddc [03:10:11] [PASSED] drm_test_connector_hdmi_init_null_product [03:10:11] [PASSED] drm_test_connector_hdmi_init_null_vendor [03:10:11] [PASSED] drm_test_connector_hdmi_init_product_length_exact [03:10:11] [PASSED] drm_test_connector_hdmi_init_product_length_too_long [03:10:11] [PASSED] drm_test_connector_hdmi_init_product_valid [03:10:11] [PASSED] drm_test_connector_hdmi_init_vendor_length_exact [03:10:11] [PASSED] drm_test_connector_hdmi_init_vendor_length_too_long [03:10:11] [PASSED] drm_test_connector_hdmi_init_vendor_valid [03:10:11] ========= drm_test_connector_hdmi_init_type_valid ========= [03:10:11] [PASSED] HDMI-A [03:10:11] [PASSED] HDMI-B [03:10:11] ===== [PASSED] drm_test_connector_hdmi_init_type_valid ===== [03:10:11] ======== drm_test_connector_hdmi_init_type_invalid ======== [03:10:11] [PASSED] Unknown [03:10:11] [PASSED] VGA [03:10:11] [PASSED] DVI-I [03:10:11] [PASSED] DVI-D [03:10:11] [PASSED] DVI-A [03:10:11] [PASSED] Composite [03:10:11] [PASSED] SVIDEO [03:10:11] [PASSED] LVDS [03:10:11] [PASSED] Component [03:10:11] [PASSED] DIN [03:10:11] [PASSED] DP [03:10:11] [PASSED] TV [03:10:11] [PASSED] eDP [03:10:11] [PASSED] Virtual [03:10:11] [PASSED] DSI [03:10:11] [PASSED] DPI [03:10:11] [PASSED] Writeback [03:10:11] [PASSED] SPI [03:10:11] [PASSED] USB [03:10:11] ==== [PASSED] drm_test_connector_hdmi_init_type_invalid ==== [03:10:11] ============ [PASSED] drmm_connector_hdmi_init ============= [03:10:11] ============= drmm_connector_init (3 subtests) ============= [03:10:11] [PASSED] drm_test_drmm_connector_init [03:10:11] [PASSED] drm_test_drmm_connector_init_null_ddc [03:10:11] ========= drm_test_drmm_connector_init_type_valid ========= [03:10:11] [PASSED] Unknown [03:10:11] [PASSED] VGA [03:10:11] [PASSED] DVI-I [03:10:11] [PASSED] DVI-D [03:10:11] [PASSED] DVI-A [03:10:11] [PASSED] Composite [03:10:11] [PASSED] SVIDEO [03:10:11] [PASSED] LVDS [03:10:11] [PASSED] Component [03:10:11] [PASSED] DIN [03:10:11] [PASSED] DP [03:10:11] [PASSED] HDMI-A [03:10:11] [PASSED] HDMI-B [03:10:11] [PASSED] TV [03:10:11] [PASSED] eDP [03:10:11] [PASSED] Virtual [03:10:11] [PASSED] DSI [03:10:11] [PASSED] DPI [03:10:11] [PASSED] Writeback [03:10:11] [PASSED] SPI [03:10:11] [PASSED] USB [03:10:11] ===== [PASSED] drm_test_drmm_connector_init_type_valid ===== [03:10:11] =============== [PASSED] drmm_connector_init =============== [03:10:11] ========= drm_connector_dynamic_init (6 subtests) ========== [03:10:11] [PASSED] drm_test_drm_connector_dynamic_init [03:10:11] [PASSED] drm_test_drm_connector_dynamic_init_null_ddc [03:10:11] [PASSED] drm_test_drm_connector_dynamic_init_not_added [03:10:11] [PASSED] drm_test_drm_connector_dynamic_init_properties [03:10:11] ===== drm_test_drm_connector_dynamic_init_type_valid ====== [03:10:11] [PASSED] Unknown [03:10:11] [PASSED] VGA [03:10:11] [PASSED] DVI-I [03:10:11] [PASSED] DVI-D [03:10:11] [PASSED] DVI-A [03:10:11] [PASSED] Composite [03:10:11] [PASSED] SVIDEO [03:10:11] [PASSED] LVDS [03:10:11] [PASSED] Component [03:10:11] [PASSED] DIN [03:10:11] [PASSED] DP [03:10:11] [PASSED] HDMI-A [03:10:11] [PASSED] HDMI-B [03:10:11] [PASSED] TV [03:10:11] [PASSED] eDP [03:10:11] [PASSED] Virtual [03:10:11] [PASSED] DSI [03:10:11] [PASSED] DPI [03:10:11] [PASSED] Writeback [03:10:11] [PASSED] SPI [03:10:11] [PASSED] USB [03:10:11] = [PASSED] drm_test_drm_connector_dynamic_init_type_valid == [03:10:11] ======== drm_test_drm_connector_dynamic_init_name ========= [03:10:11] [PASSED] Unknown [03:10:11] [PASSED] VGA [03:10:11] [PASSED] DVI-I [03:10:11] [PASSED] DVI-D [03:10:11] [PASSED] DVI-A [03:10:11] [PASSED] Composite [03:10:11] [PASSED] SVIDEO [03:10:11] [PASSED] LVDS [03:10:11] [PASSED] Component [03:10:11] [PASSED] DIN [03:10:11] [PASSED] DP [03:10:11] [PASSED] HDMI-A [03:10:11] [PASSED] HDMI-B [03:10:11] [PASSED] TV [03:10:11] [PASSED] eDP [03:10:11] [PASSED] Virtual [03:10:11] [PASSED] DSI [03:10:11] [PASSED] DPI [03:10:11] [PASSED] Writeback [03:10:11] [PASSED] SPI [03:10:11] [PASSED] USB [03:10:11] ==== [PASSED] drm_test_drm_connector_dynamic_init_name ===== [03:10:11] =========== [PASSED] drm_connector_dynamic_init ============ [03:10:11] ==== drm_connector_dynamic_register_early (4 subtests) ===== [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_early_on_list [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_early_defer [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_early_no_init [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_early_no_mode_object [03:10:11] ====== [PASSED] drm_connector_dynamic_register_early ======= [03:10:11] ======= drm_connector_dynamic_register (7 subtests) ======== [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_on_list [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_no_defer [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_no_init [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_mode_object [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_sysfs [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_sysfs_name [03:10:11] [PASSED] drm_test_drm_connector_dynamic_register_debugfs [03:10:11] ========= [PASSED] drm_connector_dynamic_register ========== [03:10:11] = drm_connector_attach_broadcast_rgb_property (2 subtests) = [03:10:11] [PASSED] drm_test_drm_connector_attach_broadcast_rgb_property [03:10:11] [PASSED] drm_test_drm_connector_attach_broadcast_rgb_property_hdmi_connector [03:10:11] === [PASSED] drm_connector_attach_broadcast_rgb_property === [03:10:11] ========== drm_get_tv_mode_from_name (2 subtests) ========== [03:10:11] ========== drm_test_get_tv_mode_from_name_valid =========== [03:10:11] [PASSED] NTSC [03:10:11] [PASSED] NTSC-443 [03:10:11] [PASSED] NTSC-J [03:10:11] [PASSED] PAL [03:10:11] [PASSED] PAL-M [03:10:11] [PASSED] PAL-N [03:10:11] [PASSED] SECAM [03:10:11] [PASSED] Mono [03:10:11] ====== [PASSED] drm_test_get_tv_mode_from_name_valid ======= [03:10:11] [PASSED] drm_test_get_tv_mode_from_name_truncated [03:10:11] ============ [PASSED] drm_get_tv_mode_from_name ============ [03:10:11] = drm_test_connector_hdmi_compute_mode_clock (12 subtests) = [03:10:11] [PASSED] drm_test_drm_hdmi_compute_mode_clock_rgb [03:10:11] [PASSED] drm_test_drm_hdmi_compute_mode_clock_rgb_10bpc [03:10:11] [PASSED] drm_test_drm_hdmi_compute_mode_clock_rgb_10bpc_vic_1 [03:10:11] [PASSED] drm_test_drm_hdmi_compute_mode_clock_rgb_12bpc [03:10:11] [PASSED] drm_test_drm_hdmi_compute_mode_clock_rgb_12bpc_vic_1 [03:10:11] [PASSED] drm_test_drm_hdmi_compute_mode_clock_rgb_double [03:10:11] = drm_test_connector_hdmi_compute_mode_clock_yuv420_valid = [03:10:11] [PASSED] VIC 96 [03:10:11] [PASSED] VIC 97 [03:10:11] [PASSED] VIC 101 [03:10:11] [PASSED] VIC 102 [03:10:11] [PASSED] VIC 106 [03:10:11] [PASSED] VIC 107 [03:10:11] === [PASSED] drm_test_connector_hdmi_compute_mode_clock_yuv420_valid === [03:10:11] [PASSED] drm_test_connector_hdmi_compute_mode_clock_yuv420_10_bpc [03:10:11] [PASSED] drm_test_connector_hdmi_compute_mode_clock_yuv420_12_bpc [03:10:11] [PASSED] drm_test_connector_hdmi_compute_mode_clock_yuv422_8_bpc [03:10:11] [PASSED] drm_test_connector_hdmi_compute_mode_clock_yuv422_10_bpc [03:10:11] [PASSED] drm_test_connector_hdmi_compute_mode_clock_yuv422_12_bpc [03:10:11] === [PASSED] drm_test_connector_hdmi_compute_mode_clock ==== [03:10:11] == drm_hdmi_connector_get_broadcast_rgb_name (2 subtests) == [03:10:11] === drm_test_drm_hdmi_connector_get_broadcast_rgb_name ==== [03:10:11] [PASSED] Automatic [03:10:11] [PASSED] Full [03:10:11] [PASSED] Limited 16:235 [03:10:11] === [PASSED] drm_test_drm_hdmi_connector_get_broadcast_rgb_name === [03:10:11] [PASSED] drm_test_drm_hdmi_connector_get_broadcast_rgb_name_invalid [03:10:11] ==== [PASSED] drm_hdmi_connector_get_broadcast_rgb_name ==== [03:10:11] == drm_hdmi_connector_get_output_format_name (2 subtests) == [03:10:11] === drm_test_drm_hdmi_connector_get_output_format_name ==== [03:10:11] [PASSED] RGB [03:10:11] [PASSED] YUV 4:2:0 [03:10:11] [PASSED] YUV 4:2:2 [03:10:11] [PASSED] YUV 4:4:4 [03:10:11] === [PASSED] drm_test_drm_hdmi_connector_get_output_format_name === [03:10:11] [PASSED] drm_test_drm_hdmi_connector_get_output_format_name_invalid [03:10:11] ==== [PASSED] drm_hdmi_connector_get_output_format_name ==== [03:10:11] ============= drm_damage_helper (21 subtests) ============== [03:10:11] [PASSED] drm_test_damage_iter_no_damage [03:10:11] [PASSED] drm_test_damage_iter_no_damage_fractional_src [03:10:11] [PASSED] drm_test_damage_iter_no_damage_src_moved [03:10:11] [PASSED] drm_test_damage_iter_no_damage_fractional_src_moved [03:10:11] [PASSED] drm_test_damage_iter_no_damage_not_visible [03:10:11] [PASSED] drm_test_damage_iter_no_damage_no_crtc [03:10:11] [PASSED] drm_test_damage_iter_no_damage_no_fb [03:10:11] [PASSED] drm_test_damage_iter_simple_damage [03:10:11] [PASSED] drm_test_damage_iter_single_damage [03:10:11] [PASSED] drm_test_damage_iter_single_damage_intersect_src [03:10:11] [PASSED] drm_test_damage_iter_single_damage_outside_src [03:10:11] [PASSED] drm_test_damage_iter_single_damage_fractional_src [03:10:11] [PASSED] drm_test_damage_iter_single_damage_intersect_fractional_src [03:10:11] [PASSED] drm_test_damage_iter_single_damage_outside_fractional_src [03:10:11] [PASSED] drm_test_damage_iter_single_damage_src_moved [03:10:11] [PASSED] drm_test_damage_iter_single_damage_fractional_src_moved [03:10:11] [PASSED] drm_test_damage_iter_damage [03:10:11] [PASSED] drm_test_damage_iter_damage_one_intersect [03:10:11] [PASSED] drm_test_damage_iter_damage_one_outside [03:10:11] [PASSED] drm_test_damage_iter_damage_src_moved [03:10:11] [PASSED] drm_test_damage_iter_damage_not_visible [03:10:11] ================ [PASSED] drm_damage_helper ================ [03:10:11] ============== drm_dp_mst_helper (3 subtests) ============== [03:10:11] ============== drm_test_dp_mst_calc_pbn_mode ============== [03:10:11] [PASSED] Clock 154000 BPP 30 DSC disabled [03:10:11] [PASSED] Clock 234000 BPP 30 DSC disabled [03:10:11] [PASSED] Clock 297000 BPP 24 DSC disabled [03:10:11] [PASSED] Clock 332880 BPP 24 DSC enabled [03:10:11] [PASSED] Clock 324540 BPP 24 DSC enabled [03:10:11] ========== [PASSED] drm_test_dp_mst_calc_pbn_mode ========== [03:10:11] ============== drm_test_dp_mst_calc_pbn_div =============== [03:10:11] [PASSED] Link rate 2000000 lane count 4 [03:10:11] [PASSED] Link rate 2000000 lane count 2 [03:10:11] [PASSED] Link rate 2000000 lane count 1 [03:10:11] [PASSED] Link rate 1350000 lane count 4 [03:10:11] [PASSED] Link rate 1350000 lane count 2 [03:10:11] [PASSED] Link rate 1350000 lane count 1 [03:10:11] [PASSED] Link rate 1000000 lane count 4 [03:10:11] [PASSED] Link rate 1000000 lane count 2 [03:10:11] [PASSED] Link rate 1000000 lane count 1 [03:10:11] [PASSED] Link rate 810000 lane count 4 [03:10:11] [PASSED] Link rate 810000 lane count 2 [03:10:11] [PASSED] Link rate 810000 lane count 1 [03:10:11] [PASSED] Link rate 540000 lane count 4 [03:10:11] [PASSED] Link rate 540000 lane count 2 [03:10:11] [PASSED] Link rate 540000 lane count 1 [03:10:11] [PASSED] Link rate 270000 lane count 4 [03:10:11] [PASSED] Link rate 270000 lane count 2 [03:10:11] [PASSED] Link rate 270000 lane count 1 [03:10:11] [PASSED] Link rate 162000 lane count 4 [03:10:11] [PASSED] Link rate 162000 lane count 2 [03:10:11] [PASSED] Link rate 162000 lane count 1 [03:10:11] ========== [PASSED] drm_test_dp_mst_calc_pbn_div =========== [03:10:11] ========= drm_test_dp_mst_sideband_msg_req_decode ========= [03:10:11] [PASSED] DP_ENUM_PATH_RESOURCES with port number [03:10:11] [PASSED] DP_POWER_UP_PHY with port number [03:10:11] [PASSED] DP_POWER_DOWN_PHY with port number [03:10:11] [PASSED] DP_ALLOCATE_PAYLOAD with SDP stream sinks [03:10:11] [PASSED] DP_ALLOCATE_PAYLOAD with port number [03:10:11] [PASSED] DP_ALLOCATE_PAYLOAD with VCPI [03:10:11] [PASSED] DP_ALLOCATE_PAYLOAD with PBN [03:10:11] [PASSED] DP_QUERY_PAYLOAD with port number [03:10:11] [PASSED] DP_QUERY_PAYLOAD with VCPI [03:10:11] [PASSED] DP_REMOTE_DPCD_READ with port number [03:10:11] [PASSED] DP_REMOTE_DPCD_READ with DPCD address [03:10:11] [PASSED] DP_REMOTE_DPCD_READ with max number of bytes [03:10:11] [PASSED] DP_REMOTE_DPCD_WRITE with port number [03:10:11] [PASSED] DP_REMOTE_DPCD_WRITE with DPCD address [03:10:11] [PASSED] DP_REMOTE_DPCD_WRITE with data array [03:10:11] [PASSED] DP_REMOTE_I2C_READ with port number [03:10:11] [PASSED] DP_REMOTE_I2C_READ with I2C device ID [03:10:11] [PASSED] DP_REMOTE_I2C_READ with transactions array [03:10:11] [PASSED] DP_REMOTE_I2C_WRITE with port number [03:10:11] [PASSED] DP_REMOTE_I2C_WRITE with I2C device ID [03:10:11] [PASSED] DP_REMOTE_I2C_WRITE with data array [03:10:11] [PASSED] DP_QUERY_STREAM_ENC_STATUS with stream ID [03:10:11] [PASSED] DP_QUERY_STREAM_ENC_STATUS with client ID [03:10:11] [PASSED] DP_QUERY_STREAM_ENC_STATUS with stream event [03:10:11] [PASSED] DP_QUERY_STREAM_ENC_STATUS with valid stream event [03:10:11] [PASSED] DP_QUERY_STREAM_ENC_STATUS with stream behavior [03:10:11] [PASSED] DP_QUERY_STREAM_ENC_STATUS with a valid stream behavior [03:10:11] ===== [PASSED] drm_test_dp_mst_sideband_msg_req_decode ===== [03:10:11] ================ [PASSED] drm_dp_mst_helper ================ [03:10:11] ================== drm_exec (7 subtests) =================== [03:10:11] [PASSED] sanitycheck [03:10:11] [PASSED] test_lock [03:10:11] [PASSED] test_lock_unlock [03:10:11] [PASSED] test_duplicates [03:10:11] [PASSED] test_prepare [03:10:11] [PASSED] test_prepare_array [03:10:11] [PASSED] test_multiple_loops [03:10:11] ==================== [PASSED] drm_exec ===================== [03:10:11] =========== drm_format_helper_test (17 subtests) =========== [03:10:11] ============== drm_test_fb_xrgb8888_to_gray8 ============== [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ========== [PASSED] drm_test_fb_xrgb8888_to_gray8 ========== [03:10:11] ============= drm_test_fb_xrgb8888_to_rgb332 ============== [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ========= [PASSED] drm_test_fb_xrgb8888_to_rgb332 ========== [03:10:11] ============= drm_test_fb_xrgb8888_to_rgb565 ============== [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ========= [PASSED] drm_test_fb_xrgb8888_to_rgb565 ========== [03:10:11] ============ drm_test_fb_xrgb8888_to_xrgb1555 ============= [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ======== [PASSED] drm_test_fb_xrgb8888_to_xrgb1555 ========= [03:10:11] ============ drm_test_fb_xrgb8888_to_argb1555 ============= [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ======== [PASSED] drm_test_fb_xrgb8888_to_argb1555 ========= [03:10:11] ============ drm_test_fb_xrgb8888_to_rgba5551 ============= [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ======== [PASSED] drm_test_fb_xrgb8888_to_rgba5551 ========= [03:10:11] ============= drm_test_fb_xrgb8888_to_rgb888 ============== [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ========= [PASSED] drm_test_fb_xrgb8888_to_rgb888 ========== [03:10:11] ============= drm_test_fb_xrgb8888_to_bgr888 ============== [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ========= [PASSED] drm_test_fb_xrgb8888_to_bgr888 ========== [03:10:11] ============ drm_test_fb_xrgb8888_to_argb8888 ============= [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ======== [PASSED] drm_test_fb_xrgb8888_to_argb8888 ========= [03:10:11] =========== drm_test_fb_xrgb8888_to_xrgb2101010 =========== [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ======= [PASSED] drm_test_fb_xrgb8888_to_xrgb2101010 ======= [03:10:11] =========== drm_test_fb_xrgb8888_to_argb2101010 =========== [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ======= [PASSED] drm_test_fb_xrgb8888_to_argb2101010 ======= [03:10:11] ============== drm_test_fb_xrgb8888_to_mono =============== [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ========== [PASSED] drm_test_fb_xrgb8888_to_mono =========== [03:10:11] ==================== drm_test_fb_swab ===================== [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ================ [PASSED] drm_test_fb_swab ================= [03:10:11] ============ drm_test_fb_xrgb8888_to_xbgr8888 ============= [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ======== [PASSED] drm_test_fb_xrgb8888_to_xbgr8888 ========= [03:10:11] ============ drm_test_fb_xrgb8888_to_abgr8888 ============= [03:10:11] [PASSED] single_pixel_source_buffer [03:10:11] [PASSED] single_pixel_clip_rectangle [03:10:11] [PASSED] well_known_colors [03:10:11] [PASSED] destination_pitch [03:10:11] ======== [PASSED] drm_test_fb_xrgb8888_to_abgr8888 ========= [03:10:11] ================= drm_test_fb_clip_offset ================= [03:10:11] [PASSED] pass through [03:10:11] [PASSED] horizontal offset [03:10:11] [PASSED] vertical offset [03:10:11] [PASSED] horizontal and vertical offset [03:10:11] [PASSED] horizontal offset (custom pitch) [03:10:11] [PASSED] vertical offset (custom pitch) [03:10:11] [PASSED] horizontal and vertical offset (custom pitch) [03:10:11] ============= [PASSED] drm_test_fb_clip_offset ============= [03:10:11] =================== drm_test_fb_memcpy ==================== [03:10:11] [PASSED] single_pixel_source_buffer: XR24 little-endian (0x34325258) [03:10:11] [PASSED] single_pixel_source_buffer: XRA8 little-endian (0x38415258) [03:10:11] [PASSED] single_pixel_source_buffer: YU24 little-endian (0x34325559) [03:10:11] [PASSED] single_pixel_clip_rectangle: XB24 little-endian (0x34324258) [03:10:11] [PASSED] single_pixel_clip_rectangle: XRA8 little-endian (0x38415258) [03:10:11] [PASSED] single_pixel_clip_rectangle: YU24 little-endian (0x34325559) [03:10:11] [PASSED] well_known_colors: XB24 little-endian (0x34324258) [03:10:11] [PASSED] well_known_colors: XRA8 little-endian (0x38415258) [03:10:11] [PASSED] well_known_colors: YU24 little-endian (0x34325559) [03:10:11] [PASSED] destination_pitch: XB24 little-endian (0x34324258) [03:10:11] [PASSED] destination_pitch: XRA8 little-endian (0x38415258) [03:10:11] [PASSED] destination_pitch: YU24 little-endian (0x34325559) [03:10:11] =============== [PASSED] drm_test_fb_memcpy ================ [03:10:11] ============= [PASSED] drm_format_helper_test ============== [03:10:11] ================= drm_format (18 subtests) ================= [03:10:11] [PASSED] drm_test_format_block_width_invalid [03:10:11] [PASSED] drm_test_format_block_width_one_plane [03:10:11] [PASSED] drm_test_format_block_width_two_plane [03:10:11] [PASSED] drm_test_format_block_width_three_plane [03:10:11] [PASSED] drm_test_format_block_width_tiled [03:10:11] [PASSED] drm_test_format_block_height_invalid [03:10:11] [PASSED] drm_test_format_block_height_one_plane [03:10:11] [PASSED] drm_test_format_block_height_two_plane [03:10:11] [PASSED] drm_test_format_block_height_three_plane [03:10:11] [PASSED] drm_test_format_block_height_tiled [03:10:11] [PASSED] drm_test_format_min_pitch_invalid [03:10:11] [PASSED] drm_test_format_min_pitch_one_plane_8bpp [03:10:11] [PASSED] drm_test_format_min_pitch_one_plane_16bpp [03:10:11] [PASSED] drm_test_format_min_pitch_one_plane_24bpp [03:10:11] [PASSED] drm_test_format_min_pitch_one_plane_32bpp [03:10:11] [PASSED] drm_test_format_min_pitch_two_plane [03:10:11] [PASSED] drm_test_format_min_pitch_three_plane_8bpp [03:10:11] [PASSED] drm_test_format_min_pitch_tiled [03:10:11] =================== [PASSED] drm_format ==================== [03:10:11] ============== drm_framebuffer (10 subtests) =============== [03:10:11] ========== drm_test_framebuffer_check_src_coords ========== [03:10:11] [PASSED] Success: source fits into fb [03:10:11] [PASSED] Fail: overflowing fb with x-axis coordinate [03:10:11] [PASSED] Fail: overflowing fb with y-axis coordinate [03:10:11] [PASSED] Fail: overflowing fb with source width [03:10:11] [PASSED] Fail: overflowing fb with source height [03:10:11] ====== [PASSED] drm_test_framebuffer_check_src_coords ====== [03:10:11] [PASSED] drm_test_framebuffer_cleanup [03:10:11] =============== drm_test_framebuffer_create =============== [03:10:11] [PASSED] ABGR8888 normal sizes [03:10:11] [PASSED] ABGR8888 max sizes [03:10:11] [PASSED] ABGR8888 pitch greater than min required [03:10:11] [PASSED] ABGR8888 pitch less than min required [03:10:11] [PASSED] ABGR8888 Invalid width [03:10:11] [PASSED] ABGR8888 Invalid buffer handle [03:10:11] [PASSED] No pixel format [03:10:11] [PASSED] ABGR8888 Width 0 [03:10:11] [PASSED] ABGR8888 Height 0 [03:10:11] [PASSED] ABGR8888 Out of bound height * pitch combination [03:10:11] [PASSED] ABGR8888 Large buffer offset [03:10:11] [PASSED] ABGR8888 Buffer offset for inexistent plane [03:10:11] [PASSED] ABGR8888 Invalid flag [03:10:11] [PASSED] ABGR8888 Set DRM_MODE_FB_MODIFIERS without modifiers [03:10:11] [PASSED] ABGR8888 Valid buffer modifier [03:10:11] [PASSED] ABGR8888 Invalid buffer modifier(DRM_FORMAT_MOD_SAMSUNG_64_32_TILE) [03:10:11] [PASSED] ABGR8888 Extra pitches without DRM_MODE_FB_MODIFIERS [03:10:11] [PASSED] ABGR8888 Extra pitches with DRM_MODE_FB_MODIFIERS [03:10:11] [PASSED] NV12 Normal sizes [03:10:11] [PASSED] NV12 Max sizes [03:10:11] [PASSED] NV12 Invalid pitch [03:10:11] [PASSED] NV12 Invalid modifier/missing DRM_MODE_FB_MODIFIERS flag [03:10:11] [PASSED] NV12 different modifier per-plane [03:10:11] [PASSED] NV12 with DRM_FORMAT_MOD_SAMSUNG_64_32_TILE [03:10:11] [PASSED] NV12 Valid modifiers without DRM_MODE_FB_MODIFIERS [03:10:11] [PASSED] NV12 Modifier for inexistent plane [03:10:11] [PASSED] NV12 Handle for inexistent plane [03:10:11] [PASSED] NV12 Handle for inexistent plane without DRM_MODE_FB_MODIFIERS [03:10:11] [PASSED] YVU420 DRM_MODE_FB_MODIFIERS set without modifier [03:10:11] [PASSED] YVU420 Normal sizes [03:10:11] [PASSED] YVU420 Max sizes [03:10:11] [PASSED] YVU420 Invalid pitch [03:10:11] [PASSED] YVU420 Different pitches [03:10:11] [PASSED] YVU420 Different buffer offsets/pitches [03:10:11] [PASSED] YVU420 Modifier set just for plane 0, without DRM_MODE_FB_MODIFIERS [03:10:11] [PASSED] YVU420 Modifier set just for planes 0, 1, without DRM_MODE_FB_MODIFIERS [03:10:11] [PASSED] YVU420 Modifier set just for plane 0, 1, with DRM_MODE_FB_MODIFIERS [03:10:11] [PASSED] YVU420 Valid modifier [03:10:11] [PASSED] YVU420 Different modifiers per plane [03:10:11] [PASSED] YVU420 Modifier for inexistent plane [03:10:11] [PASSED] YUV420_10BIT Invalid modifier(DRM_FORMAT_MOD_LINEAR) [03:10:11] [PASSED] X0L2 Normal sizes [03:10:11] [PASSED] X0L2 Max sizes [03:10:11] [PASSED] X0L2 Invalid pitch [03:10:11] [PASSED] X0L2 Pitch greater than minimum required [03:10:11] [PASSED] X0L2 Handle for inexistent plane [03:10:11] [PASSED] X0L2 Offset for inexistent plane, without DRM_MODE_FB_MODIFIERS set [03:10:11] [PASSED] X0L2 Modifier without DRM_MODE_FB_MODIFIERS set [03:10:11] [PASSED] X0L2 Valid modifier [03:10:11] [PASSED] X0L2 Modifier for inexistent plane [03:10:11] =========== [PASSED] drm_test_framebuffer_create =========== [03:10:11] [PASSED] drm_test_framebuffer_free [03:10:11] [PASSED] drm_test_framebuffer_init [03:10:11] [PASSED] drm_test_framebuffer_init_bad_format [03:10:11] [PASSED] drm_test_framebuffer_init_dev_mismatch [03:10:11] [PASSED] drm_test_framebuffer_lookup [03:10:11] [PASSED] drm_test_framebuffer_lookup_inexistent [03:10:11] [PASSED] drm_test_framebuffer_modifiers_not_supported [03:10:11] ================= [PASSED] drm_framebuffer ================= [03:10:11] ================ drm_gem_shmem (8 subtests) ================ [03:10:11] [PASSED] drm_gem_shmem_test_obj_create [03:10:11] [PASSED] drm_gem_shmem_test_obj_create_private [03:10:11] [PASSED] drm_gem_shmem_test_pin_pages [03:10:11] [PASSED] drm_gem_shmem_test_vmap [03:10:11] [PASSED] drm_gem_shmem_test_get_sg_table [03:10:11] [PASSED] drm_gem_shmem_test_get_pages_sgt [03:10:11] [PASSED] drm_gem_shmem_test_madvise [03:10:11] [PASSED] drm_gem_shmem_test_purge [03:10:11] ================== [PASSED] drm_gem_shmem ================== [03:10:11] === drm_atomic_helper_connector_hdmi_check (27 subtests) === [03:10:11] [PASSED] drm_test_check_broadcast_rgb_auto_cea_mode [03:10:11] [PASSED] drm_test_check_broadcast_rgb_auto_cea_mode_vic_1 [03:10:11] [PASSED] drm_test_check_broadcast_rgb_full_cea_mode [03:10:11] [PASSED] drm_test_check_broadcast_rgb_full_cea_mode_vic_1 [03:10:11] [PASSED] drm_test_check_broadcast_rgb_limited_cea_mode [03:10:11] [PASSED] drm_test_check_broadcast_rgb_limited_cea_mode_vic_1 [03:10:11] ====== drm_test_check_broadcast_rgb_cea_mode_yuv420 ======= [03:10:11] [PASSED] Automatic [03:10:11] [PASSED] Full [03:10:11] [PASSED] Limited 16:235 [03:10:11] == [PASSED] drm_test_check_broadcast_rgb_cea_mode_yuv420 === [03:10:11] [PASSED] drm_test_check_broadcast_rgb_crtc_mode_changed [03:10:11] [PASSED] drm_test_check_broadcast_rgb_crtc_mode_not_changed [03:10:11] [PASSED] drm_test_check_disable_connector [03:10:11] [PASSED] drm_test_check_hdmi_funcs_reject_rate [03:10:11] [PASSED] drm_test_check_max_tmds_rate_bpc_fallback_rgb [03:10:11] [PASSED] drm_test_check_max_tmds_rate_bpc_fallback_yuv420 [03:10:11] [PASSED] drm_test_check_max_tmds_rate_bpc_fallback_ignore_yuv422 [03:10:11] [PASSED] drm_test_check_max_tmds_rate_bpc_fallback_ignore_yuv420 [03:10:11] [PASSED] drm_test_check_driver_unsupported_fallback_yuv420 [03:10:11] [PASSED] drm_test_check_output_bpc_crtc_mode_changed [03:10:11] [PASSED] drm_test_check_output_bpc_crtc_mode_not_changed [03:10:11] [PASSED] drm_test_check_output_bpc_dvi [03:10:11] [PASSED] drm_test_check_output_bpc_format_vic_1 [03:10:11] [PASSED] drm_test_check_output_bpc_format_display_8bpc_only [03:10:11] [PASSED] drm_test_check_output_bpc_format_display_rgb_only [03:10:11] [PASSED] drm_test_check_output_bpc_format_driver_8bpc_only [03:10:11] [PASSED] drm_test_check_output_bpc_format_driver_rgb_only [03:10:11] [PASSED] drm_test_check_tmds_char_rate_rgb_8bpc [03:10:11] [PASSED] drm_test_check_tmds_char_rate_rgb_10bpc [03:10:11] [PASSED] drm_test_check_tmds_char_rate_rgb_12bpc [03:10:11] ===== [PASSED] drm_atomic_helper_connector_hdmi_check ====== [03:10:11] === drm_atomic_helper_connector_hdmi_reset (6 subtests) ==== [03:10:11] [PASSED] drm_test_check_broadcast_rgb_value [03:10:11] [PASSED] drm_test_check_bpc_8_value [03:10:11] [PASSED] drm_test_check_bpc_10_value [03:10:11] [PASSED] drm_test_check_bpc_12_value [03:10:11] [PASSED] drm_test_check_format_value [03:10:11] [PASSED] drm_test_check_tmds_char_value [03:10:11] ===== [PASSED] drm_atomic_helper_connector_hdmi_reset ====== [03:10:11] = drm_atomic_helper_connector_hdmi_mode_valid (4 subtests) = [03:10:11] [PASSED] drm_test_check_mode_valid [03:10:11] [PASSED] drm_test_check_mode_valid_reject [03:10:11] [PASSED] drm_test_check_mode_valid_reject_rate [03:10:11] [PASSED] drm_test_check_mode_valid_reject_max_clock [03:10:11] === [PASSED] drm_atomic_helper_connector_hdmi_mode_valid === [03:10:11] = drm_atomic_helper_connector_hdmi_infoframes (5 subtests) = [03:10:11] [PASSED] drm_test_check_infoframes [03:10:11] [PASSED] drm_test_check_reject_avi_infoframe [03:10:11] [PASSED] drm_test_check_reject_hdr_infoframe_bpc_8 [03:10:11] [PASSED] drm_test_check_reject_hdr_infoframe_bpc_10 [03:10:11] [PASSED] drm_test_check_reject_audio_infoframe [03:10:11] === [PASSED] drm_atomic_helper_connector_hdmi_infoframes === [03:10:11] ================= drm_managed (2 subtests) ================= [03:10:11] [PASSED] drm_test_managed_release_action [03:10:11] [PASSED] drm_test_managed_run_action [03:10:11] =================== [PASSED] drm_managed =================== [03:10:11] =================== drm_mm (6 subtests) ==================== [03:10:11] [PASSED] drm_test_mm_init [03:10:11] [PASSED] drm_test_mm_debug [03:10:11] [PASSED] drm_test_mm_align32 [03:10:11] [PASSED] drm_test_mm_align64 [03:10:11] [PASSED] drm_test_mm_lowest [03:10:11] [PASSED] drm_test_mm_highest [03:10:11] ===================== [PASSED] drm_mm ====================== [03:10:11] ============= drm_modes_analog_tv (5 subtests) ============= [03:10:11] [PASSED] drm_test_modes_analog_tv_mono_576i [03:10:11] [PASSED] drm_test_modes_analog_tv_ntsc_480i [03:10:11] [PASSED] drm_test_modes_analog_tv_ntsc_480i_inlined [03:10:11] [PASSED] drm_test_modes_analog_tv_pal_576i [03:10:11] [PASSED] drm_test_modes_analog_tv_pal_576i_inlined [03:10:11] =============== [PASSED] drm_modes_analog_tv =============== [03:10:11] ============== drm_plane_helper (2 subtests) =============== [03:10:11] =============== drm_test_check_plane_state ================ [03:10:11] [PASSED] clipping_simple [03:10:11] [PASSED] clipping_rotate_reflect [03:10:11] [PASSED] positioning_simple [03:10:11] [PASSED] upscaling [03:10:11] [PASSED] downscaling [03:10:11] [PASSED] rounding1 [03:10:11] [PASSED] rounding2 [03:10:11] [PASSED] rounding3 [03:10:11] [PASSED] rounding4 [03:10:11] =========== [PASSED] drm_test_check_plane_state ============ [03:10:11] =========== drm_test_check_invalid_plane_state ============ [03:10:11] [PASSED] positioning_invalid [03:10:11] [PASSED] upscaling_invalid [03:10:11] [PASSED] downscaling_invalid [03:10:11] ======= [PASSED] drm_test_check_invalid_plane_state ======== [03:10:11] ================ [PASSED] drm_plane_helper ================= [03:10:11] ====== drm_connector_helper_tv_get_modes (1 subtest) ======= [03:10:11] ====== drm_test_connector_helper_tv_get_modes_check ======= [03:10:11] [PASSED] None [03:10:11] [PASSED] PAL [03:10:11] [PASSED] NTSC [03:10:11] [PASSED] Both, NTSC Default [03:10:11] [PASSED] Both, PAL Default [03:10:11] [PASSED] Both, NTSC Default, with PAL on command-line [03:10:11] [PASSED] Both, PAL Default, with NTSC on command-line [03:10:11] == [PASSED] drm_test_connector_helper_tv_get_modes_check === [03:10:11] ======== [PASSED] drm_connector_helper_tv_get_modes ======== [03:10:11] ================== drm_rect (9 subtests) =================== [03:10:11] [PASSED] drm_test_rect_clip_scaled_div_by_zero [03:10:11] [PASSED] drm_test_rect_clip_scaled_not_clipped [03:10:11] [PASSED] drm_test_rect_clip_scaled_clipped [03:10:11] [PASSED] drm_test_rect_clip_scaled_signed_vs_unsigned [03:10:11] ================= drm_test_rect_intersect ================= [03:10:11] [PASSED] top-left x bottom-right: 2x2+1+1 x 2x2+0+0 [03:10:11] [PASSED] top-right x bottom-left: 2x2+0+0 x 2x2+1-1 [03:10:11] [PASSED] bottom-left x top-right: 2x2+1-1 x 2x2+0+0 [03:10:11] [PASSED] bottom-right x top-left: 2x2+0+0 x 2x2+1+1 [03:10:11] [PASSED] right x left: 2x1+0+0 x 3x1+1+0 [03:10:11] [PASSED] left x right: 3x1+1+0 x 2x1+0+0 [03:10:11] [PASSED] up x bottom: 1x2+0+0 x 1x3+0-1 [03:10:11] [PASSED] bottom x up: 1x3+0-1 x 1x2+0+0 [03:10:11] [PASSED] touching corner: 1x1+0+0 x 2x2+1+1 [03:10:11] [PASSED] touching side: 1x1+0+0 x 1x1+1+0 [03:10:11] [PASSED] equal rects: 2x2+0+0 x 2x2+0+0 [03:10:11] [PASSED] inside another: 2x2+0+0 x 1x1+1+1 [03:10:11] [PASSED] far away: 1x1+0+0 x 1x1+3+6 [03:10:11] [PASSED] points intersecting: 0x0+5+10 x 0x0+5+10 [03:10:11] [PASSED] points not intersecting: 0x0+0+0 x 0x0+5+10 [03:10:11] ============= [PASSED] drm_test_rect_intersect ============= [03:10:11] ================ drm_test_rect_calc_hscale ================ [03:10:11] [PASSED] normal use [03:10:11] [PASSED] out of max range [03:10:11] [PASSED] out of min range [03:10:11] [PASSED] zero dst [03:10:11] [PASSED] negative src [03:10:11] [PASSED] negative dst [03:10:11] ============ [PASSED] drm_test_rect_calc_hscale ============ [03:10:11] ================ drm_test_rect_calc_vscale ================ [03:10:11] [PASSED] normal use [03:10:11] [PASSED] out of max range [03:10:11] [PASSED] out of min range [03:10:11] [PASSED] zero dst [03:10:11] [PASSED] negative src [03:10:11] [PASSED] negative dst stty: 'standard input': Inappropriate ioctl for device [03:10:11] ============ [PASSED] drm_test_rect_calc_vscale ============ [03:10:11] ================== drm_test_rect_rotate =================== [03:10:11] [PASSED] reflect-x [03:10:11] [PASSED] reflect-y [03:10:11] [PASSED] rotate-0 [03:10:11] [PASSED] rotate-90 [03:10:11] [PASSED] rotate-180 [03:10:11] [PASSED] rotate-270 [03:10:11] ============== [PASSED] drm_test_rect_rotate =============== [03:10:11] ================ drm_test_rect_rotate_inv ================= [03:10:11] [PASSED] reflect-x [03:10:11] [PASSED] reflect-y [03:10:11] [PASSED] rotate-0 [03:10:11] [PASSED] rotate-90 [03:10:11] [PASSED] rotate-180 [03:10:11] [PASSED] rotate-270 [03:10:11] ============ [PASSED] drm_test_rect_rotate_inv ============= [03:10:11] ==================== [PASSED] drm_rect ===================== [03:10:11] ============ drm_sysfb_modeset_test (1 subtest) ============ [03:10:11] ============ drm_test_sysfb_build_fourcc_list ============= [03:10:11] [PASSED] no native formats [03:10:11] [PASSED] XRGB8888 as native format [03:10:11] [PASSED] remove duplicates [03:10:11] [PASSED] convert alpha formats [03:10:11] [PASSED] random formats [03:10:11] ======== [PASSED] drm_test_sysfb_build_fourcc_list ========= [03:10:11] ============= [PASSED] drm_sysfb_modeset_test ============== [03:10:11] ================== drm_fixp (2 subtests) =================== [03:10:11] [PASSED] drm_test_int2fixp [03:10:11] [PASSED] drm_test_sm2fixp [03:10:11] ==================== [PASSED] drm_fixp ===================== [03:10:11] ============================================================ [03:10:11] Testing complete. Ran 621 tests: passed: 621 [03:10:11] Elapsed time: 26.546s total, 1.654s configuring, 24.671s building, 0.180s running + /kernel/tools/testing/kunit/kunit.py run --kunitconfig /kernel/drivers/gpu/drm/ttm/tests/.kunitconfig [03:10:11] Configuring KUnit Kernel ... Regenerating .config ... Populating config with: $ make ARCH=um O=.kunit olddefconfig [03:10:13] Building KUnit Kernel ... Populating config with: $ make ARCH=um O=.kunit olddefconfig Building with: $ make all compile_commands.json scripts_gdb ARCH=um O=.kunit --jobs=48 [03:10:22] Starting KUnit Kernel (1/1)... [03:10:22] ============================================================ Running tests with: $ .kunit/linux kunit.enable=1 mem=1G console=tty kunit_shutdown=halt [03:10:23] ================= ttm_device (5 subtests) ================== [03:10:23] [PASSED] ttm_device_init_basic [03:10:23] [PASSED] ttm_device_init_multiple [03:10:23] [PASSED] ttm_device_fini_basic [03:10:23] [PASSED] ttm_device_init_no_vma_man [03:10:23] ================== ttm_device_init_pools ================== [03:10:23] [PASSED] No DMA allocations, no DMA32 required [03:10:23] [PASSED] DMA allocations, DMA32 required [03:10:23] [PASSED] No DMA allocations, DMA32 required [03:10:23] [PASSED] DMA allocations, no DMA32 required [03:10:23] ============== [PASSED] ttm_device_init_pools ============== [03:10:23] =================== [PASSED] ttm_device ==================== [03:10:23] ================== ttm_pool (8 subtests) =================== [03:10:23] ================== ttm_pool_alloc_basic =================== [03:10:23] [PASSED] One page [03:10:23] [PASSED] More than one page [03:10:23] [PASSED] Above the allocation limit [03:10:23] [PASSED] One page, with coherent DMA mappings enabled [03:10:23] [PASSED] Above the allocation limit, with coherent DMA mappings enabled [03:10:23] ============== [PASSED] ttm_pool_alloc_basic =============== [03:10:23] ============== ttm_pool_alloc_basic_dma_addr ============== [03:10:23] [PASSED] One page [03:10:23] [PASSED] More than one page [03:10:23] [PASSED] Above the allocation limit [03:10:23] [PASSED] One page, with coherent DMA mappings enabled [03:10:23] [PASSED] Above the allocation limit, with coherent DMA mappings enabled [03:10:23] ========== [PASSED] ttm_pool_alloc_basic_dma_addr ========== [03:10:23] [PASSED] ttm_pool_alloc_order_caching_match [03:10:23] [PASSED] ttm_pool_alloc_caching_mismatch [03:10:23] [PASSED] ttm_pool_alloc_order_mismatch [03:10:23] [PASSED] ttm_pool_free_dma_alloc [03:10:23] [PASSED] ttm_pool_free_no_dma_alloc [03:10:23] [PASSED] ttm_pool_fini_basic [03:10:23] ==================== [PASSED] ttm_pool ===================== [03:10:23] ================ ttm_resource (8 subtests) ================= [03:10:23] ================= ttm_resource_init_basic ================= [03:10:23] [PASSED] Init resource in TTM_PL_SYSTEM [03:10:23] [PASSED] Init resource in TTM_PL_VRAM [03:10:23] [PASSED] Init resource in a private placement [03:10:23] [PASSED] Init resource in TTM_PL_SYSTEM, set placement flags [03:10:23] ============= [PASSED] ttm_resource_init_basic ============= [03:10:23] [PASSED] ttm_resource_init_pinned [03:10:23] [PASSED] ttm_resource_fini_basic [03:10:23] [PASSED] ttm_resource_manager_init_basic [03:10:23] [PASSED] ttm_resource_manager_usage_basic [03:10:23] [PASSED] ttm_resource_manager_set_used_basic [03:10:23] [PASSED] ttm_sys_man_alloc_basic [03:10:23] [PASSED] ttm_sys_man_free_basic [03:10:23] ================== [PASSED] ttm_resource =================== [03:10:23] =================== ttm_tt (15 subtests) =================== [03:10:23] ==================== ttm_tt_init_basic ==================== [03:10:23] [PASSED] Page-aligned size [03:10:23] [PASSED] Extra pages requested [03:10:23] ================ [PASSED] ttm_tt_init_basic ================ [03:10:23] [PASSED] ttm_tt_init_misaligned [03:10:23] [PASSED] ttm_tt_fini_basic [03:10:23] [PASSED] ttm_tt_fini_sg [03:10:23] [PASSED] ttm_tt_fini_shmem [03:10:23] [PASSED] ttm_tt_create_basic [03:10:23] [PASSED] ttm_tt_create_invalid_bo_type [03:10:23] [PASSED] ttm_tt_create_ttm_exists [03:10:23] [PASSED] ttm_tt_create_failed [03:10:23] [PASSED] ttm_tt_destroy_basic [03:10:23] [PASSED] ttm_tt_populate_null_ttm [03:10:23] [PASSED] ttm_tt_populate_populated_ttm [03:10:23] [PASSED] ttm_tt_unpopulate_basic [03:10:23] [PASSED] ttm_tt_unpopulate_empty_ttm [03:10:23] [PASSED] ttm_tt_swapin_basic [03:10:23] ===================== [PASSED] ttm_tt ====================== [03:10:23] =================== ttm_bo (14 subtests) =================== [03:10:23] =========== ttm_bo_reserve_optimistic_no_ticket =========== [03:10:23] [PASSED] Cannot be interrupted and sleeps [03:10:23] [PASSED] Cannot be interrupted, locks straight away [03:10:23] [PASSED] Can be interrupted, sleeps [03:10:23] ======= [PASSED] ttm_bo_reserve_optimistic_no_ticket ======= [03:10:23] [PASSED] ttm_bo_reserve_locked_no_sleep [03:10:23] [PASSED] ttm_bo_reserve_no_wait_ticket [03:10:23] [PASSED] ttm_bo_reserve_double_resv [03:10:23] [PASSED] ttm_bo_reserve_interrupted [03:10:23] [PASSED] ttm_bo_reserve_deadlock [03:10:23] [PASSED] ttm_bo_unreserve_basic [03:10:23] [PASSED] ttm_bo_unreserve_pinned [03:10:23] [PASSED] ttm_bo_unreserve_bulk [03:10:23] [PASSED] ttm_bo_fini_basic [03:10:23] [PASSED] ttm_bo_fini_shared_resv [03:10:23] [PASSED] ttm_bo_pin_basic [03:10:23] [PASSED] ttm_bo_pin_unpin_resource [03:10:23] [PASSED] ttm_bo_multiple_pin_one_unpin [03:10:23] ===================== [PASSED] ttm_bo ====================== [03:10:23] ============== ttm_bo_validate (22 subtests) =============== [03:10:23] ============== ttm_bo_init_reserved_sys_man =============== [03:10:23] [PASSED] Buffer object for userspace [03:10:23] [PASSED] Kernel buffer object [03:10:23] [PASSED] Shared buffer object [03:10:23] ========== [PASSED] ttm_bo_init_reserved_sys_man =========== [03:10:23] ============== ttm_bo_init_reserved_mock_man ============== [03:10:23] [PASSED] Buffer object for userspace [03:10:23] [PASSED] Kernel buffer object [03:10:23] [PASSED] Shared buffer object [03:10:23] ========== [PASSED] ttm_bo_init_reserved_mock_man ========== [03:10:23] [PASSED] ttm_bo_init_reserved_resv [03:10:23] ================== ttm_bo_validate_basic ================== [03:10:23] [PASSED] Buffer object for userspace [03:10:23] [PASSED] Kernel buffer object [03:10:23] [PASSED] Shared buffer object [03:10:23] ============== [PASSED] ttm_bo_validate_basic ============== [03:10:23] [PASSED] ttm_bo_validate_invalid_placement [03:10:23] ============= ttm_bo_validate_same_placement ============== [03:10:23] [PASSED] System manager [03:10:23] [PASSED] VRAM manager [03:10:23] ========= [PASSED] ttm_bo_validate_same_placement ========== [03:10:23] [PASSED] ttm_bo_validate_failed_alloc [03:10:23] [PASSED] ttm_bo_validate_pinned [03:10:23] [PASSED] ttm_bo_validate_busy_placement [03:10:23] ================ ttm_bo_validate_multihop ================= [03:10:23] [PASSED] Buffer object for userspace [03:10:23] [PASSED] Kernel buffer object [03:10:23] [PASSED] Shared buffer object [03:10:23] ============ [PASSED] ttm_bo_validate_multihop ============= [03:10:23] ========== ttm_bo_validate_no_placement_signaled ========== [03:10:23] [PASSED] Buffer object in system domain, no page vector [03:10:23] [PASSED] Buffer object in system domain with an existing page vector [03:10:23] ====== [PASSED] ttm_bo_validate_no_placement_signaled ====== [03:10:23] ======== ttm_bo_validate_no_placement_not_signaled ======== [03:10:23] [PASSED] Buffer object for userspace [03:10:23] [PASSED] Kernel buffer object [03:10:23] [PASSED] Shared buffer object [03:10:23] ==== [PASSED] ttm_bo_validate_no_placement_not_signaled ==== [03:10:23] [PASSED] ttm_bo_validate_move_fence_signaled [03:10:23] ========= ttm_bo_validate_move_fence_not_signaled ========= [03:10:23] [PASSED] Waits for GPU [03:10:23] [PASSED] Tries to lock straight away [03:10:23] ===== [PASSED] ttm_bo_validate_move_fence_not_signaled ===== [03:10:23] [PASSED] ttm_bo_validate_swapout [03:10:23] [PASSED] ttm_bo_validate_happy_evict [03:10:23] [PASSED] ttm_bo_validate_all_pinned_evict [03:10:23] [PASSED] ttm_bo_validate_allowed_only_evict [03:10:23] [PASSED] ttm_bo_validate_deleted_evict [03:10:23] [PASSED] ttm_bo_validate_busy_domain_evict [03:10:23] [PASSED] ttm_bo_validate_evict_gutting [03:10:23] [PASSED] ttm_bo_validate_recrusive_evict stty: 'standard input': Inappropriate ioctl for device [03:10:23] ================= [PASSED] ttm_bo_validate ================= [03:10:23] ============================================================ [03:10:23] Testing complete. Ran 102 tests: passed: 102 [03:10:23] Elapsed time: 11.264s total, 1.687s configuring, 9.361s building, 0.179s running + cleanup ++ stat -c %u:%g /kernel + chown -R 1003:1003 /kernel ^ permalink raw reply [flat|nested] 30+ messages in thread
* ✓ Xe.CI.BAT: success for drm/xe/oa: Wa_14026633728 2026-04-07 3:02 [PATCH 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit ` (3 preceding siblings ...) 2026-04-07 3:10 ` ✓ CI.KUnit: success for drm/xe/oa: Wa_14026633728 Patchwork @ 2026-04-07 3:51 ` Patchwork 2026-04-07 5:02 ` ✓ Xe.CI.FULL: " Patchwork 5 siblings, 0 replies; 30+ messages in thread From: Patchwork @ 2026-04-07 3:51 UTC (permalink / raw) To: Ashutosh Dixit; +Cc: intel-xe [-- Attachment #1: Type: text/plain, Size: 848 bytes --] == Series Details == Series: drm/xe/oa: Wa_14026633728 URL : https://patchwork.freedesktop.org/series/164412/ State : success == Summary == CI Bug Log - changes from xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688_BAT -> xe-pw-164412v1_BAT ==================================================== Summary ------- **SUCCESS** No regressions found. Participating hosts (13 -> 13) ------------------------------ No changes in participating hosts Changes ------- No changes found Build changes ------------- * Linux: xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688 -> xe-pw-164412v1 IGT_8849: 8849 xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688: 7b7217a9e27a82ef10be22ab3a55ad5bbc849688 xe-pw-164412v1: 164412v1 == Logs == For more details see: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/index.html [-- Attachment #2: Type: text/html, Size: 1396 bytes --] ^ permalink raw reply [flat|nested] 30+ messages in thread
* ✓ Xe.CI.FULL: success for drm/xe/oa: Wa_14026633728 2026-04-07 3:02 [PATCH 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit ` (4 preceding siblings ...) 2026-04-07 3:51 ` ✓ Xe.CI.BAT: " Patchwork @ 2026-04-07 5:02 ` Patchwork 5 siblings, 0 replies; 30+ messages in thread From: Patchwork @ 2026-04-07 5:02 UTC (permalink / raw) To: Ashutosh Dixit; +Cc: intel-xe [-- Attachment #1: Type: text/plain, Size: 16251 bytes --] == Series Details == Series: drm/xe/oa: Wa_14026633728 URL : https://patchwork.freedesktop.org/series/164412/ State : success == Summary == CI Bug Log - changes from xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688_FULL -> xe-pw-164412v1_FULL ==================================================== Summary ------- **SUCCESS** No regressions found. Participating hosts (2 -> 2) ------------------------------ No changes in participating hosts Known issues ------------ Here are the changes found in xe-pw-164412v1_FULL that come from known issues: ### IGT changes ### #### Issues hit #### * igt@core_hotunplug@unplug-rescan: - shard-bmg: [PASS][1] -> [ABORT][2] ([Intel XE#7578]) [1]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-3/igt@core_hotunplug@unplug-rescan.html [2]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-10/igt@core_hotunplug@unplug-rescan.html * igt@kms_big_fb@linear-max-hw-stride-32bpp-rotate-0-hflip: - shard-bmg: NOTRUN -> [SKIP][3] ([Intel XE#7059] / [Intel XE#7085]) [3]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_big_fb@linear-max-hw-stride-32bpp-rotate-0-hflip.html * igt@kms_big_fb@y-tiled-max-hw-stride-32bpp-rotate-180-hflip: - shard-bmg: NOTRUN -> [SKIP][4] ([Intel XE#1124]) [4]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_big_fb@y-tiled-max-hw-stride-32bpp-rotate-180-hflip.html * igt@kms_ccs@bad-aux-stride-y-tiled-ccs: - shard-bmg: NOTRUN -> [SKIP][5] ([Intel XE#2887]) [5]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_ccs@bad-aux-stride-y-tiled-ccs.html * igt@kms_ccs@crc-primary-suspend-y-tiled-gen12-rc-ccs-cc: - shard-bmg: NOTRUN -> [SKIP][6] ([Intel XE#3432]) [6]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@kms_ccs@crc-primary-suspend-y-tiled-gen12-rc-ccs-cc.html * igt@kms_content_protection@dp-mst-type-0: - shard-bmg: NOTRUN -> [SKIP][7] ([Intel XE#2390] / [Intel XE#6974]) [7]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_content_protection@dp-mst-type-0.html * igt@kms_cursor_crc@cursor-sliding-512x512: - shard-bmg: NOTRUN -> [SKIP][8] ([Intel XE#2321] / [Intel XE#7355]) [8]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@kms_cursor_crc@cursor-sliding-512x512.html * igt@kms_cursor_legacy@flip-vs-cursor-atomic: - shard-bmg: [PASS][9] -> [FAIL][10] ([Intel XE#7571]) [9]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-10/igt@kms_cursor_legacy@flip-vs-cursor-atomic.html [10]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_cursor_legacy@flip-vs-cursor-atomic.html * igt@kms_dirtyfb@drrs-dirtyfb-ioctl: - shard-bmg: NOTRUN -> [SKIP][11] ([Intel XE#1508]) [11]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_dirtyfb@drrs-dirtyfb-ioctl.html * igt@kms_dsc@dsc-with-output-formats: - shard-bmg: NOTRUN -> [SKIP][12] ([Intel XE#2244]) [12]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_dsc@dsc-with-output-formats.html * igt@kms_flip@flip-vs-suspend: - shard-bmg: [PASS][13] -> [INCOMPLETE][14] ([Intel XE#2049] / [Intel XE#2597]) +1 other test incomplete [13]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-5/igt@kms_flip@flip-vs-suspend.html [14]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-2/igt@kms_flip@flip-vs-suspend.html * igt@kms_flip_scaled_crc@flip-64bpp-yftile-to-32bpp-yftile-downscaling: - shard-bmg: NOTRUN -> [SKIP][15] ([Intel XE#7178] / [Intel XE#7351]) [15]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_flip_scaled_crc@flip-64bpp-yftile-to-32bpp-yftile-downscaling.html * igt@kms_flip_scaled_crc@flip-nv12-linear-to-nv12-linear-reflect-x: - shard-bmg: NOTRUN -> [SKIP][16] ([Intel XE#7179]) [16]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@kms_flip_scaled_crc@flip-nv12-linear-to-nv12-linear-reflect-x.html * igt@kms_frontbuffer_tracking@drrs-2p-primscrn-cur-indfb-draw-render: - shard-bmg: NOTRUN -> [SKIP][17] ([Intel XE#2311]) +2 other tests skip [17]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@kms_frontbuffer_tracking@drrs-2p-primscrn-cur-indfb-draw-render.html * igt@kms_frontbuffer_tracking@fbc-2p-scndscrn-cur-indfb-onoff: - shard-bmg: NOTRUN -> [SKIP][18] ([Intel XE#4141]) +2 other tests skip [18]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_frontbuffer_tracking@fbc-2p-scndscrn-cur-indfb-onoff.html * igt@kms_frontbuffer_tracking@psr-2p-scndscrn-cur-indfb-onoff: - shard-bmg: NOTRUN -> [SKIP][19] ([Intel XE#2313]) +1 other test skip [19]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_frontbuffer_tracking@psr-2p-scndscrn-cur-indfb-onoff.html * igt@kms_joiner@invalid-modeset-ultra-joiner: - shard-bmg: NOTRUN -> [SKIP][20] ([Intel XE#6911] / [Intel XE#7378]) [20]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@kms_joiner@invalid-modeset-ultra-joiner.html * igt@kms_plane@pixel-format-y-tiled-ccs-modifier: - shard-bmg: NOTRUN -> [SKIP][21] ([Intel XE#7283]) [21]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@kms_plane@pixel-format-y-tiled-ccs-modifier.html * igt@kms_pm_backlight@fade-with-dpms: - shard-bmg: NOTRUN -> [SKIP][22] ([Intel XE#7376] / [Intel XE#870]) [22]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@kms_pm_backlight@fade-with-dpms.html * igt@kms_psr2_sf@fbc-psr2-overlay-plane-move-continuous-exceed-sf: - shard-bmg: NOTRUN -> [SKIP][23] ([Intel XE#1489]) [23]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_psr2_sf@fbc-psr2-overlay-plane-move-continuous-exceed-sf.html * igt@kms_psr@fbc-psr2-basic: - shard-bmg: NOTRUN -> [SKIP][24] ([Intel XE#2234] / [Intel XE#2850]) +3 other tests skip [24]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@kms_psr@fbc-psr2-basic.html * igt@kms_sharpness_filter@invalid-filter-with-plane: - shard-bmg: NOTRUN -> [SKIP][25] ([Intel XE#6503]) [25]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_sharpness_filter@invalid-filter-with-plane.html * igt@xe_eudebug@vma-ufence-faultable: - shard-bmg: NOTRUN -> [SKIP][26] ([Intel XE#7636]) +1 other test skip [26]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@xe_eudebug@vma-ufence-faultable.html * igt@xe_eudebug_sriov@deny-sriov: - shard-bmg: NOTRUN -> [SKIP][27] ([Intel XE#5793] / [Intel XE#7320] / [Intel XE#7464]) [27]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@xe_eudebug_sriov@deny-sriov.html * igt@xe_evict@evict-mixed-many-threads-small: - shard-bmg: [PASS][28] -> [INCOMPLETE][29] ([Intel XE#6321]) [28]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-5/igt@xe_evict@evict-mixed-many-threads-small.html [29]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-2/igt@xe_evict@evict-mixed-many-threads-small.html * igt@xe_exec_basic@multigpu-many-execqueues-many-vm-bindexecqueue-userptr-invalidate: - shard-bmg: NOTRUN -> [SKIP][30] ([Intel XE#2322] / [Intel XE#7372]) [30]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@xe_exec_basic@multigpu-many-execqueues-many-vm-bindexecqueue-userptr-invalidate.html * igt@xe_exec_fault_mode@many-multi-queue-userptr-invalidate-race-prefetch: - shard-bmg: NOTRUN -> [SKIP][31] ([Intel XE#7136]) +1 other test skip [31]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@xe_exec_fault_mode@many-multi-queue-userptr-invalidate-race-prefetch.html * igt@xe_exec_multi_queue@one-queue-preempt-mode-fault-userptr: - shard-bmg: NOTRUN -> [SKIP][32] ([Intel XE#6874]) +2 other tests skip [32]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@xe_exec_multi_queue@one-queue-preempt-mode-fault-userptr.html * igt@xe_exec_system_allocator@many-execqueues-mmap-prefetch: - shard-bmg: [PASS][33] -> [ABORT][34] ([Intel XE#6652]) [33]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-6/igt@xe_exec_system_allocator@many-execqueues-mmap-prefetch.html [34]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-2/igt@xe_exec_system_allocator@many-execqueues-mmap-prefetch.html * igt@xe_exec_threads@threads-multi-queue-mixed-fd-userptr: - shard-bmg: NOTRUN -> [SKIP][35] ([Intel XE#7138]) +2 other tests skip [35]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@xe_exec_threads@threads-multi-queue-mixed-fd-userptr.html * igt@xe_module_load@force-load: - shard-bmg: ([PASS][36], [PASS][37]) -> [ABORT][38] ([Intel XE#7578]) [36]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-6/igt@xe_module_load@force-load.html [37]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-2/igt@xe_module_load@force-load.html [38]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-9/igt@xe_module_load@force-load.html #### Possible fixes #### * igt@kms_ccs@crc-primary-suspend-4-tiled-bmg-ccs: - shard-bmg: [INCOMPLETE][39] ([Intel XE#7084]) -> [PASS][40] +1 other test pass [39]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-10/igt@kms_ccs@crc-primary-suspend-4-tiled-bmg-ccs.html [40]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-5/igt@kms_ccs@crc-primary-suspend-4-tiled-bmg-ccs.html * igt@kms_flip@2x-dpms-vs-vblank-race: - shard-bmg: [INCOMPLETE][41] -> [PASS][42] +1 other test pass [41]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-2/igt@kms_flip@2x-dpms-vs-vblank-race.html [42]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-1/igt@kms_flip@2x-dpms-vs-vblank-race.html * igt@kms_vrr@seamless-rr-switch-virtual@pipe-a-edp-1: - shard-lnl: [FAIL][43] ([Intel XE#2142]) -> [PASS][44] +1 other test pass [43]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-lnl-6/igt@kms_vrr@seamless-rr-switch-virtual@pipe-a-edp-1.html [44]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-lnl-8/igt@kms_vrr@seamless-rr-switch-virtual@pipe-a-edp-1.html * igt@xe_exec_system_allocator@pat-index-madvise-pat-idx-uc-single-vma: - shard-lnl: [FAIL][45] ([Intel XE#5625]) -> [PASS][46] [45]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-lnl-2/igt@xe_exec_system_allocator@pat-index-madvise-pat-idx-uc-single-vma.html [46]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-lnl-6/igt@xe_exec_system_allocator@pat-index-madvise-pat-idx-uc-single-vma.html #### Warnings #### * igt@kms_tiled_display@basic-test-pattern: - shard-bmg: [SKIP][47] ([Intel XE#2426] / [Intel XE#5848]) -> [FAIL][48] ([Intel XE#1729] / [Intel XE#7424]) [47]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688/shard-bmg-7/igt@kms_tiled_display@basic-test-pattern.html [48]: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/shard-bmg-6/igt@kms_tiled_display@basic-test-pattern.html [Intel XE#1124]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/1124 [Intel XE#1489]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/1489 [Intel XE#1508]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/1508 [Intel XE#1729]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/1729 [Intel XE#2049]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2049 [Intel XE#2142]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2142 [Intel XE#2234]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2234 [Intel XE#2244]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2244 [Intel XE#2311]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2311 [Intel XE#2313]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2313 [Intel XE#2321]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2321 [Intel XE#2322]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2322 [Intel XE#2390]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2390 [Intel XE#2426]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2426 [Intel XE#2597]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2597 [Intel XE#2850]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2850 [Intel XE#2887]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2887 [Intel XE#3432]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/3432 [Intel XE#4141]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/4141 [Intel XE#5625]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/5625 [Intel XE#5793]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/5793 [Intel XE#5848]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/5848 [Intel XE#6321]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/6321 [Intel XE#6503]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/6503 [Intel XE#6652]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/6652 [Intel XE#6874]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/6874 [Intel XE#6911]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/6911 [Intel XE#6974]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/6974 [Intel XE#7059]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7059 [Intel XE#7084]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7084 [Intel XE#7085]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7085 [Intel XE#7136]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7136 [Intel XE#7138]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7138 [Intel XE#7178]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7178 [Intel XE#7179]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7179 [Intel XE#7283]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7283 [Intel XE#7320]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7320 [Intel XE#7351]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7351 [Intel XE#7355]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7355 [Intel XE#7372]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7372 [Intel XE#7376]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7376 [Intel XE#7378]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7378 [Intel XE#7424]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7424 [Intel XE#7464]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7464 [Intel XE#7571]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7571 [Intel XE#7578]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7578 [Intel XE#7636]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/7636 [Intel XE#870]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/870 Build changes ------------- * Linux: xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688 -> xe-pw-164412v1 IGT_8849: 8849 xe-4853-7b7217a9e27a82ef10be22ab3a55ad5bbc849688: 7b7217a9e27a82ef10be22ab3a55ad5bbc849688 xe-pw-164412v1: 164412v1 == Logs == For more details see: https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-164412v1/index.html [-- Attachment #2: Type: text/html, Size: 17587 bytes --] ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v2 0/3] drm/xe/oa: Wa_14026633728 @ 2026-04-09 23:17 Ashutosh Dixit 2026-04-09 23:17 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-09 23:17 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa v2: Change to Patch 1 Ashutosh Dixit (3): drm/xe/oa: Use xe_map layer drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap drm/xe/oa: Implement Wa_14026633728 drivers/gpu/drm/xe/xe_oa.c | 127 ++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 3 - drivers/gpu/drm/xe/xe_wa_oob.rules | 1 + 3 files changed, 72 insertions(+), 59 deletions(-) -- 2.48.1 ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-09 23:17 [PATCH v2 0/3] " Ashutosh Dixit @ 2026-04-09 23:17 ` Ashutosh Dixit 2026-04-11 0:49 ` Dixit, Ashutosh 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-09 23:17 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa OA code should have used xe_map layer to begin with. In CRI, the OA buffer can be located both in system and device memory. For these reasons, move OA code to use the xe_map layer when accessing the OA buffer. v2: Use xe_map layer also in xe_oa_copy_to_user (Umesh) Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> --- drivers/gpu/drm/xe/xe_oa.c | 98 +++++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 3 - 2 files changed, 59 insertions(+), 42 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index 6337e671c97ae..ff2ddec8ffa65 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) #define oa_report_header_64bit(__s) \ ((__s)->oa_buffer.format->header == HDR_64_BIT) -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) +static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) { - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + return oa_report_header_64bit(stream) ? + xe_map_rd(stream->oa->xe, map, tail, u64) : + xe_map_rd(stream->oa->xe, map, tail, u32); } -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)report = 0; - else - *report = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head, u64, 0) : + xe_map_wr(stream->oa->xe, map, head, u32, 0); } -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + return oa_report_header_64bit(stream) ? - *((u64 *)report + 1) : - *((u32 *)report + 1); + xe_map_rd(stream->oa->xe, map, tail + 2, u64) : + xe_map_rd(stream->oa->xe, map, tail + 1, u32); } -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)&report[2] = 0; - else - report[1] = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : + xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); } static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) * they were written. If not : (╯°□°）╯︵ ┻━┻ */ while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { - void *report = stream->oa_buffer.vaddr + tail; - - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) break; tail = xe_oa_circ_diff(stream, tail, report_size); @@ -311,30 +317,47 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) return HRTIMER_RESTART; } +/* Because there is no iosys_map_copy_to_user() or copy_to_user_from_iomem() */ +static int xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) +{ + u32 __user *dptr = dst; + + xe_gt_assert(stream->gt, len % 4 == 0); + + for (int i = 0; i < len / 4; i++) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + u32 val = xe_map_rd(stream->oa->xe, map, head, u32); + int ret = put_user(val, dptr++); + + if (ret) + return ret; + head += 4; + } + + return 0; +} + static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, - size_t count, size_t *offset, const u8 *report) + size_t count, size_t *offset, u32 head) { int report_size = stream->oa_buffer.format->size; int report_size_partial; - u8 *oa_buf_end; if ((count - *offset) < report_size) return -ENOSPC; buf += *offset; - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - report_size_partial = oa_buf_end - report; + report_size_partial = stream->oa_buffer.circ_size - head; if (report_size_partial < report_size) { - if (copy_to_user(buf, report, report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, head, report_size_partial)) return -EFAULT; buf += report_size_partial; - if (copy_to_user(buf, stream->oa_buffer.vaddr, - report_size - report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) return -EFAULT; - } else if (copy_to_user(buf, report, report_size)) { + } else if (xe_oa_copy_to_user(stream, buf, head, report_size)) { return -EFAULT; } @@ -347,7 +370,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, size_t count, size_t *offset) { int report_size = stream->oa_buffer.format->size; - u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); size_t start_offset = *offset; unsigned long flags; @@ -364,26 +386,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, for (; xe_oa_circ_diff(stream, tail, head); head = xe_oa_circ_incr(stream, head, report_size)) { - u8 *report = oa_buf_base + head; - - ret = xe_oa_append_report(stream, buf, count, offset, report); + ret = xe_oa_append_report(stream, buf, count, offset, head); if (ret) break; if (!(stream->oa_buffer.circ_size % report_size)) { /* Clear out report id and timestamp to detect unlanded reports */ - oa_report_id_clear(stream, (void *)report); - oa_timestamp_clear(stream, (void *)report); + oa_report_id_clear(stream, head); + oa_timestamp_clear(stream, head); } else { - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - u32 part = oa_buf_end - report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + u32 part = stream->oa_buffer.circ_size - head; /* Zero out the entire report */ if (report_size <= part) { - memset(report, 0, report_size); + xe_map_memset(stream->oa->xe, map, head, 0, report_size); } else { - memset(report, 0, part); - memset(oa_buf_base, 0, report_size - part); + xe_map_memset(stream->oa->xe, map, head, 0, part); + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); } } } @@ -436,7 +456,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, + xe_bo_size(stream->oa_buffer.bo)); } static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) @@ -891,7 +912,6 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) stream->oa_buffer.bo = bo; /* mmap implementation requires OA buffer to be in system memory */ xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); - stream->oa_buffer.vaddr = bo->vmap.vaddr; return 0; } diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h index b03ffd5134834..2dd550c93b8ac 100644 --- a/drivers/gpu/drm/xe/xe_oa_types.h +++ b/drivers/gpu/drm/xe/xe_oa_types.h @@ -162,9 +162,6 @@ struct xe_oa_buffer { /** @format: xe_bo backing the OA buffer */ struct xe_bo *bo; - /** @vaddr: mapped vaddr of the OA buffer */ - u8 *vaddr; - /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ spinlock_t ptr_lock; -- 2.48.1 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-09 23:17 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit @ 2026-04-11 0:49 ` Dixit, Ashutosh 2026-04-13 18:05 ` Umesh Nerlige Ramappa 0 siblings, 1 reply; 30+ messages in thread From: Dixit, Ashutosh @ 2026-04-11 0:49 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa On Thu, 09 Apr 2026 16:17:52 -0700, Ashutosh Dixit wrote: > Hi Umesh, > +/* Because there is no iosys_map_copy_to_user() or copy_to_user_from_iomem() */ > +static int xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) > +{ > + u32 __user *dptr = dst; > + > + xe_gt_assert(stream->gt, len % 4 == 0); > + > + for (int i = 0; i < len / 4; i++) { > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > + u32 val = xe_map_rd(stream->oa->xe, map, head, u32); > + int ret = put_user(val, dptr++); > + > + if (ret) > + return ret; > + head += 4; > + } > + > + return 0; > +} Unfortunately, with this copy_to_user() implementation we are seeing buffer overflow, implying there is significant performance impact, even when OA buffer is in system memory: Starting subtest: non-zero-reason Starting dynamic subtest: oag-0 (xe_oa:6798) CRITICAL: Test assertion failure function test_non_zero_reason, file ../tests/intel/xe_oa.c:2693: (xe_oa:6798) CRITICAL: Failed assertion: !(oa_status & DRM_XE_OASTATUS_BUFFER_OVERFLOW) That is, we are unable to read data at the rate at which HW is writing data, causing buffer overrun. Therefore in v3 I have basically reverted to the copy_to_user() implementation of v1, except that I have taken the copy_to_user() out of xe_oa.c and moved it to xe_map.h, to make that implementation "official" (and OA code to strictly use the xe_map layer). This seems to be the best approach to me, it avoids unnecessary copies and works correctly on all of our systems of interest. Also, the previous code, without this patch, is of course doing this exact same thing, except that the use of the xe_map layer make these assumptions explicit. If you still don't think we should be doing this, I can try to get a second opinion about this patch. Thanks. ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-11 0:49 ` Dixit, Ashutosh @ 2026-04-13 18:05 ` Umesh Nerlige Ramappa 2026-04-13 22:06 ` Dixit, Ashutosh 0 siblings, 1 reply; 30+ messages in thread From: Umesh Nerlige Ramappa @ 2026-04-13 18:05 UTC (permalink / raw) To: Dixit, Ashutosh; +Cc: intel-xe On Fri, Apr 10, 2026 at 05:49:03PM -0700, Dixit, Ashutosh wrote: >On Thu, 09 Apr 2026 16:17:52 -0700, Ashutosh Dixit wrote: >> > >Hi Umesh, > >> +/* Because there is no iosys_map_copy_to_user() or copy_to_user_from_iomem() */ >> +static int xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) >> +{ >> + u32 __user *dptr = dst; >> + >> + xe_gt_assert(stream->gt, len % 4 == 0); >> + >> + for (int i = 0; i < len / 4; i++) { >> + struct iosys_map *map = &stream->oa_buffer.bo->vmap; >> + u32 val = xe_map_rd(stream->oa->xe, map, head, u32); >> + int ret = put_user(val, dptr++); >> + >> + if (ret) >> + return ret; >> + head += 4; >> + } >> + >> + return 0; >> +} > >Unfortunately, with this copy_to_user() implementation we are seeing buffer >overflow, implying there is significant performance impact, even when OA >buffer is in system memory: > > Starting subtest: non-zero-reason > Starting dynamic subtest: oag-0 > (xe_oa:6798) CRITICAL: Test assertion failure function test_non_zero_reason, file ../tests/intel/xe_oa.c:2693: > (xe_oa:6798) CRITICAL: Failed assertion: !(oa_status & DRM_XE_OASTATUS_BUFFER_OVERFLOW) > >That is, we are unable to read data at the rate at which HW is writing >data, causing buffer overrun. > >Therefore in v3 I have basically reverted to the copy_to_user() >implementation of v1, except that I have taken the copy_to_user() out of >xe_oa.c and moved it to xe_map.h, to make that implementation "official" >(and OA code to strictly use the xe_map layer). This seems to be the best >approach to me, it avoids unnecessary copies and works correctly on all of >our systems of interest. > >Also, the previous code, without this patch, is of course doing this exact >same thing, except that the use of the xe_map layer make these assumptions >explicit. > >If you still don't think we should be doing this, I can try to get a second >opinion about this patch. Thanks. Yes, please include someone who can give us an Ack or some inputs on this approach because we are hitting some issues. Thanks, Umesh ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-13 18:05 ` Umesh Nerlige Ramappa @ 2026-04-13 22:06 ` Dixit, Ashutosh 0 siblings, 0 replies; 30+ messages in thread From: Dixit, Ashutosh @ 2026-04-13 22:06 UTC (permalink / raw) To: Umesh Nerlige Ramappa; +Cc: intel-xe On Mon, 13 Apr 2026 11:05:09 -0700, Umesh Nerlige Ramappa wrote: > > On Fri, Apr 10, 2026 at 05:49:03PM -0700, Dixit, Ashutosh wrote: > > On Thu, 09 Apr 2026 16:17:52 -0700, Ashutosh Dixit wrote: > >> > > > > Hi Umesh, > > > >> +/* Because there is no iosys_map_copy_to_user() or copy_to_user_from_iomem() */ > >> +static int xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) > >> +{ > >> + u32 __user *dptr = dst; > >> + > >> + xe_gt_assert(stream->gt, len % 4 == 0); > >> + > >> + for (int i = 0; i < len / 4; i++) { > >> + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > >> + u32 val = xe_map_rd(stream->oa->xe, map, head, u32); > >> + int ret = put_user(val, dptr++); > >> + > >> + if (ret) > >> + return ret; > >> + head += 4; > >> + } > >> + > >> + return 0; > >> +} > > > > Unfortunately, with this copy_to_user() implementation we are seeing buffer > > overflow, implying there is significant performance impact, even when OA > > buffer is in system memory: > > > > Starting subtest: non-zero-reason > > Starting dynamic subtest: oag-0 > > (xe_oa:6798) CRITICAL: Test assertion failure function test_non_zero_reason, file ../tests/intel/xe_oa.c:2693: > > (xe_oa:6798) CRITICAL: Failed assertion: !(oa_status & DRM_XE_OASTATUS_BUFFER_OVERFLOW) > > > > That is, we are unable to read data at the rate at which HW is writing > > data, causing buffer overrun. > > > > Therefore in v3 I have basically reverted to the copy_to_user() > > implementation of v1, except that I have taken the copy_to_user() out of > > xe_oa.c and moved it to xe_map.h, to make that implementation "official" > > (and OA code to strictly use the xe_map layer). This seems to be the best > > approach to me, it avoids unnecessary copies and works correctly on all of > > our systems of interest. > > > > Also, the previous code, without this patch, is of course doing this exact > > same thing, except that the use of the xe_map layer make these assumptions > > explicit. > > > > If you still don't think we should be doing this, I can try to get a second > > opinion about this patch. Thanks. > > Yes, please include someone who can give us an Ack or some inputs on this > approach because we are hitting some issues. I have asked. What I posted in v3 is my preferred option since it avoids the extra copy. But if that doesn't go through, the function below has the extra copy, but seems to have the same performace as v3, so that is also an option: static int xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) { struct iosys_map *map = &stream->oa_buffer.bo->vmap; u8 buf[1024]; // Allocate as part of stream xe_map_memcpy_from(xe, buf, map, head, len); return copy_to_user(dst, buf, len); } So I am planning to try this next, if v3 doesn't go through. Thanks. ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v3 0/3] drm/xe/oa: Wa_14026633728 @ 2026-04-11 0:48 Ashutosh Dixit 2026-04-11 0:48 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-11 0:48 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa v3: Change to Patch 1 again (see changelog in Patch 1) v2: Change to Patch 1 Ashutosh Dixit (3): drm/xe/oa: Use xe_map layer drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap drm/xe/oa: Implement Wa_14026633728 drivers/gpu/drm/xe/xe_map.h | 14 ++++ drivers/gpu/drm/xe/xe_oa.c | 109 ++++++++++++++--------------- drivers/gpu/drm/xe/xe_oa_types.h | 3 - drivers/gpu/drm/xe/xe_wa_oob.rules | 1 + 4 files changed, 68 insertions(+), 59 deletions(-) -- 2.48.1 ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-11 0:48 [PATCH v3 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit @ 2026-04-11 0:48 ` Ashutosh Dixit 0 siblings, 0 replies; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-11 0:48 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa OA code should have used xe_map layer to begin with. In CRI, the OA buffer can be located both in system and device memory. For these reasons, move OA code to use the xe_map layer when accessing the OA buffer. v2: Use xe_map layer also in xe_oa_copy_to_user (Umesh) v3: To avoid performance impact in v2, revert to v1 but move xe_map_copy_to_user() to xe_map.h Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> --- drivers/gpu/drm/xe/xe_map.h | 14 ++++++ drivers/gpu/drm/xe/xe_oa.c | 80 ++++++++++++++++---------------- drivers/gpu/drm/xe/xe_oa_types.h | 3 -- 3 files changed, 55 insertions(+), 42 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_map.h b/drivers/gpu/drm/xe/xe_map.h index c447771256913..8007fcb342880 100644 --- a/drivers/gpu/drm/xe/xe_map.h +++ b/drivers/gpu/drm/xe/xe_map.h @@ -44,6 +44,20 @@ static inline void xe_map_memset(struct xe_device *xe, iosys_map_memset(dst, offset, value, len); } +static inline unsigned long +xe_map_copy_to_user(struct xe_device *xe, void __user *dst, + const struct iosys_map *src, + size_t src_offset, size_t len) +{ + xe_device_assert_mem_access(xe); + + /* Because there is no iosys_map_copy_to_user() or copy_to_user_from_iomem() */ + if (src->is_iomem) + return copy_to_user(dst, src->vaddr_iomem + src_offset, len); + else + return copy_to_user(dst, src->vaddr + src_offset, len); +} + /* FIXME: We likely should kill these two functions sooner or later */ static inline u32 xe_map_read32(struct xe_device *xe, struct iosys_map *map) { diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index 6337e671c97ae..d25beb6ac6040 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) #define oa_report_header_64bit(__s) \ ((__s)->oa_buffer.format->header == HDR_64_BIT) -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) +static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) { - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + return oa_report_header_64bit(stream) ? + xe_map_rd(stream->oa->xe, map, tail, u64) : + xe_map_rd(stream->oa->xe, map, tail, u32); } -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)report = 0; - else - *report = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head, u64, 0) : + xe_map_wr(stream->oa->xe, map, head, u32, 0); } -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + return oa_report_header_64bit(stream) ? - *((u64 *)report + 1) : - *((u32 *)report + 1); + xe_map_rd(stream->oa->xe, map, tail + 2, u64) : + xe_map_rd(stream->oa->xe, map, tail + 1, u32); } -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)&report[2] = 0; - else - report[1] = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : + xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); } static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) * they were written. If not : (╯°□°）╯︵ ┻━┻ */ while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { - void *report = stream->oa_buffer.vaddr + tail; - - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) break; tail = xe_oa_circ_diff(stream, tail, report_size); @@ -312,29 +318,28 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) } static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, - size_t count, size_t *offset, const u8 *report) + size_t count, size_t *offset, u32 head) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; int report_size = stream->oa_buffer.format->size; + struct xe_device *xe = stream->oa->xe; int report_size_partial; - u8 *oa_buf_end; if ((count - *offset) < report_size) return -ENOSPC; buf += *offset; - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - report_size_partial = oa_buf_end - report; + report_size_partial = stream->oa_buffer.circ_size - head; if (report_size_partial < report_size) { - if (copy_to_user(buf, report, report_size_partial)) + if (xe_map_copy_to_user(xe, buf, map, head, report_size_partial)) return -EFAULT; buf += report_size_partial; - if (copy_to_user(buf, stream->oa_buffer.vaddr, - report_size - report_size_partial)) + if (xe_map_copy_to_user(xe, buf, map, 0, report_size - report_size_partial)) return -EFAULT; - } else if (copy_to_user(buf, report, report_size)) { + } else if (xe_map_copy_to_user(xe, buf, map, head, report_size)) { return -EFAULT; } @@ -347,7 +352,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, size_t count, size_t *offset) { int report_size = stream->oa_buffer.format->size; - u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); size_t start_offset = *offset; unsigned long flags; @@ -364,26 +368,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, for (; xe_oa_circ_diff(stream, tail, head); head = xe_oa_circ_incr(stream, head, report_size)) { - u8 *report = oa_buf_base + head; - - ret = xe_oa_append_report(stream, buf, count, offset, report); + ret = xe_oa_append_report(stream, buf, count, offset, head); if (ret) break; if (!(stream->oa_buffer.circ_size % report_size)) { /* Clear out report id and timestamp to detect unlanded reports */ - oa_report_id_clear(stream, (void *)report); - oa_timestamp_clear(stream, (void *)report); + oa_report_id_clear(stream, head); + oa_timestamp_clear(stream, head); } else { - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - u32 part = oa_buf_end - report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + u32 part = stream->oa_buffer.circ_size - head; /* Zero out the entire report */ if (report_size <= part) { - memset(report, 0, report_size); + xe_map_memset(stream->oa->xe, map, head, 0, report_size); } else { - memset(report, 0, part); - memset(oa_buf_base, 0, report_size - part); + xe_map_memset(stream->oa->xe, map, head, 0, part); + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); } } } @@ -436,7 +438,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, + xe_bo_size(stream->oa_buffer.bo)); } static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) @@ -891,7 +894,6 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) stream->oa_buffer.bo = bo; /* mmap implementation requires OA buffer to be in system memory */ xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); - stream->oa_buffer.vaddr = bo->vmap.vaddr; return 0; } diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h index b03ffd5134834..2dd550c93b8ac 100644 --- a/drivers/gpu/drm/xe/xe_oa_types.h +++ b/drivers/gpu/drm/xe/xe_oa_types.h @@ -162,9 +162,6 @@ struct xe_oa_buffer { /** @format: xe_bo backing the OA buffer */ struct xe_bo *bo; - /** @vaddr: mapped vaddr of the OA buffer */ - u8 *vaddr; - /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ spinlock_t ptr_lock; -- 2.48.1 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* [PATCH v4 0/3] drm/xe/oa: Wa_14026633728 @ 2026-04-15 2:03 Ashutosh Dixit 2026-04-15 2:03 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-15 2:03 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa v4: Change to Patch 1 again (see changelog in Patch 1) v3: Change to Patch 1 again (see changelog in Patch 1) v2: Change to Patch 1 Ashutosh Dixit (3): drm/xe/oa: Use xe_map layer drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap drm/xe/oa: Implement Wa_14026633728 drivers/gpu/drm/xe/xe_oa.c | 123 ++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- drivers/gpu/drm/xe/xe_wa_oob.rules | 1 + 3 files changed, 70 insertions(+), 58 deletions(-) -- 2.48.1 ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-15 2:03 [PATCH v4 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit @ 2026-04-15 2:03 ` Ashutosh Dixit 2026-04-24 21:05 ` Umesh Nerlige Ramappa 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-15 2:03 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa OA code should have used xe_map layer to begin with. In CRI, the OA buffer can be located both in system and device memory. For these reasons, move OA code to use the xe_map layer when accessing the OA buffer. v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) v3: To avoid performance impact in v2, revert to v1 but move xe_map_copy_to_user() to xe_map.h v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> --- drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- 2 files changed, 58 insertions(+), 41 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index 6337e671c97ae..4d13d8ef8103f 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) #define oa_report_header_64bit(__s) \ ((__s)->oa_buffer.format->header == HDR_64_BIT) -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) +static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) { - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + return oa_report_header_64bit(stream) ? + xe_map_rd(stream->oa->xe, map, tail, u64) : + xe_map_rd(stream->oa->xe, map, tail, u32); } -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)report = 0; - else - *report = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head, u64, 0) : + xe_map_wr(stream->oa->xe, map, head, u32, 0); } -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + return oa_report_header_64bit(stream) ? - *((u64 *)report + 1) : - *((u32 *)report + 1); + xe_map_rd(stream->oa->xe, map, tail + 2, u64) : + xe_map_rd(stream->oa->xe, map, tail + 1, u32); } -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)&report[2] = 0; - else - report[1] = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : + xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); } static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) * they were written. If not : (╯°□°）╯︵ ┻━┻ */ while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { - void *report = stream->oa_buffer.vaddr + tail; - - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) break; tail = xe_oa_circ_diff(stream, tail, report_size); @@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) return HRTIMER_RESTART; } +static unsigned long +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) +{ + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, + &stream->oa_buffer.bo->vmap, head, len); + return copy_to_user(dst, stream->oa_buffer.bounce, len); +} + static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, - size_t count, size_t *offset, const u8 *report) + size_t count, size_t *offset, u32 head) { int report_size = stream->oa_buffer.format->size; int report_size_partial; - u8 *oa_buf_end; if ((count - *offset) < report_size) return -ENOSPC; buf += *offset; - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - report_size_partial = oa_buf_end - report; + report_size_partial = stream->oa_buffer.circ_size - head; if (report_size_partial < report_size) { - if (copy_to_user(buf, report, report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, head, report_size_partial)) return -EFAULT; buf += report_size_partial; - if (copy_to_user(buf, stream->oa_buffer.vaddr, - report_size - report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) return -EFAULT; - } else if (copy_to_user(buf, report, report_size)) { + } else if (xe_oa_copy_to_user(stream, buf, head, report_size)) { return -EFAULT; } @@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, size_t count, size_t *offset) { int report_size = stream->oa_buffer.format->size; - u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); size_t start_offset = *offset; unsigned long flags; @@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, for (; xe_oa_circ_diff(stream, tail, head); head = xe_oa_circ_incr(stream, head, report_size)) { - u8 *report = oa_buf_base + head; - - ret = xe_oa_append_report(stream, buf, count, offset, report); + ret = xe_oa_append_report(stream, buf, count, offset, head); if (ret) break; if (!(stream->oa_buffer.circ_size % report_size)) { /* Clear out report id and timestamp to detect unlanded reports */ - oa_report_id_clear(stream, (void *)report); - oa_timestamp_clear(stream, (void *)report); + oa_report_id_clear(stream, head); + oa_timestamp_clear(stream, head); } else { - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - u32 part = oa_buf_end - report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + u32 part = stream->oa_buffer.circ_size - head; /* Zero out the entire report */ if (report_size <= part) { - memset(report, 0, report_size); + xe_map_memset(stream->oa->xe, map, head, 0, report_size); } else { - memset(report, 0, part); - memset(oa_buf_base, 0, report_size - part); + xe_map_memset(stream->oa->xe, map, head, 0, part); + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); } } } @@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, + xe_bo_size(stream->oa_buffer.bo)); } static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) @@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) { xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + kfree(stream->oa_buffer.bounce); } static void xe_oa_free_configs(struct xe_oa_stream *stream) @@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) return PTR_ERR(bo); stream->oa_buffer.bo = bo; + /* mmap implementation requires OA buffer to be in system memory */ xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); - stream->oa_buffer.vaddr = bo->vmap.vaddr; + + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); + if (!stream->oa_buffer.bounce) { + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + return -ENOMEM; + } + return 0; } diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h index b03ffd5134834..99a04b8d586b5 100644 --- a/drivers/gpu/drm/xe/xe_oa_types.h +++ b/drivers/gpu/drm/xe/xe_oa_types.h @@ -162,8 +162,8 @@ struct xe_oa_buffer { /** @format: xe_bo backing the OA buffer */ struct xe_bo *bo; - /** @vaddr: mapped vaddr of the OA buffer */ - u8 *vaddr; + /** @bounce: bounce buffer used with xe_map layer */ + void *bounce; /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ spinlock_t ptr_lock; -- 2.48.1 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-15 2:03 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit @ 2026-04-24 21:05 ` Umesh Nerlige Ramappa 2026-04-24 23:56 ` Dixit, Ashutosh 0 siblings, 1 reply; 30+ messages in thread From: Umesh Nerlige Ramappa @ 2026-04-24 21:05 UTC (permalink / raw) To: Ashutosh Dixit; +Cc: intel-xe On Tue, Apr 14, 2026 at 07:03:40PM -0700, Ashutosh Dixit wrote: >OA code should have used xe_map layer to begin with. In CRI, the OA buffer >can be located both in system and device memory. For these reasons, move OA >code to use the xe_map layer when accessing the OA buffer. > >v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) >v3: To avoid performance impact in v2, revert to v1 but move > xe_map_copy_to_user() to xe_map.h >v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user > >Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> >--- > drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- > drivers/gpu/drm/xe/xe_oa_types.h | 4 +- > 2 files changed, 58 insertions(+), 41 deletions(-) > >diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c >index 6337e671c97ae..4d13d8ef8103f 100644 >--- a/drivers/gpu/drm/xe/xe_oa.c >+++ b/drivers/gpu/drm/xe/xe_oa.c >@@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) > #define oa_report_header_64bit(__s) \ > ((__s)->oa_buffer.format->header == HDR_64_BIT) > >-static u64 oa_report_id(struct xe_oa_stream *stream, void *report) >+static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) > { >- return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ return oa_report_header_64bit(stream) ? >+ xe_map_rd(stream->oa->xe, map, tail, u64) : >+ xe_map_rd(stream->oa->xe, map, tail, u32); > } (1) The helper is just returning the report id for a report passed. In this case it is tail, but the code would work for any report. I would keep the name as report or report_offset. Let the caller use specific names like tail. > >-static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) >+static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) > { >- if (oa_report_header_64bit(stream)) >- *(u64 *)report = 0; >- else >- *report = 0; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ oa_report_header_64bit(stream) ? >+ xe_map_wr(stream->oa->xe, map, head, u64, 0) : >+ xe_map_wr(stream->oa->xe, map, head, u32, 0); > } > >-static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) >+static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) > { >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ > return oa_report_header_64bit(stream) ? >- *((u64 *)report + 1) : >- *((u32 *)report + 1); >+ xe_map_rd(stream->oa->xe, map, tail + 2, u64) : >+ xe_map_rd(stream->oa->xe, map, tail + 1, u32); (2) I think this should be tail + 8 and tail + 4. Can you please double check? The u32/u64 is only being used to access the data at the offset rather than determine how to calculate the actual offset. > } > >-static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) >+static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) > { >- if (oa_report_header_64bit(stream)) >- *(u64 *)&report[2] = 0; >- else >- report[1] = 0; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ oa_report_header_64bit(stream) ? >+ xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : >+ xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); Same here - comment (2) > } Same comment (1) for all the above helpers - the name can be generic - report_offset. I see similar change in other functions below as well. If you think it is necessary, I suggest trying to move it to a separate patch with commit message explaining why you want to rename them. That would also help review this patch easily. > > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) >@@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > * they were written. If not : (╯°□°）╯︵ ┻━┻ > */ > while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { >- void *report = stream->oa_buffer.vaddr + tail; >- >- if (oa_report_id(stream, report) || oa_timestamp(stream, report)) >+ if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) > break; > > tail = xe_oa_circ_diff(stream, tail, report_size); >@@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) > return HRTIMER_RESTART; > } > >+static unsigned long >+xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) >+{ >+ xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, >+ &stream->oa_buffer.bo->vmap, head, len); >+ return copy_to_user(dst, stream->oa_buffer.bounce, len); >+} >+ Still need to look at the patch below this point. Will send that out next week... Regards, Umesh > static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, >- size_t count, size_t *offset, const u8 *report) >+ size_t count, size_t *offset, u32 head) > { > int report_size = stream->oa_buffer.format->size; > int report_size_partial; >- u8 *oa_buf_end; > > if ((count - *offset) < report_size) > return -ENOSPC; > > buf += *offset; > >- oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >- report_size_partial = oa_buf_end - report; >+ report_size_partial = stream->oa_buffer.circ_size - head; > > if (report_size_partial < report_size) { >- if (copy_to_user(buf, report, report_size_partial)) >+ if (xe_oa_copy_to_user(stream, buf, head, report_size_partial)) > return -EFAULT; > buf += report_size_partial; > >- if (copy_to_user(buf, stream->oa_buffer.vaddr, >- report_size - report_size_partial)) >+ if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) > return -EFAULT; >- } else if (copy_to_user(buf, report, report_size)) { >+ } else if (xe_oa_copy_to_user(stream, buf, head, report_size)) { > return -EFAULT; > } > >@@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > size_t count, size_t *offset) > { > int report_size = stream->oa_buffer.format->size; >- u8 *oa_buf_base = stream->oa_buffer.vaddr; > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); > size_t start_offset = *offset; > unsigned long flags; >@@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > for (; xe_oa_circ_diff(stream, tail, head); > head = xe_oa_circ_incr(stream, head, report_size)) { >- u8 *report = oa_buf_base + head; >- >- ret = xe_oa_append_report(stream, buf, count, offset, report); >+ ret = xe_oa_append_report(stream, buf, count, offset, head); > if (ret) > break; > > if (!(stream->oa_buffer.circ_size % report_size)) { > /* Clear out report id and timestamp to detect unlanded reports */ >- oa_report_id_clear(stream, (void *)report); >- oa_timestamp_clear(stream, (void *)report); >+ oa_report_id_clear(stream, head); >+ oa_timestamp_clear(stream, head); > } else { >- u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >- u32 part = oa_buf_end - report; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ u32 part = stream->oa_buffer.circ_size - head; > > /* Zero out the entire report */ > if (report_size <= part) { >- memset(report, 0, report_size); >+ xe_map_memset(stream->oa->xe, map, head, 0, report_size); > } else { >- memset(report, 0, part); >- memset(oa_buf_base, 0, report_size - part); >+ xe_map_memset(stream->oa->xe, map, head, 0, part); >+ xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); > } > } > } >@@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) > spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); > > /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ >- memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); >+ xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, >+ xe_bo_size(stream->oa_buffer.bo)); > } > > static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) >@@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) > static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) > { > xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); >+ kfree(stream->oa_buffer.bounce); > } > > static void xe_oa_free_configs(struct xe_oa_stream *stream) >@@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > return PTR_ERR(bo); > > stream->oa_buffer.bo = bo; >+ > /* mmap implementation requires OA buffer to be in system memory */ > xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); >- stream->oa_buffer.vaddr = bo->vmap.vaddr; >+ >+ stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); >+ if (!stream->oa_buffer.bounce) { >+ xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); >+ return -ENOMEM; >+ } >+ > return 0; > } > >diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h >index b03ffd5134834..99a04b8d586b5 100644 >--- a/drivers/gpu/drm/xe/xe_oa_types.h >+++ b/drivers/gpu/drm/xe/xe_oa_types.h >@@ -162,8 +162,8 @@ struct xe_oa_buffer { > /** @format: xe_bo backing the OA buffer */ > struct xe_bo *bo; > >- /** @vaddr: mapped vaddr of the OA buffer */ >- u8 *vaddr; >+ /** @bounce: bounce buffer used with xe_map layer */ >+ void *bounce; > > /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ > spinlock_t ptr_lock; >-- >2.48.1 > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-24 21:05 ` Umesh Nerlige Ramappa @ 2026-04-24 23:56 ` Dixit, Ashutosh 2026-04-25 0:16 ` Dixit, Ashutosh 0 siblings, 1 reply; 30+ messages in thread From: Dixit, Ashutosh @ 2026-04-24 23:56 UTC (permalink / raw) To: Umesh Nerlige Ramappa; +Cc: intel-xe On Fri, 24 Apr 2026 14:05:16 -0700, Umesh Nerlige Ramappa wrote: > Hi Umesh, > On Tue, Apr 14, 2026 at 07:03:40PM -0700, Ashutosh Dixit wrote: > > OA code should have used xe_map layer to begin with. In CRI, the OA buffer > > can be located both in system and device memory. For these reasons, move OA > > code to use the xe_map layer when accessing the OA buffer. > > > > v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) > > v3: To avoid performance impact in v2, revert to v1 but move > > xe_map_copy_to_user() to xe_map.h > > v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user > > > > Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> > > --- > > drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- > > drivers/gpu/drm/xe/xe_oa_types.h | 4 +- > > 2 files changed, 58 insertions(+), 41 deletions(-) > > > > diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c > > index 6337e671c97ae..4d13d8ef8103f 100644 > > --- a/drivers/gpu/drm/xe/xe_oa.c > > +++ b/drivers/gpu/drm/xe/xe_oa.c > > @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) > > #define oa_report_header_64bit(__s) \ > > ((__s)->oa_buffer.format->header == HDR_64_BIT) > > > > -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) > > +static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) > > { > > - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + > > + return oa_report_header_64bit(stream) ? > > + xe_map_rd(stream->oa->xe, map, tail, u64) : > > + xe_map_rd(stream->oa->xe, map, tail, u32); > > } > > (1) The helper is just returning the report id for a report passed. In this > case it is tail, but the code would work for any report. I would keep the > name as report or report_offset. Let the caller use specific names like > tail. > > > > > -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) > > +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) > > { > > - if (oa_report_header_64bit(stream)) > > - *(u64 *)report = 0; > > - else > > - *report = 0; > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + > > + oa_report_header_64bit(stream) ? > > + xe_map_wr(stream->oa->xe, map, head, u64, 0) : > > + xe_map_wr(stream->oa->xe, map, head, u32, 0); > > } > > > > -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) > > +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) > > { > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + > > return oa_report_header_64bit(stream) ? > > - *((u64 *)report + 1) : > > - *((u32 *)report + 1); > > + xe_map_rd(stream->oa->xe, map, tail + 2, u64) : > > + xe_map_rd(stream->oa->xe, map, tail + 1, u32); > > (2) I think this should be tail + 8 and tail + 4. Can you please double > check? The u32/u64 is only being used to access the data at the offset > rather than determine how to calculate the actual offset. Hmm, you are right!!! I don't believe I made such a basic mistake :/ And all the xe_oa tests work fine even with this. Anyway thanks for catching this, I'm fixing this in the next rev. > > > } > > > > -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) > > +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) > > { > > - if (oa_report_header_64bit(stream)) > > - *(u64 *)&report[2] = 0; > > - else > > - report[1] = 0; > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + > > + oa_report_header_64bit(stream) ? > > + xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : > > + xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); > > Same here - comment (2) > > > } > > Same comment (1) for all the above helpers - the name can be generic - > report_offset. I see similar change in other functions below as well. If > you think it is necessary, I suggest trying to move it to a separate patch > with commit message explaining why you want to rename them. That would also > help review this patch easily. So, as we know HW writes to tail and SW reads from head. So the helpers with head as the arg are invoked from the read side only and the helpers with tail as the arg are invoked from the write side only. That was the reason for naming the arg as head/tail, rather than just "offset". I thought it will make the code easier to understand and also maybe prevent mistakes in the future. So anyway, let me ask once again if you think I should change it. Of course, there's no problem changing the arg name in the helpers. I'll also see if it's possible to split the patch some way to make the review a bit easier. Thanks. -- Ashutosh > > > > > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > > @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > > * they were written. If not : (╯°□°）╯︵ ┻━┻ > > */ > > while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { > > - void *report = stream->oa_buffer.vaddr + tail; > > - > > - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) > > + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) > > break; > > > > tail = xe_oa_circ_diff(stream, tail, report_size); > > @@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) > > return HRTIMER_RESTART; > > } > > > > +static unsigned long > > +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) > > +{ > > + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, > > + &stream->oa_buffer.bo->vmap, head, len); > > + return copy_to_user(dst, stream->oa_buffer.bounce, len); > > +} > > + > > Still need to look at the patch below this point. Will send that out next > week... > > Regards, > Umesh > > > static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, > > - size_t count, size_t *offset, const u8 *report) > > + size_t count, size_t *offset, u32 head) > > { > > int report_size = stream->oa_buffer.format->size; > > int report_size_partial; > > - u8 *oa_buf_end; > > > > if ((count - *offset) < report_size) > > return -ENOSPC; > > > > buf += *offset; > > > > - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; > > - report_size_partial = oa_buf_end - report; > > + report_size_partial = stream->oa_buffer.circ_size - head; > > > > if (report_size_partial < report_size) { > > - if (copy_to_user(buf, report, report_size_partial)) > > + if (xe_oa_copy_to_user(stream, buf, head, report_size_partial)) > > return -EFAULT; > > buf += report_size_partial; > > > > - if (copy_to_user(buf, stream->oa_buffer.vaddr, > > - report_size - report_size_partial)) > > + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) > > return -EFAULT; > > - } else if (copy_to_user(buf, report, report_size)) { > > + } else if (xe_oa_copy_to_user(stream, buf, head, report_size)) { > > return -EFAULT; > > } > > > > @@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > size_t count, size_t *offset) > > { > > int report_size = stream->oa_buffer.format->size; > > - u8 *oa_buf_base = stream->oa_buffer.vaddr; > > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); > > size_t start_offset = *offset; > > unsigned long flags; > > @@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > > > for (; xe_oa_circ_diff(stream, tail, head); > > head = xe_oa_circ_incr(stream, head, report_size)) { > > - u8 *report = oa_buf_base + head; > > - > > - ret = xe_oa_append_report(stream, buf, count, offset, report); > > + ret = xe_oa_append_report(stream, buf, count, offset, head); > > if (ret) > > break; > > > > if (!(stream->oa_buffer.circ_size % report_size)) { > > /* Clear out report id and timestamp to detect unlanded reports */ > > - oa_report_id_clear(stream, (void *)report); > > - oa_timestamp_clear(stream, (void *)report); > > + oa_report_id_clear(stream, head); > > + oa_timestamp_clear(stream, head); > > } else { > > - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; > > - u32 part = oa_buf_end - report; > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + u32 part = stream->oa_buffer.circ_size - head; > > > > /* Zero out the entire report */ > > if (report_size <= part) { > > - memset(report, 0, report_size); > > + xe_map_memset(stream->oa->xe, map, head, 0, report_size); > > } else { > > - memset(report, 0, part); > > - memset(oa_buf_base, 0, report_size - part); > > + xe_map_memset(stream->oa->xe, map, head, 0, part); > > + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); > > } > > } > > } > > @@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) > > spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); > > > > /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ > > - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); > > + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, > > + xe_bo_size(stream->oa_buffer.bo)); > > } > > > > static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) > > @@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) > > static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) > > { > > xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); > > + kfree(stream->oa_buffer.bounce); > > } > > > > static void xe_oa_free_configs(struct xe_oa_stream *stream) > > @@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > > return PTR_ERR(bo); > > > > stream->oa_buffer.bo = bo; > > + > > /* mmap implementation requires OA buffer to be in system memory */ > > xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); > > - stream->oa_buffer.vaddr = bo->vmap.vaddr; > > + > > + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); > > + if (!stream->oa_buffer.bounce) { > > + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); > > + return -ENOMEM; > > + } > > + > > return 0; > > } > > > > diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h > > index b03ffd5134834..99a04b8d586b5 100644 > > --- a/drivers/gpu/drm/xe/xe_oa_types.h > > +++ b/drivers/gpu/drm/xe/xe_oa_types.h > > @@ -162,8 +162,8 @@ struct xe_oa_buffer { > > /** @format: xe_bo backing the OA buffer */ > > struct xe_bo *bo; > > > > - /** @vaddr: mapped vaddr of the OA buffer */ > > - u8 *vaddr; > > + /** @bounce: bounce buffer used with xe_map layer */ > > + void *bounce; > > > > /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ > > spinlock_t ptr_lock; > > -- > > 2.48.1 > > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-24 23:56 ` Dixit, Ashutosh @ 2026-04-25 0:16 ` Dixit, Ashutosh 2026-04-27 18:27 ` Umesh Nerlige Ramappa 0 siblings, 1 reply; 30+ messages in thread From: Dixit, Ashutosh @ 2026-04-25 0:16 UTC (permalink / raw) To: Umesh Nerlige Ramappa; +Cc: intel-xe On Fri, 24 Apr 2026 16:56:04 -0700, Dixit, Ashutosh wrote: > > On Fri, 24 Apr 2026 14:05:16 -0700, Umesh Nerlige Ramappa wrote: > > > > Hi Umesh, > > > On Tue, Apr 14, 2026 at 07:03:40PM -0700, Ashutosh Dixit wrote: > > > OA code should have used xe_map layer to begin with. In CRI, the OA buffer > > > can be located both in system and device memory. For these reasons, move OA > > > code to use the xe_map layer when accessing the OA buffer. > > > > > > v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) > > > v3: To avoid performance impact in v2, revert to v1 but move > > > xe_map_copy_to_user() to xe_map.h > > > v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user > > > > > > Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> > > > --- > > > drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- > > > drivers/gpu/drm/xe/xe_oa_types.h | 4 +- > > > 2 files changed, 58 insertions(+), 41 deletions(-) > > > > > > diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c > > > index 6337e671c97ae..4d13d8ef8103f 100644 > > > --- a/drivers/gpu/drm/xe/xe_oa.c > > > +++ b/drivers/gpu/drm/xe/xe_oa.c > > > @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) > > > #define oa_report_header_64bit(__s) \ > > > ((__s)->oa_buffer.format->header == HDR_64_BIT) > > > > > > -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) > > > +static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) > > > { > > > - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; > > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > > + > > > + return oa_report_header_64bit(stream) ? > > > + xe_map_rd(stream->oa->xe, map, tail, u64) : > > > + xe_map_rd(stream->oa->xe, map, tail, u32); > > > } > > > > (1) The helper is just returning the report id for a report passed. In this > > case it is tail, but the code would work for any report. I would keep the > > name as report or report_offset. Let the caller use specific names like > > tail. > > > > > > > > -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) > > > +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) > > > { > > > - if (oa_report_header_64bit(stream)) > > > - *(u64 *)report = 0; > > > - else > > > - *report = 0; > > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > > + > > > + oa_report_header_64bit(stream) ? > > > + xe_map_wr(stream->oa->xe, map, head, u64, 0) : > > > + xe_map_wr(stream->oa->xe, map, head, u32, 0); > > > } > > > > > > -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) > > > +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) > > > { > > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > > + > > > return oa_report_header_64bit(stream) ? > > > - *((u64 *)report + 1) : > > > - *((u32 *)report + 1); > > > + xe_map_rd(stream->oa->xe, map, tail + 2, u64) : > > > + xe_map_rd(stream->oa->xe, map, tail + 1, u32); > > > > (2) I think this should be tail + 8 and tail + 4. Can you please double > > check? The u32/u64 is only being used to access the data at the offset > > rather than determine how to calculate the actual offset. > > Hmm, you are right!!! I don't believe I made such a basic mistake :/ And > all the xe_oa tests work fine even with this. > > Anyway thanks for catching this, I'm fixing this in the next rev. > > > > > > } > > > > > > -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) > > > +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) > > > { > > > - if (oa_report_header_64bit(stream)) > > > - *(u64 *)&report[2] = 0; > > > - else > > > - report[1] = 0; > > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > > + > > > + oa_report_header_64bit(stream) ? > > > + xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : > > > + xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); > > > > Same here - comment (2) > > > > > } > > > > Same comment (1) for all the above helpers - the name can be generic - > > report_offset. I see similar change in other functions below as well. If > > you think it is necessary, I suggest trying to move it to a separate patch > > with commit message explaining why you want to rename them. That would also > > help review this patch easily. > > So, as we know HW writes to tail and SW reads from head. So the helpers > with head as the arg are invoked from the read side only and the helpers > with tail as the arg are invoked from the write side only. That was the > reason for naming the arg as head/tail, rather than just "offset". I > thought it will make the code easier to understand and also maybe prevent > mistakes in the future. > > So anyway, let me ask once again if you think I should change it. Of > course, there's no problem changing the arg name in the helpers. > > I'll also see if it's possible to split the patch some way to make the > review a bit easier. I just sent out v5 in which I have let all this stay as is. Anyway, please let me know what you prefer. Thanks. > > Thanks. > -- > Ashutosh > > > > > > > > > > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > > > @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > > > * they were written. If not : (╯°□°）╯︵ ┻━┻ > > > */ > > > while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { > > > - void *report = stream->oa_buffer.vaddr + tail; > > > - > > > - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) > > > + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) > > > break; > > > > > > tail = xe_oa_circ_diff(stream, tail, report_size); > > > @@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) > > > return HRTIMER_RESTART; > > > } > > > > > > +static unsigned long > > > +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) > > > +{ > > > + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, > > > + &stream->oa_buffer.bo->vmap, head, len); > > > + return copy_to_user(dst, stream->oa_buffer.bounce, len); > > > +} > > > + > > > > Still need to look at the patch below this point. Will send that out next > > week... > > > > Regards, > > Umesh > > > > > static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, > > > - size_t count, size_t *offset, const u8 *report) > > > + size_t count, size_t *offset, u32 head) > > > { > > > int report_size = stream->oa_buffer.format->size; > > > int report_size_partial; > > > - u8 *oa_buf_end; > > > > > > if ((count - *offset) < report_size) > > > return -ENOSPC; > > > > > > buf += *offset; > > > > > > - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; > > > - report_size_partial = oa_buf_end - report; > > > + report_size_partial = stream->oa_buffer.circ_size - head; > > > > > > if (report_size_partial < report_size) { > > > - if (copy_to_user(buf, report, report_size_partial)) > > > + if (xe_oa_copy_to_user(stream, buf, head, report_size_partial)) > > > return -EFAULT; > > > buf += report_size_partial; > > > > > > - if (copy_to_user(buf, stream->oa_buffer.vaddr, > > > - report_size - report_size_partial)) > > > + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) > > > return -EFAULT; > > > - } else if (copy_to_user(buf, report, report_size)) { > > > + } else if (xe_oa_copy_to_user(stream, buf, head, report_size)) { > > > return -EFAULT; > > > } > > > > > > @@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > > size_t count, size_t *offset) > > > { > > > int report_size = stream->oa_buffer.format->size; > > > - u8 *oa_buf_base = stream->oa_buffer.vaddr; > > > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); > > > size_t start_offset = *offset; > > > unsigned long flags; > > > @@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > > > > > for (; xe_oa_circ_diff(stream, tail, head); > > > head = xe_oa_circ_incr(stream, head, report_size)) { > > > - u8 *report = oa_buf_base + head; > > > - > > > - ret = xe_oa_append_report(stream, buf, count, offset, report); > > > + ret = xe_oa_append_report(stream, buf, count, offset, head); > > > if (ret) > > > break; > > > > > > if (!(stream->oa_buffer.circ_size % report_size)) { > > > /* Clear out report id and timestamp to detect unlanded reports */ > > > - oa_report_id_clear(stream, (void *)report); > > > - oa_timestamp_clear(stream, (void *)report); > > > + oa_report_id_clear(stream, head); > > > + oa_timestamp_clear(stream, head); > > > } else { > > > - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; > > > - u32 part = oa_buf_end - report; > > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > > + u32 part = stream->oa_buffer.circ_size - head; > > > > > > /* Zero out the entire report */ > > > if (report_size <= part) { > > > - memset(report, 0, report_size); > > > + xe_map_memset(stream->oa->xe, map, head, 0, report_size); > > > } else { > > > - memset(report, 0, part); > > > - memset(oa_buf_base, 0, report_size - part); > > > + xe_map_memset(stream->oa->xe, map, head, 0, part); > > > + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); > > > } > > > } > > > } > > > @@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) > > > spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); > > > > > > /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ > > > - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); > > > + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, > > > + xe_bo_size(stream->oa_buffer.bo)); > > > } > > > > > > static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) > > > @@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) > > > static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) > > > { > > > xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); > > > + kfree(stream->oa_buffer.bounce); > > > } > > > > > > static void xe_oa_free_configs(struct xe_oa_stream *stream) > > > @@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > > > return PTR_ERR(bo); > > > > > > stream->oa_buffer.bo = bo; > > > + > > > /* mmap implementation requires OA buffer to be in system memory */ > > > xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); > > > - stream->oa_buffer.vaddr = bo->vmap.vaddr; > > > + > > > + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); > > > + if (!stream->oa_buffer.bounce) { > > > + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); > > > + return -ENOMEM; > > > + } > > > + > > > return 0; > > > } > > > > > > diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h > > > index b03ffd5134834..99a04b8d586b5 100644 > > > --- a/drivers/gpu/drm/xe/xe_oa_types.h > > > +++ b/drivers/gpu/drm/xe/xe_oa_types.h > > > @@ -162,8 +162,8 @@ struct xe_oa_buffer { > > > /** @format: xe_bo backing the OA buffer */ > > > struct xe_bo *bo; > > > > > > - /** @vaddr: mapped vaddr of the OA buffer */ > > > - u8 *vaddr; > > > + /** @bounce: bounce buffer used with xe_map layer */ > > > + void *bounce; > > > > > > /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ > > > spinlock_t ptr_lock; > > > -- > > > 2.48.1 > > > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-25 0:16 ` Dixit, Ashutosh @ 2026-04-27 18:27 ` Umesh Nerlige Ramappa 2026-04-27 19:04 ` Dixit, Ashutosh 0 siblings, 1 reply; 30+ messages in thread From: Umesh Nerlige Ramappa @ 2026-04-27 18:27 UTC (permalink / raw) To: Dixit, Ashutosh; +Cc: intel-xe On Fri, Apr 24, 2026 at 05:16:48PM -0700, Dixit, Ashutosh wrote: >On Fri, 24 Apr 2026 16:56:04 -0700, Dixit, Ashutosh wrote: >> >> On Fri, 24 Apr 2026 14:05:16 -0700, Umesh Nerlige Ramappa wrote: >> > >> >> Hi Umesh, >> >> > On Tue, Apr 14, 2026 at 07:03:40PM -0700, Ashutosh Dixit wrote: >> > > OA code should have used xe_map layer to begin with. In CRI, the OA buffer >> > > can be located both in system and device memory. For these reasons, move OA >> > > code to use the xe_map layer when accessing the OA buffer. >> > > >> > > v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) >> > > v3: To avoid performance impact in v2, revert to v1 but move >> > > xe_map_copy_to_user() to xe_map.h >> > > v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user >> > > >> > > Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> >> > > --- >> > > drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- >> > > drivers/gpu/drm/xe/xe_oa_types.h | 4 +- >> > > 2 files changed, 58 insertions(+), 41 deletions(-) >> > > >> > > diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c >> > > index 6337e671c97ae..4d13d8ef8103f 100644 >> > > --- a/drivers/gpu/drm/xe/xe_oa.c >> > > +++ b/drivers/gpu/drm/xe/xe_oa.c >> > > @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) >> > > #define oa_report_header_64bit(__s) \ >> > > ((__s)->oa_buffer.format->header == HDR_64_BIT) >> > > >> > > -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) >> > > +static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) >> > > { >> > > - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; >> > > + >> > > + return oa_report_header_64bit(stream) ? >> > > + xe_map_rd(stream->oa->xe, map, tail, u64) : >> > > + xe_map_rd(stream->oa->xe, map, tail, u32); >> > > } >> > >> > (1) The helper is just returning the report id for a report passed. In this >> > case it is tail, but the code would work for any report. I would keep the >> > name as report or report_offset. Let the caller use specific names like >> > tail. >> > >> > > >> > > -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) >> > > +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) >> > > { >> > > - if (oa_report_header_64bit(stream)) >> > > - *(u64 *)report = 0; >> > > - else >> > > - *report = 0; >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; >> > > + >> > > + oa_report_header_64bit(stream) ? >> > > + xe_map_wr(stream->oa->xe, map, head, u64, 0) : >> > > + xe_map_wr(stream->oa->xe, map, head, u32, 0); >> > > } >> > > >> > > -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) >> > > +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) >> > > { >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; >> > > + >> > > return oa_report_header_64bit(stream) ? >> > > - *((u64 *)report + 1) : >> > > - *((u32 *)report + 1); >> > > + xe_map_rd(stream->oa->xe, map, tail + 2, u64) : >> > > + xe_map_rd(stream->oa->xe, map, tail + 1, u32); >> > >> > (2) I think this should be tail + 8 and tail + 4. Can you please double >> > check? The u32/u64 is only being used to access the data at the offset >> > rather than determine how to calculate the actual offset. >> >> Hmm, you are right!!! I don't believe I made such a basic mistake :/ And >> all the xe_oa tests work fine even with this. >> >> Anyway thanks for catching this, I'm fixing this in the next rev. >> >> > >> > > } >> > > >> > > -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) >> > > +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) >> > > { >> > > - if (oa_report_header_64bit(stream)) >> > > - *(u64 *)&report[2] = 0; >> > > - else >> > > - report[1] = 0; >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; >> > > + >> > > + oa_report_header_64bit(stream) ? >> > > + xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : >> > > + xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); >> > >> > Same here - comment (2) >> > >> > > } >> > >> > Same comment (1) for all the above helpers - the name can be generic - >> > report_offset. I see similar change in other functions below as well. If >> > you think it is necessary, I suggest trying to move it to a separate patch >> > with commit message explaining why you want to rename them. That would also >> > help review this patch easily. >> >> So, as we know HW writes to tail and SW reads from head. So the helpers >> with head as the arg are invoked from the read side only and the helpers >> with tail as the arg are invoked from the write side only. That was the >> reason for naming the arg as head/tail, rather than just "offset". I >> thought it will make the code easier to understand and also maybe prevent >> mistakes in the future. To me, the report_offset (or offset) is clearer. When I read the code I would remember this helper as returning report_id/timestamp given any report offset as an argument. The tail/head names makes me wonder if the caller has to be accounted for as well. Anyways, please add the above description you have to the commit message and you can leave it as is. Thanks, Umesh >> >> So anyway, let me ask once again if you think I should change it. Of >> course, there's no problem changing the arg name in the helpers. >> >> I'll also see if it's possible to split the patch some way to make the >> review a bit easier. > >I just sent out v5 in which I have let all this stay as is. Anyway, please >let me know what you prefer. Thanks. > >> >> Thanks. >> -- >> Ashutosh >> >> >> > >> > > >> > > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) >> > > @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) >> > > * they were written. If not : (╯°□°）╯︵ ┻━┻ >> > > */ >> > > while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { >> > > - void *report = stream->oa_buffer.vaddr + tail; >> > > - >> > > - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) >> > > + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) >> > > break; >> > > >> > > tail = xe_oa_circ_diff(stream, tail, report_size); >> > > @@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) >> > > return HRTIMER_RESTART; >> > > } >> > > >> > > +static unsigned long >> > > +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) >> > > +{ >> > > + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, >> > > + &stream->oa_buffer.bo->vmap, head, len); >> > > + return copy_to_user(dst, stream->oa_buffer.bounce, len); >> > > +} >> > > + >> > >> > Still need to look at the patch below this point. Will send that out next >> > week... >> > >> > Regards, >> > Umesh >> > >> > > static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, >> > > - size_t count, size_t *offset, const u8 *report) >> > > + size_t count, size_t *offset, u32 head) >> > > { >> > > int report_size = stream->oa_buffer.format->size; >> > > int report_size_partial; >> > > - u8 *oa_buf_end; >> > > >> > > if ((count - *offset) < report_size) >> > > return -ENOSPC; >> > > >> > > buf += *offset; >> > > >> > > - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >> > > - report_size_partial = oa_buf_end - report; >> > > + report_size_partial = stream->oa_buffer.circ_size - head; >> > > >> > > if (report_size_partial < report_size) { >> > > - if (copy_to_user(buf, report, report_size_partial)) >> > > + if (xe_oa_copy_to_user(stream, buf, head, report_size_partial)) >> > > return -EFAULT; >> > > buf += report_size_partial; >> > > >> > > - if (copy_to_user(buf, stream->oa_buffer.vaddr, >> > > - report_size - report_size_partial)) >> > > + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) >> > > return -EFAULT; >> > > - } else if (copy_to_user(buf, report, report_size)) { >> > > + } else if (xe_oa_copy_to_user(stream, buf, head, report_size)) { >> > > return -EFAULT; >> > > } >> > > >> > > @@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, >> > > size_t count, size_t *offset) >> > > { >> > > int report_size = stream->oa_buffer.format->size; >> > > - u8 *oa_buf_base = stream->oa_buffer.vaddr; >> > > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); >> > > size_t start_offset = *offset; >> > > unsigned long flags; >> > > @@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, >> > > >> > > for (; xe_oa_circ_diff(stream, tail, head); >> > > head = xe_oa_circ_incr(stream, head, report_size)) { >> > > - u8 *report = oa_buf_base + head; >> > > - >> > > - ret = xe_oa_append_report(stream, buf, count, offset, report); >> > > + ret = xe_oa_append_report(stream, buf, count, offset, head); >> > > if (ret) >> > > break; >> > > >> > > if (!(stream->oa_buffer.circ_size % report_size)) { >> > > /* Clear out report id and timestamp to detect unlanded reports */ >> > > - oa_report_id_clear(stream, (void *)report); >> > > - oa_timestamp_clear(stream, (void *)report); >> > > + oa_report_id_clear(stream, head); >> > > + oa_timestamp_clear(stream, head); >> > > } else { >> > > - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >> > > - u32 part = oa_buf_end - report; >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; >> > > + u32 part = stream->oa_buffer.circ_size - head; >> > > >> > > /* Zero out the entire report */ >> > > if (report_size <= part) { >> > > - memset(report, 0, report_size); >> > > + xe_map_memset(stream->oa->xe, map, head, 0, report_size); >> > > } else { >> > > - memset(report, 0, part); >> > > - memset(oa_buf_base, 0, report_size - part); >> > > + xe_map_memset(stream->oa->xe, map, head, 0, part); >> > > + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); >> > > } >> > > } >> > > } >> > > @@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) >> > > spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); >> > > >> > > /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ >> > > - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); >> > > + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, >> > > + xe_bo_size(stream->oa_buffer.bo)); >> > > } >> > > >> > > static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) >> > > @@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) >> > > static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) >> > > { >> > > xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); >> > > + kfree(stream->oa_buffer.bounce); >> > > } >> > > >> > > static void xe_oa_free_configs(struct xe_oa_stream *stream) >> > > @@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) >> > > return PTR_ERR(bo); >> > > >> > > stream->oa_buffer.bo = bo; >> > > + >> > > /* mmap implementation requires OA buffer to be in system memory */ >> > > xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); >> > > - stream->oa_buffer.vaddr = bo->vmap.vaddr; >> > > + >> > > + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); >> > > + if (!stream->oa_buffer.bounce) { >> > > + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); >> > > + return -ENOMEM; >> > > + } >> > > + >> > > return 0; >> > > } >> > > >> > > diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h >> > > index b03ffd5134834..99a04b8d586b5 100644 >> > > --- a/drivers/gpu/drm/xe/xe_oa_types.h >> > > +++ b/drivers/gpu/drm/xe/xe_oa_types.h >> > > @@ -162,8 +162,8 @@ struct xe_oa_buffer { >> > > /** @format: xe_bo backing the OA buffer */ >> > > struct xe_bo *bo; >> > > >> > > - /** @vaddr: mapped vaddr of the OA buffer */ >> > > - u8 *vaddr; >> > > + /** @bounce: bounce buffer used with xe_map layer */ >> > > + void *bounce; >> > > >> > > /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ >> > > spinlock_t ptr_lock; >> > > -- >> > > 2.48.1 >> > > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-27 18:27 ` Umesh Nerlige Ramappa @ 2026-04-27 19:04 ` Dixit, Ashutosh 0 siblings, 0 replies; 30+ messages in thread From: Dixit, Ashutosh @ 2026-04-27 19:04 UTC (permalink / raw) To: Umesh Nerlige Ramappa; +Cc: intel-xe On Mon, 27 Apr 2026 11:27:59 -0700, Umesh Nerlige Ramappa wrote: > > On Fri, Apr 24, 2026 at 05:16:48PM -0700, Dixit, Ashutosh wrote: > > On Fri, 24 Apr 2026 16:56:04 -0700, Dixit, Ashutosh wrote: > >> > >> On Fri, 24 Apr 2026 14:05:16 -0700, Umesh Nerlige Ramappa wrote: > >> > > >> > >> Hi Umesh, > >> > >> > On Tue, Apr 14, 2026 at 07:03:40PM -0700, Ashutosh Dixit wrote: > >> > > OA code should have used xe_map layer to begin with. In CRI, the OA buffer > >> > > can be located both in system and device memory. For these reasons, move OA > >> > > code to use the xe_map layer when accessing the OA buffer. > >> > > > >> > > v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) > >> > > v3: To avoid performance impact in v2, revert to v1 but move > >> > > xe_map_copy_to_user() to xe_map.h > >> > > v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user > >> > > > >> > > Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> > >> > > --- > >> > > drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- > >> > > drivers/gpu/drm/xe/xe_oa_types.h | 4 +- > >> > > 2 files changed, 58 insertions(+), 41 deletions(-) > >> > > > >> > > diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c > >> > > index 6337e671c97ae..4d13d8ef8103f 100644 > >> > > --- a/drivers/gpu/drm/xe/xe_oa.c > >> > > +++ b/drivers/gpu/drm/xe/xe_oa.c > >> > > @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) > >> > > #define oa_report_header_64bit(__s) \ > >> > > ((__s)->oa_buffer.format->header == HDR_64_BIT) > >> > > > >> > > -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) > >> > > +static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) > >> > > { > >> > > - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; > >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > >> > > + > >> > > + return oa_report_header_64bit(stream) ? > >> > > + xe_map_rd(stream->oa->xe, map, tail, u64) : > >> > > + xe_map_rd(stream->oa->xe, map, tail, u32); > >> > > } > >> > > >> > (1) The helper is just returning the report id for a report passed. In this > >> > case it is tail, but the code would work for any report. I would keep the > >> > name as report or report_offset. Let the caller use specific names like > >> > tail. > >> > > >> > > > >> > > -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) > >> > > +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) > >> > > { > >> > > - if (oa_report_header_64bit(stream)) > >> > > - *(u64 *)report = 0; > >> > > - else > >> > > - *report = 0; > >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > >> > > + > >> > > + oa_report_header_64bit(stream) ? > >> > > + xe_map_wr(stream->oa->xe, map, head, u64, 0) : > >> > > + xe_map_wr(stream->oa->xe, map, head, u32, 0); > >> > > } > >> > > > >> > > -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) > >> > > +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) > >> > > { > >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > >> > > + > >> > > return oa_report_header_64bit(stream) ? > >> > > - *((u64 *)report + 1) : > >> > > - *((u32 *)report + 1); > >> > > + xe_map_rd(stream->oa->xe, map, tail + 2, u64) : > >> > > + xe_map_rd(stream->oa->xe, map, tail + 1, u32); > >> > > >> > (2) I think this should be tail + 8 and tail + 4. Can you please double > >> > check? The u32/u64 is only being used to access the data at the offset > >> > rather than determine how to calculate the actual offset. > >> > >> Hmm, you are right!!! I don't believe I made such a basic mistake :/ And > >> all the xe_oa tests work fine even with this. > >> > >> Anyway thanks for catching this, I'm fixing this in the next rev. > >> > >> > > >> > > } > >> > > > >> > > -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) > >> > > +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) > >> > > { > >> > > - if (oa_report_header_64bit(stream)) > >> > > - *(u64 *)&report[2] = 0; > >> > > - else > >> > > - report[1] = 0; > >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > >> > > + > >> > > + oa_report_header_64bit(stream) ? > >> > > + xe_map_wr(stream->oa->xe, map, head + 2, u64, 0) : > >> > > + xe_map_wr(stream->oa->xe, map, head + 1, u32, 0); > >> > > >> > Same here - comment (2) > >> > > >> > > } > >> > > >> > Same comment (1) for all the above helpers - the name can be generic - > >> > report_offset. I see similar change in other functions below as well. If > >> > you think it is necessary, I suggest trying to move it to a separate patch > >> > with commit message explaining why you want to rename them. That would also > >> > help review this patch easily. > >> > >> So, as we know HW writes to tail and SW reads from head. So the helpers > >> with head as the arg are invoked from the read side only and the helpers > >> with tail as the arg are invoked from the write side only. That was the > >> reason for naming the arg as head/tail, rather than just "offset". I > >> thought it will make the code easier to understand and also maybe prevent > >> mistakes in the future. > > To me, the report_offset (or offset) is clearer. When I read the code I > would remember this helper as returning report_id/timestamp given any > report offset as an argument. The tail/head names makes me wonder if the > caller has to be accounted for as well. OK, understood now. I changed the arg to report_offset in v6. Thanks. > Anyways, please add the above description you have to the commit message > and you can leave it as is. > > Thanks, > Umesh > > >> > >> So anyway, let me ask once again if you think I should change it. Of > >> course, there's no problem changing the arg name in the helpers. > >> > >> I'll also see if it's possible to split the patch some way to make the > >> review a bit easier. > > > > I just sent out v5 in which I have let all this stay as is. Anyway, please > > let me know what you prefer. Thanks. > > > >> > >> Thanks. > >> -- > >> Ashutosh > >> > >> > >> > > >> > > > >> > > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > >> > > @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > >> > > * they were written. If not : (╯°□°）╯︵ ┻━┻ > >> > > */ > >> > > while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { > >> > > - void *report = stream->oa_buffer.vaddr + tail; > >> > > - > >> > > - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) > >> > > + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) > >> > > break; > >> > > > >> > > tail = xe_oa_circ_diff(stream, tail, report_size); > >> > > @@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) > >> > > return HRTIMER_RESTART; > >> > > } > >> > > > >> > > +static unsigned long > >> > > +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) > >> > > +{ > >> > > + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, > >> > > + &stream->oa_buffer.bo->vmap, head, len); > >> > > + return copy_to_user(dst, stream->oa_buffer.bounce, len); > >> > > +} > >> > > + > >> > > >> > Still need to look at the patch below this point. Will send that out next > >> > week... > >> > > >> > Regards, > >> > Umesh > >> > > >> > > static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, > >> > > - size_t count, size_t *offset, const u8 *report) > >> > > + size_t count, size_t *offset, u32 head) > >> > > { > >> > > int report_size = stream->oa_buffer.format->size; > >> > > int report_size_partial; > >> > > - u8 *oa_buf_end; > >> > > > >> > > if ((count - *offset) < report_size) > >> > > return -ENOSPC; > >> > > > >> > > buf += *offset; > >> > > > >> > > - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; > >> > > - report_size_partial = oa_buf_end - report; > >> > > + report_size_partial = stream->oa_buffer.circ_size - head; > >> > > > >> > > if (report_size_partial < report_size) { > >> > > - if (copy_to_user(buf, report, report_size_partial)) > >> > > + if (xe_oa_copy_to_user(stream, buf, head, report_size_partial)) > >> > > return -EFAULT; > >> > > buf += report_size_partial; > >> > > > >> > > - if (copy_to_user(buf, stream->oa_buffer.vaddr, > >> > > - report_size - report_size_partial)) > >> > > + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) > >> > > return -EFAULT; > >> > > - } else if (copy_to_user(buf, report, report_size)) { > >> > > + } else if (xe_oa_copy_to_user(stream, buf, head, report_size)) { > >> > > return -EFAULT; > >> > > } > >> > > > >> > > @@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > >> > > size_t count, size_t *offset) > >> > > { > >> > > int report_size = stream->oa_buffer.format->size; > >> > > - u8 *oa_buf_base = stream->oa_buffer.vaddr; > >> > > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); > >> > > size_t start_offset = *offset; > >> > > unsigned long flags; > >> > > @@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > >> > > > >> > > for (; xe_oa_circ_diff(stream, tail, head); > >> > > head = xe_oa_circ_incr(stream, head, report_size)) { > >> > > - u8 *report = oa_buf_base + head; > >> > > - > >> > > - ret = xe_oa_append_report(stream, buf, count, offset, report); > >> > > + ret = xe_oa_append_report(stream, buf, count, offset, head); > >> > > if (ret) > >> > > break; > >> > > > >> > > if (!(stream->oa_buffer.circ_size % report_size)) { > >> > > /* Clear out report id and timestamp to detect unlanded reports */ > >> > > - oa_report_id_clear(stream, (void *)report); > >> > > - oa_timestamp_clear(stream, (void *)report); > >> > > + oa_report_id_clear(stream, head); > >> > > + oa_timestamp_clear(stream, head); > >> > > } else { > >> > > - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; > >> > > - u32 part = oa_buf_end - report; > >> > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > >> > > + u32 part = stream->oa_buffer.circ_size - head; > >> > > > >> > > /* Zero out the entire report */ > >> > > if (report_size <= part) { > >> > > - memset(report, 0, report_size); > >> > > + xe_map_memset(stream->oa->xe, map, head, 0, report_size); > >> > > } else { > >> > > - memset(report, 0, part); > >> > > - memset(oa_buf_base, 0, report_size - part); > >> > > + xe_map_memset(stream->oa->xe, map, head, 0, part); > >> > > + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); > >> > > } > >> > > } > >> > > } > >> > > @@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) > >> > > spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); > >> > > > >> > > /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ > >> > > - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); > >> > > + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, > >> > > + xe_bo_size(stream->oa_buffer.bo)); > >> > > } > >> > > > >> > > static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) > >> > > @@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) > >> > > static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) > >> > > { > >> > > xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); > >> > > + kfree(stream->oa_buffer.bounce); > >> > > } > >> > > > >> > > static void xe_oa_free_configs(struct xe_oa_stream *stream) > >> > > @@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > >> > > return PTR_ERR(bo); > >> > > > >> > > stream->oa_buffer.bo = bo; > >> > > + > >> > > /* mmap implementation requires OA buffer to be in system memory */ > >> > > xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); > >> > > - stream->oa_buffer.vaddr = bo->vmap.vaddr; > >> > > + > >> > > + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); > >> > > + if (!stream->oa_buffer.bounce) { > >> > > + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); > >> > > + return -ENOMEM; > >> > > + } > >> > > + > >> > > return 0; > >> > > } > >> > > > >> > > diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h > >> > > index b03ffd5134834..99a04b8d586b5 100644 > >> > > --- a/drivers/gpu/drm/xe/xe_oa_types.h > >> > > +++ b/drivers/gpu/drm/xe/xe_oa_types.h > >> > > @@ -162,8 +162,8 @@ struct xe_oa_buffer { > >> > > /** @format: xe_bo backing the OA buffer */ > >> > > struct xe_bo *bo; > >> > > > >> > > - /** @vaddr: mapped vaddr of the OA buffer */ > >> > > - u8 *vaddr; > >> > > + /** @bounce: bounce buffer used with xe_map layer */ > >> > > + void *bounce; > >> > > > >> > > /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ > >> > > spinlock_t ptr_lock; > >> > > -- > >> > > 2.48.1 > >> > > ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v5 0/3] drm/xe/oa: Wa_14026633728 @ 2026-04-25 0:14 Ashutosh Dixit 2026-04-25 0:14 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-25 0:14 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa v5 though v2: Change to Patch 1 (see changelog in Patch 1) Ashutosh Dixit (3): drm/xe/oa: Use xe_map layer drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap drm/xe/oa: Implement Wa_14026633728 drivers/gpu/drm/xe/xe_oa.c | 123 ++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- drivers/gpu/drm/xe/xe_wa_oob.rules | 1 + 3 files changed, 70 insertions(+), 58 deletions(-) -- 2.54.0 ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-25 0:14 [PATCH v5 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit @ 2026-04-25 0:14 ` Ashutosh Dixit 2026-04-27 19:23 ` Umesh Nerlige Ramappa 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-25 0:14 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa OA code should have used xe_map layer to begin with. In CRI, the OA buffer can be located both in system and device memory. For these reasons, move OA code to use the xe_map layer when accessing the OA buffer. v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) v3: To avoid performance impact in v2, revert to v1 but move xe_map_copy_to_user() to xe_map.h v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user v5: Fix offsets in oa_timestamp() and oa_timestamp_clear() (Umesh) Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> --- drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- 2 files changed, 58 insertions(+), 41 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index 6337e671c97ae..3ae30dc609bd7 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) #define oa_report_header_64bit(__s) \ ((__s)->oa_buffer.format->header == HDR_64_BIT) -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) +static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) { - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + return oa_report_header_64bit(stream) ? + xe_map_rd(stream->oa->xe, map, tail, u64) : + xe_map_rd(stream->oa->xe, map, tail, u32); } -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)report = 0; - else - *report = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head, u64, 0) : + xe_map_wr(stream->oa->xe, map, head, u32, 0); } -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + return oa_report_header_64bit(stream) ? - *((u64 *)report + 1) : - *((u32 *)report + 1); + xe_map_rd(stream->oa->xe, map, tail + 8, u64) : + xe_map_rd(stream->oa->xe, map, tail + 4, u32); } -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) { - if (oa_report_header_64bit(stream)) - *(u64 *)&report[2] = 0; - else - report[1] = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, head + 8, u64, 0) : + xe_map_wr(stream->oa->xe, map, head + 4, u32, 0); } static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) * they were written. If not : (╯°□°）╯︵ ┻━┻ */ while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { - void *report = stream->oa_buffer.vaddr + tail; - - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) break; tail = xe_oa_circ_diff(stream, tail, report_size); @@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) return HRTIMER_RESTART; } +static unsigned long +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) +{ + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, + &stream->oa_buffer.bo->vmap, head, len); + return copy_to_user(dst, stream->oa_buffer.bounce, len); +} + static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, - size_t count, size_t *offset, const u8 *report) + size_t count, size_t *offset, u32 head) { int report_size = stream->oa_buffer.format->size; int report_size_partial; - u8 *oa_buf_end; if ((count - *offset) < report_size) return -ENOSPC; buf += *offset; - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - report_size_partial = oa_buf_end - report; + report_size_partial = stream->oa_buffer.circ_size - head; if (report_size_partial < report_size) { - if (copy_to_user(buf, report, report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, head, report_size_partial)) return -EFAULT; buf += report_size_partial; - if (copy_to_user(buf, stream->oa_buffer.vaddr, - report_size - report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) return -EFAULT; - } else if (copy_to_user(buf, report, report_size)) { + } else if (xe_oa_copy_to_user(stream, buf, head, report_size)) { return -EFAULT; } @@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, size_t count, size_t *offset) { int report_size = stream->oa_buffer.format->size; - u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); size_t start_offset = *offset; unsigned long flags; @@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, for (; xe_oa_circ_diff(stream, tail, head); head = xe_oa_circ_incr(stream, head, report_size)) { - u8 *report = oa_buf_base + head; - - ret = xe_oa_append_report(stream, buf, count, offset, report); + ret = xe_oa_append_report(stream, buf, count, offset, head); if (ret) break; if (!(stream->oa_buffer.circ_size % report_size)) { /* Clear out report id and timestamp to detect unlanded reports */ - oa_report_id_clear(stream, (void *)report); - oa_timestamp_clear(stream, (void *)report); + oa_report_id_clear(stream, head); + oa_timestamp_clear(stream, head); } else { - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - u32 part = oa_buf_end - report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + u32 part = stream->oa_buffer.circ_size - head; /* Zero out the entire report */ if (report_size <= part) { - memset(report, 0, report_size); + xe_map_memset(stream->oa->xe, map, head, 0, report_size); } else { - memset(report, 0, part); - memset(oa_buf_base, 0, report_size - part); + xe_map_memset(stream->oa->xe, map, head, 0, part); + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); } } } @@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, + xe_bo_size(stream->oa_buffer.bo)); } static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) @@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) { xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + kfree(stream->oa_buffer.bounce); } static void xe_oa_free_configs(struct xe_oa_stream *stream) @@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) return PTR_ERR(bo); stream->oa_buffer.bo = bo; + /* mmap implementation requires OA buffer to be in system memory */ xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); - stream->oa_buffer.vaddr = bo->vmap.vaddr; + + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); + if (!stream->oa_buffer.bounce) { + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + return -ENOMEM; + } + return 0; } diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h index 8906c3084b5f8..3d9ec8490899c 100644 --- a/drivers/gpu/drm/xe/xe_oa_types.h +++ b/drivers/gpu/drm/xe/xe_oa_types.h @@ -164,8 +164,8 @@ struct xe_oa_buffer { /** @bo: xe_bo backing the OA buffer */ struct xe_bo *bo; - /** @vaddr: mapped vaddr of the OA buffer */ - u8 *vaddr; + /** @bounce: bounce buffer used with xe_map layer */ + void *bounce; /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ spinlock_t ptr_lock; -- 2.54.0 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-25 0:14 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit @ 2026-04-27 19:23 ` Umesh Nerlige Ramappa 0 siblings, 0 replies; 30+ messages in thread From: Umesh Nerlige Ramappa @ 2026-04-27 19:23 UTC (permalink / raw) To: Ashutosh Dixit; +Cc: intel-xe On Fri, Apr 24, 2026 at 05:14:37PM -0700, Ashutosh Dixit wrote: >OA code should have used xe_map layer to begin with. In CRI, the OA buffer >can be located both in system and device memory. For these reasons, move OA >code to use the xe_map layer when accessing the OA buffer. (1) Please also add the description from the thread in v4 on why you are renaming some stuff. > >v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) >v3: To avoid performance impact in v2, revert to v1 but move > xe_map_copy_to_user() to xe_map.h >v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user >v5: Fix offsets in oa_timestamp() and oa_timestamp_clear() (Umesh) > >Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> >--- > drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- > drivers/gpu/drm/xe/xe_oa_types.h | 4 +- > 2 files changed, 58 insertions(+), 41 deletions(-) > >diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c >index 6337e671c97ae..3ae30dc609bd7 100644 >--- a/drivers/gpu/drm/xe/xe_oa.c >+++ b/drivers/gpu/drm/xe/xe_oa.c >@@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) > #define oa_report_header_64bit(__s) \ > ((__s)->oa_buffer.format->header == HDR_64_BIT) > >-static u64 oa_report_id(struct xe_oa_stream *stream, void *report) >+static u64 oa_report_id(struct xe_oa_stream *stream, u32 tail) > { >- return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ return oa_report_header_64bit(stream) ? >+ xe_map_rd(stream->oa->xe, map, tail, u64) : >+ xe_map_rd(stream->oa->xe, map, tail, u32); > } > >-static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) >+static void oa_report_id_clear(struct xe_oa_stream *stream, u32 head) > { >- if (oa_report_header_64bit(stream)) >- *(u64 *)report = 0; >- else >- *report = 0; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ oa_report_header_64bit(stream) ? >+ xe_map_wr(stream->oa->xe, map, head, u64, 0) : >+ xe_map_wr(stream->oa->xe, map, head, u32, 0); > } > >-static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) >+static u64 oa_timestamp(struct xe_oa_stream *stream, u32 tail) > { >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ > return oa_report_header_64bit(stream) ? >- *((u64 *)report + 1) : >- *((u32 *)report + 1); >+ xe_map_rd(stream->oa->xe, map, tail + 8, u64) : >+ xe_map_rd(stream->oa->xe, map, tail + 4, u32); > } > >-static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) >+static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 head) > { >- if (oa_report_header_64bit(stream)) >- *(u64 *)&report[2] = 0; >- else >- report[1] = 0; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ oa_report_header_64bit(stream) ? >+ xe_map_wr(stream->oa->xe, map, head + 8, u64, 0) : >+ xe_map_wr(stream->oa->xe, map, head + 4, u32, 0); > } Well, since this is cleared after the report is copied out, it's difficult to catch it. The only code this will affect is when we try to detect fully landed reports. I don't think existing tests can catch that. Can we plan to implement a test for this at some point? Something like - mmap OA buffer, read the data using read() call and then checked the mmapped location to ensure you zeroed out the timestamp and any other field we are clearing. > > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) >@@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > * they were written. If not : (╯°□°）╯︵ ┻━┻ > */ > while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { >- void *report = stream->oa_buffer.vaddr + tail; >- >- if (oa_report_id(stream, report) || oa_timestamp(stream, report)) >+ if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) > break; > > tail = xe_oa_circ_diff(stream, tail, report_size); >@@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) > return HRTIMER_RESTART; > } > >+static unsigned long >+xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 head, u32 len) >+{ (2) Check and fail if len is greater than report_size. I see callers right now are still within limits, so this should at least assert that the len is <= report_size because your bounce buffer can only hold that much data. With 1 and 2, this is Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com> Thanks, Umesh >+ xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, >+ &stream->oa_buffer.bo->vmap, head, len); >+ return copy_to_user(dst, stream->oa_buffer.bounce, len); >+} >+ > static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, >- size_t count, size_t *offset, const u8 *report) >+ size_t count, size_t *offset, u32 head) > { > int report_size = stream->oa_buffer.format->size; > int report_size_partial; >- u8 *oa_buf_end; > > if ((count - *offset) < report_size) > return -ENOSPC; > > buf += *offset; > >- oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >- report_size_partial = oa_buf_end - report; >+ report_size_partial = stream->oa_buffer.circ_size - head; > > if (report_size_partial < report_size) { >- if (copy_to_user(buf, report, report_size_partial)) >+ if (xe_oa_copy_to_user(stream, buf, head, report_size_partial)) > return -EFAULT; > buf += report_size_partial; > >- if (copy_to_user(buf, stream->oa_buffer.vaddr, >- report_size - report_size_partial)) >+ if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) > return -EFAULT; >- } else if (copy_to_user(buf, report, report_size)) { >+ } else if (xe_oa_copy_to_user(stream, buf, head, report_size)) { > return -EFAULT; > } > >@@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > size_t count, size_t *offset) > { > int report_size = stream->oa_buffer.format->size; >- u8 *oa_buf_base = stream->oa_buffer.vaddr; > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); > size_t start_offset = *offset; > unsigned long flags; >@@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > for (; xe_oa_circ_diff(stream, tail, head); > head = xe_oa_circ_incr(stream, head, report_size)) { >- u8 *report = oa_buf_base + head; >- >- ret = xe_oa_append_report(stream, buf, count, offset, report); >+ ret = xe_oa_append_report(stream, buf, count, offset, head); > if (ret) > break; > > if (!(stream->oa_buffer.circ_size % report_size)) { > /* Clear out report id and timestamp to detect unlanded reports */ >- oa_report_id_clear(stream, (void *)report); >- oa_timestamp_clear(stream, (void *)report); >+ oa_report_id_clear(stream, head); >+ oa_timestamp_clear(stream, head); > } else { >- u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >- u32 part = oa_buf_end - report; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ u32 part = stream->oa_buffer.circ_size - head; > > /* Zero out the entire report */ > if (report_size <= part) { >- memset(report, 0, report_size); >+ xe_map_memset(stream->oa->xe, map, head, 0, report_size); > } else { >- memset(report, 0, part); >- memset(oa_buf_base, 0, report_size - part); >+ xe_map_memset(stream->oa->xe, map, head, 0, part); >+ xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); > } > } > } >@@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) > spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); > > /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ >- memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); >+ xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, >+ xe_bo_size(stream->oa_buffer.bo)); > } > > static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) >@@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) > static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) > { > xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); >+ kfree(stream->oa_buffer.bounce); > } > > static void xe_oa_free_configs(struct xe_oa_stream *stream) >@@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > return PTR_ERR(bo); > > stream->oa_buffer.bo = bo; >+ > /* mmap implementation requires OA buffer to be in system memory */ > xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); >- stream->oa_buffer.vaddr = bo->vmap.vaddr; >+ >+ stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); >+ if (!stream->oa_buffer.bounce) { >+ xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); >+ return -ENOMEM; >+ } >+ > return 0; > } > >diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h >index 8906c3084b5f8..3d9ec8490899c 100644 >--- a/drivers/gpu/drm/xe/xe_oa_types.h >+++ b/drivers/gpu/drm/xe/xe_oa_types.h >@@ -164,8 +164,8 @@ struct xe_oa_buffer { > /** @bo: xe_bo backing the OA buffer */ > struct xe_bo *bo; > >- /** @vaddr: mapped vaddr of the OA buffer */ >- u8 *vaddr; >+ /** @bounce: bounce buffer used with xe_map layer */ >+ void *bounce; > > /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ > spinlock_t ptr_lock; >-- >2.54.0 > ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v6 0/3] drm/xe/oa: Wa_14026633728 @ 2026-04-27 19:02 Ashutosh Dixit 2026-04-27 19:02 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-27 19:02 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa v6 though v2: Change to Patch 1 (see changelog in Patch 1) Ashutosh Dixit (3): drm/xe/oa: Use xe_map layer drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap drm/xe/oa: Implement Wa_14026633728 drivers/gpu/drm/xe/xe_oa.c | 123 ++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- drivers/gpu/drm/xe/xe_wa_oob.rules | 1 + 3 files changed, 70 insertions(+), 58 deletions(-) -- 2.54.0 ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-27 19:02 [PATCH v6 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit @ 2026-04-27 19:02 ` Ashutosh Dixit 2026-04-27 19:34 ` Umesh Nerlige Ramappa 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-27 19:02 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa OA code should have used xe_map layer to begin with. In CRI, the OA buffer can be located both in system and device memory. For these reasons, move OA code to use the xe_map layer when accessing the OA buffer. v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) v3: To avoid performance impact in v2, revert to v1 but move xe_map_copy_to_user() to xe_map.h v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user v5: Fix offsets in oa_timestamp() and oa_timestamp_clear() (Umesh) v6: Rename head/tail in helper args to report_offset (Umesh) Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> --- drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- 2 files changed, 58 insertions(+), 41 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index 6337e671c97ae..1fff0e8e9e78e 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) #define oa_report_header_64bit(__s) \ ((__s)->oa_buffer.format->header == HDR_64_BIT) -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) +static u64 oa_report_id(struct xe_oa_stream *stream, u32 report_offset) { - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + return oa_report_header_64bit(stream) ? + xe_map_rd(stream->oa->xe, map, report_offset, u64) : + xe_map_rd(stream->oa->xe, map, report_offset, u32); } -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 report_offset) { - if (oa_report_header_64bit(stream)) - *(u64 *)report = 0; - else - *report = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, report_offset, u64, 0) : + xe_map_wr(stream->oa->xe, map, report_offset, u32, 0); } -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 report_offset) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + return oa_report_header_64bit(stream) ? - *((u64 *)report + 1) : - *((u32 *)report + 1); + xe_map_rd(stream->oa->xe, map, report_offset + 8, u64) : + xe_map_rd(stream->oa->xe, map, report_offset + 4, u32); } -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 report_offset) { - if (oa_report_header_64bit(stream)) - *(u64 *)&report[2] = 0; - else - report[1] = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, report_offset + 8, u64, 0) : + xe_map_wr(stream->oa->xe, map, report_offset + 4, u32, 0); } static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) * they were written. If not : (╯°□°）╯︵ ┻━┻ */ while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { - void *report = stream->oa_buffer.vaddr + tail; - - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) break; tail = xe_oa_circ_diff(stream, tail, report_size); @@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) return HRTIMER_RESTART; } +static unsigned long +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 report_offset, u32 len) +{ + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, + &stream->oa_buffer.bo->vmap, report_offset, len); + return copy_to_user(dst, stream->oa_buffer.bounce, len); +} + static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, - size_t count, size_t *offset, const u8 *report) + size_t count, size_t *offset, u32 report_offset) { int report_size = stream->oa_buffer.format->size; int report_size_partial; - u8 *oa_buf_end; if ((count - *offset) < report_size) return -ENOSPC; buf += *offset; - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - report_size_partial = oa_buf_end - report; + report_size_partial = stream->oa_buffer.circ_size - report_offset; if (report_size_partial < report_size) { - if (copy_to_user(buf, report, report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, report_offset, report_size_partial)) return -EFAULT; buf += report_size_partial; - if (copy_to_user(buf, stream->oa_buffer.vaddr, - report_size - report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) return -EFAULT; - } else if (copy_to_user(buf, report, report_size)) { + } else if (xe_oa_copy_to_user(stream, buf, report_offset, report_size)) { return -EFAULT; } @@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, size_t count, size_t *offset) { int report_size = stream->oa_buffer.format->size; - u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); size_t start_offset = *offset; unsigned long flags; @@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, for (; xe_oa_circ_diff(stream, tail, head); head = xe_oa_circ_incr(stream, head, report_size)) { - u8 *report = oa_buf_base + head; - - ret = xe_oa_append_report(stream, buf, count, offset, report); + ret = xe_oa_append_report(stream, buf, count, offset, head); if (ret) break; if (!(stream->oa_buffer.circ_size % report_size)) { /* Clear out report id and timestamp to detect unlanded reports */ - oa_report_id_clear(stream, (void *)report); - oa_timestamp_clear(stream, (void *)report); + oa_report_id_clear(stream, head); + oa_timestamp_clear(stream, head); } else { - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - u32 part = oa_buf_end - report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + u32 part = stream->oa_buffer.circ_size - head; /* Zero out the entire report */ if (report_size <= part) { - memset(report, 0, report_size); + xe_map_memset(stream->oa->xe, map, head, 0, report_size); } else { - memset(report, 0, part); - memset(oa_buf_base, 0, report_size - part); + xe_map_memset(stream->oa->xe, map, head, 0, part); + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); } } } @@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, + xe_bo_size(stream->oa_buffer.bo)); } static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) @@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) { xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + kfree(stream->oa_buffer.bounce); } static void xe_oa_free_configs(struct xe_oa_stream *stream) @@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) return PTR_ERR(bo); stream->oa_buffer.bo = bo; + /* mmap implementation requires OA buffer to be in system memory */ xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); - stream->oa_buffer.vaddr = bo->vmap.vaddr; + + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); + if (!stream->oa_buffer.bounce) { + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + return -ENOMEM; + } + return 0; } diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h index 8906c3084b5f8..3d9ec8490899c 100644 --- a/drivers/gpu/drm/xe/xe_oa_types.h +++ b/drivers/gpu/drm/xe/xe_oa_types.h @@ -164,8 +164,8 @@ struct xe_oa_buffer { /** @bo: xe_bo backing the OA buffer */ struct xe_bo *bo; - /** @vaddr: mapped vaddr of the OA buffer */ - u8 *vaddr; + /** @bounce: bounce buffer used with xe_map layer */ + void *bounce; /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ spinlock_t ptr_lock; -- 2.54.0 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-27 19:02 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit @ 2026-04-27 19:34 ` Umesh Nerlige Ramappa 2026-04-27 20:30 ` Dixit, Ashutosh 0 siblings, 1 reply; 30+ messages in thread From: Umesh Nerlige Ramappa @ 2026-04-27 19:34 UTC (permalink / raw) To: Ashutosh Dixit; +Cc: intel-xe On Mon, Apr 27, 2026 at 12:02:21PM -0700, Ashutosh Dixit wrote: >OA code should have used xe_map layer to begin with. In CRI, the OA buffer >can be located both in system and device memory. For these reasons, move OA >code to use the xe_map layer when accessing the OA buffer. > >v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) >v3: To avoid performance impact in v2, revert to v1 but move > xe_map_copy_to_user() to xe_map.h >v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user >v5: Fix offsets in oa_timestamp() and oa_timestamp_clear() (Umesh) >v6: Rename head/tail in helper args to report_offset (Umesh) Thanks for renaming. My comments are in the v5 version of the series. With an assert/check in the xe_oa_copy_to_user(), this should be Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com> Umesh > >Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> >--- > drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- > drivers/gpu/drm/xe/xe_oa_types.h | 4 +- > 2 files changed, 58 insertions(+), 41 deletions(-) > >diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c >index 6337e671c97ae..1fff0e8e9e78e 100644 >--- a/drivers/gpu/drm/xe/xe_oa.c >+++ b/drivers/gpu/drm/xe/xe_oa.c >@@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) > #define oa_report_header_64bit(__s) \ > ((__s)->oa_buffer.format->header == HDR_64_BIT) > >-static u64 oa_report_id(struct xe_oa_stream *stream, void *report) >+static u64 oa_report_id(struct xe_oa_stream *stream, u32 report_offset) > { >- return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ return oa_report_header_64bit(stream) ? >+ xe_map_rd(stream->oa->xe, map, report_offset, u64) : >+ xe_map_rd(stream->oa->xe, map, report_offset, u32); > } > >-static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) >+static void oa_report_id_clear(struct xe_oa_stream *stream, u32 report_offset) > { >- if (oa_report_header_64bit(stream)) >- *(u64 *)report = 0; >- else >- *report = 0; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ oa_report_header_64bit(stream) ? >+ xe_map_wr(stream->oa->xe, map, report_offset, u64, 0) : >+ xe_map_wr(stream->oa->xe, map, report_offset, u32, 0); > } > >-static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) >+static u64 oa_timestamp(struct xe_oa_stream *stream, u32 report_offset) > { >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ > return oa_report_header_64bit(stream) ? >- *((u64 *)report + 1) : >- *((u32 *)report + 1); >+ xe_map_rd(stream->oa->xe, map, report_offset + 8, u64) : >+ xe_map_rd(stream->oa->xe, map, report_offset + 4, u32); > } > >-static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) >+static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 report_offset) > { >- if (oa_report_header_64bit(stream)) >- *(u64 *)&report[2] = 0; >- else >- report[1] = 0; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ >+ oa_report_header_64bit(stream) ? >+ xe_map_wr(stream->oa->xe, map, report_offset + 8, u64, 0) : >+ xe_map_wr(stream->oa->xe, map, report_offset + 4, u32, 0); > } > > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) >@@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > * they were written. If not : (╯°□°）╯︵ ┻━┻ > */ > while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { >- void *report = stream->oa_buffer.vaddr + tail; >- >- if (oa_report_id(stream, report) || oa_timestamp(stream, report)) >+ if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) > break; > > tail = xe_oa_circ_diff(stream, tail, report_size); >@@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) > return HRTIMER_RESTART; > } > >+static unsigned long >+xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 report_offset, u32 len) >+{ >+ xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, >+ &stream->oa_buffer.bo->vmap, report_offset, len); >+ return copy_to_user(dst, stream->oa_buffer.bounce, len); >+} >+ > static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, >- size_t count, size_t *offset, const u8 *report) >+ size_t count, size_t *offset, u32 report_offset) > { > int report_size = stream->oa_buffer.format->size; > int report_size_partial; >- u8 *oa_buf_end; > > if ((count - *offset) < report_size) > return -ENOSPC; > > buf += *offset; > >- oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >- report_size_partial = oa_buf_end - report; >+ report_size_partial = stream->oa_buffer.circ_size - report_offset; > > if (report_size_partial < report_size) { >- if (copy_to_user(buf, report, report_size_partial)) >+ if (xe_oa_copy_to_user(stream, buf, report_offset, report_size_partial)) > return -EFAULT; > buf += report_size_partial; > >- if (copy_to_user(buf, stream->oa_buffer.vaddr, >- report_size - report_size_partial)) >+ if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) > return -EFAULT; >- } else if (copy_to_user(buf, report, report_size)) { >+ } else if (xe_oa_copy_to_user(stream, buf, report_offset, report_size)) { > return -EFAULT; > } > >@@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > size_t count, size_t *offset) > { > int report_size = stream->oa_buffer.format->size; >- u8 *oa_buf_base = stream->oa_buffer.vaddr; > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); > size_t start_offset = *offset; > unsigned long flags; >@@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > for (; xe_oa_circ_diff(stream, tail, head); > head = xe_oa_circ_incr(stream, head, report_size)) { >- u8 *report = oa_buf_base + head; >- >- ret = xe_oa_append_report(stream, buf, count, offset, report); >+ ret = xe_oa_append_report(stream, buf, count, offset, head); > if (ret) > break; > > if (!(stream->oa_buffer.circ_size % report_size)) { > /* Clear out report id and timestamp to detect unlanded reports */ >- oa_report_id_clear(stream, (void *)report); >- oa_timestamp_clear(stream, (void *)report); >+ oa_report_id_clear(stream, head); >+ oa_timestamp_clear(stream, head); > } else { >- u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; >- u32 part = oa_buf_end - report; >+ struct iosys_map *map = &stream->oa_buffer.bo->vmap; >+ u32 part = stream->oa_buffer.circ_size - head; > > /* Zero out the entire report */ > if (report_size <= part) { >- memset(report, 0, report_size); >+ xe_map_memset(stream->oa->xe, map, head, 0, report_size); > } else { >- memset(report, 0, part); >- memset(oa_buf_base, 0, report_size - part); >+ xe_map_memset(stream->oa->xe, map, head, 0, part); >+ xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); > } > } > } >@@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) > spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); > > /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ >- memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); >+ xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, >+ xe_bo_size(stream->oa_buffer.bo)); > } > > static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) >@@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) > static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) > { > xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); >+ kfree(stream->oa_buffer.bounce); > } > > static void xe_oa_free_configs(struct xe_oa_stream *stream) >@@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > return PTR_ERR(bo); > > stream->oa_buffer.bo = bo; >+ > /* mmap implementation requires OA buffer to be in system memory */ > xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); >- stream->oa_buffer.vaddr = bo->vmap.vaddr; >+ >+ stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); >+ if (!stream->oa_buffer.bounce) { >+ xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); >+ return -ENOMEM; >+ } >+ > return 0; > } > >diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h >index 8906c3084b5f8..3d9ec8490899c 100644 >--- a/drivers/gpu/drm/xe/xe_oa_types.h >+++ b/drivers/gpu/drm/xe/xe_oa_types.h >@@ -164,8 +164,8 @@ struct xe_oa_buffer { > /** @bo: xe_bo backing the OA buffer */ > struct xe_bo *bo; > >- /** @vaddr: mapped vaddr of the OA buffer */ >- u8 *vaddr; >+ /** @bounce: bounce buffer used with xe_map layer */ >+ void *bounce; > > /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ > spinlock_t ptr_lock; >-- >2.54.0 > ^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-27 19:34 ` Umesh Nerlige Ramappa @ 2026-04-27 20:30 ` Dixit, Ashutosh 0 siblings, 0 replies; 30+ messages in thread From: Dixit, Ashutosh @ 2026-04-27 20:30 UTC (permalink / raw) To: Umesh Nerlige Ramappa; +Cc: intel-xe On Mon, 27 Apr 2026 12:34:24 -0700, Umesh Nerlige Ramappa wrote: > > On Mon, Apr 27, 2026 at 12:02:21PM -0700, Ashutosh Dixit wrote: > > OA code should have used xe_map layer to begin with. In CRI, the OA buffer > > can be located both in system and device memory. For these reasons, move OA > > code to use the xe_map layer when accessing the OA buffer. > > > > v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) > > v3: To avoid performance impact in v2, revert to v1 but move > > xe_map_copy_to_user() to xe_map.h > > v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user > > v5: Fix offsets in oa_timestamp() and oa_timestamp_clear() (Umesh) > > v6: Rename head/tail in helper args to report_offset (Umesh) > > Thanks for renaming. My comments are in the v5 version of the series. > > With an assert/check in the xe_oa_copy_to_user(), this should be Add in v7. The only remaining comment is the one about the test, will make a note of it. > > Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com> Thanks. -- Ashutosh > > > > > Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> > > --- > > drivers/gpu/drm/xe/xe_oa.c | 95 +++++++++++++++++++------------- > > drivers/gpu/drm/xe/xe_oa_types.h | 4 +- > > 2 files changed, 58 insertions(+), 41 deletions(-) > > > > diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c > > index 6337e671c97ae..1fff0e8e9e78e 100644 > > --- a/drivers/gpu/drm/xe/xe_oa.c > > +++ b/drivers/gpu/drm/xe/xe_oa.c > > @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) > > #define oa_report_header_64bit(__s) \ > > ((__s)->oa_buffer.format->header == HDR_64_BIT) > > > > -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) > > +static u64 oa_report_id(struct xe_oa_stream *stream, u32 report_offset) > > { > > - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + > > + return oa_report_header_64bit(stream) ? > > + xe_map_rd(stream->oa->xe, map, report_offset, u64) : > > + xe_map_rd(stream->oa->xe, map, report_offset, u32); > > } > > > > -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) > > +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 report_offset) > > { > > - if (oa_report_header_64bit(stream)) > > - *(u64 *)report = 0; > > - else > > - *report = 0; > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + > > + oa_report_header_64bit(stream) ? > > + xe_map_wr(stream->oa->xe, map, report_offset, u64, 0) : > > + xe_map_wr(stream->oa->xe, map, report_offset, u32, 0); > > } > > > > -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) > > +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 report_offset) > > { > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + > > return oa_report_header_64bit(stream) ? > > - *((u64 *)report + 1) : > > - *((u32 *)report + 1); > > + xe_map_rd(stream->oa->xe, map, report_offset + 8, u64) : > > + xe_map_rd(stream->oa->xe, map, report_offset + 4, u32); > > } > > > > -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) > > +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 report_offset) > > { > > - if (oa_report_header_64bit(stream)) > > - *(u64 *)&report[2] = 0; > > - else > > - report[1] = 0; > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + > > + oa_report_header_64bit(stream) ? > > + xe_map_wr(stream->oa->xe, map, report_offset + 8, u64, 0) : > > + xe_map_wr(stream->oa->xe, map, report_offset + 4, u32, 0); > > } > > > > static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > > @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) > > * they were written. If not : (╯°□°）╯︵ ┻━┻ > > */ > > while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { > > - void *report = stream->oa_buffer.vaddr + tail; > > - > > - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) > > + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) > > break; > > > > tail = xe_oa_circ_diff(stream, tail, report_size); > > @@ -311,30 +317,35 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) > > return HRTIMER_RESTART; > > } > > > > +static unsigned long > > +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 report_offset, u32 len) > > +{ > > + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, > > + &stream->oa_buffer.bo->vmap, report_offset, len); > > + return copy_to_user(dst, stream->oa_buffer.bounce, len); > > +} > > + > > static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, > > - size_t count, size_t *offset, const u8 *report) > > + size_t count, size_t *offset, u32 report_offset) > > { > > int report_size = stream->oa_buffer.format->size; > > int report_size_partial; > > - u8 *oa_buf_end; > > > > if ((count - *offset) < report_size) > > return -ENOSPC; > > > > buf += *offset; > > > > - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; > > - report_size_partial = oa_buf_end - report; > > + report_size_partial = stream->oa_buffer.circ_size - report_offset; > > > > if (report_size_partial < report_size) { > > - if (copy_to_user(buf, report, report_size_partial)) > > + if (xe_oa_copy_to_user(stream, buf, report_offset, report_size_partial)) > > return -EFAULT; > > buf += report_size_partial; > > > > - if (copy_to_user(buf, stream->oa_buffer.vaddr, > > - report_size - report_size_partial)) > > + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) > > return -EFAULT; > > - } else if (copy_to_user(buf, report, report_size)) { > > + } else if (xe_oa_copy_to_user(stream, buf, report_offset, report_size)) { > > return -EFAULT; > > } > > > > @@ -347,7 +358,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > size_t count, size_t *offset) > > { > > int report_size = stream->oa_buffer.format->size; > > - u8 *oa_buf_base = stream->oa_buffer.vaddr; > > u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); > > size_t start_offset = *offset; > > unsigned long flags; > > @@ -364,26 +374,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, > > > > for (; xe_oa_circ_diff(stream, tail, head); > > head = xe_oa_circ_incr(stream, head, report_size)) { > > - u8 *report = oa_buf_base + head; > > - > > - ret = xe_oa_append_report(stream, buf, count, offset, report); > > + ret = xe_oa_append_report(stream, buf, count, offset, head); > > if (ret) > > break; > > > > if (!(stream->oa_buffer.circ_size % report_size)) { > > /* Clear out report id and timestamp to detect unlanded reports */ > > - oa_report_id_clear(stream, (void *)report); > > - oa_timestamp_clear(stream, (void *)report); > > + oa_report_id_clear(stream, head); > > + oa_timestamp_clear(stream, head); > > } else { > > - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; > > - u32 part = oa_buf_end - report; > > + struct iosys_map *map = &stream->oa_buffer.bo->vmap; > > + u32 part = stream->oa_buffer.circ_size - head; > > > > /* Zero out the entire report */ > > if (report_size <= part) { > > - memset(report, 0, report_size); > > + xe_map_memset(stream->oa->xe, map, head, 0, report_size); > > } else { > > - memset(report, 0, part); > > - memset(oa_buf_base, 0, report_size - part); > > + xe_map_memset(stream->oa->xe, map, head, 0, part); > > + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); > > } > > } > > } > > @@ -436,7 +444,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) > > spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); > > > > /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ > > - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); > > + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, > > + xe_bo_size(stream->oa_buffer.bo)); > > } > > > > static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) > > @@ -699,6 +708,7 @@ static int num_lri_dwords(int num_regs) > > static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) > > { > > xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); > > + kfree(stream->oa_buffer.bounce); > > } > > > > static void xe_oa_free_configs(struct xe_oa_stream *stream) > > @@ -889,9 +899,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) > > return PTR_ERR(bo); > > > > stream->oa_buffer.bo = bo; > > + > > /* mmap implementation requires OA buffer to be in system memory */ > > xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); > > - stream->oa_buffer.vaddr = bo->vmap.vaddr; > > + > > + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); > > + if (!stream->oa_buffer.bounce) { > > + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); > > + return -ENOMEM; > > + } > > + > > return 0; > > } > > > > diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h > > index 8906c3084b5f8..3d9ec8490899c 100644 > > --- a/drivers/gpu/drm/xe/xe_oa_types.h > > +++ b/drivers/gpu/drm/xe/xe_oa_types.h > > @@ -164,8 +164,8 @@ struct xe_oa_buffer { > > /** @bo: xe_bo backing the OA buffer */ > > struct xe_bo *bo; > > > > - /** @vaddr: mapped vaddr of the OA buffer */ > > - u8 *vaddr; > > + /** @bounce: bounce buffer used with xe_map layer */ > > + void *bounce; > > > > /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ > > spinlock_t ptr_lock; > > -- > > 2.54.0 > > ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v7 0/3] drm/xe/oa: Wa_14026633728 @ 2026-04-27 20:26 Ashutosh Dixit 2026-04-27 20:26 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-27 20:26 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa v7 though v2: Change to Patch 1 (see changelog in Patch 1) Ashutosh Dixit (3): drm/xe/oa: Use xe_map layer drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap drm/xe/oa: Implement Wa_14026633728 drivers/gpu/drm/xe/xe_oa.c | 125 ++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- drivers/gpu/drm/xe/xe_wa_oob.rules | 1 + 3 files changed, 72 insertions(+), 58 deletions(-) -- 2.54.0 ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-27 20:26 [PATCH v7 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit @ 2026-04-27 20:26 ` Ashutosh Dixit 0 siblings, 0 replies; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-27 20:26 UTC (permalink / raw) To: intel-xe; +Cc: Umesh Nerlige Ramappa OA code should have used xe_map layer to begin with. In CRI, the OA buffer can be located both in system and device memory. For these reasons, move OA code to use the xe_map layer when accessing the OA buffer. v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) v3: To avoid performance impact in v2, revert to v1 but move xe_map_copy_to_user() to xe_map.h v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user v5: Fix offsets in oa_timestamp() and oa_timestamp_clear() (Umesh) v6: Rename head/tail in helper args to report_offset (Umesh) v7: Add xe_assert in xe_oa_copy_to_user (Umesh) Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com> --- drivers/gpu/drm/xe/xe_oa.c | 97 +++++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- 2 files changed, 60 insertions(+), 41 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index 6337e671c97ae..6ded3c2fce34e 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) #define oa_report_header_64bit(__s) \ ((__s)->oa_buffer.format->header == HDR_64_BIT) -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) +static u64 oa_report_id(struct xe_oa_stream *stream, u32 report_offset) { - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + return oa_report_header_64bit(stream) ? + xe_map_rd(stream->oa->xe, map, report_offset, u64) : + xe_map_rd(stream->oa->xe, map, report_offset, u32); } -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 report_offset) { - if (oa_report_header_64bit(stream)) - *(u64 *)report = 0; - else - *report = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, report_offset, u64, 0) : + xe_map_wr(stream->oa->xe, map, report_offset, u32, 0); } -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 report_offset) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + return oa_report_header_64bit(stream) ? - *((u64 *)report + 1) : - *((u32 *)report + 1); + xe_map_rd(stream->oa->xe, map, report_offset + 8, u64) : + xe_map_rd(stream->oa->xe, map, report_offset + 4, u32); } -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 report_offset) { - if (oa_report_header_64bit(stream)) - *(u64 *)&report[2] = 0; - else - report[1] = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, report_offset + 8, u64, 0) : + xe_map_wr(stream->oa->xe, map, report_offset + 4, u32, 0); } static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) * they were written. If not : (╯°□°）╯︵ ┻━┻ */ while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { - void *report = stream->oa_buffer.vaddr + tail; - - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) break; tail = xe_oa_circ_diff(stream, tail, report_size); @@ -311,30 +317,37 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) return HRTIMER_RESTART; } +static unsigned long +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 report_offset, u32 len) +{ + xe_assert(stream->oa->xe, len <= stream->oa_buffer.format->size); + + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, + &stream->oa_buffer.bo->vmap, report_offset, len); + return copy_to_user(dst, stream->oa_buffer.bounce, len); +} + static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, - size_t count, size_t *offset, const u8 *report) + size_t count, size_t *offset, u32 report_offset) { int report_size = stream->oa_buffer.format->size; int report_size_partial; - u8 *oa_buf_end; if ((count - *offset) < report_size) return -ENOSPC; buf += *offset; - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - report_size_partial = oa_buf_end - report; + report_size_partial = stream->oa_buffer.circ_size - report_offset; if (report_size_partial < report_size) { - if (copy_to_user(buf, report, report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, report_offset, report_size_partial)) return -EFAULT; buf += report_size_partial; - if (copy_to_user(buf, stream->oa_buffer.vaddr, - report_size - report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) return -EFAULT; - } else if (copy_to_user(buf, report, report_size)) { + } else if (xe_oa_copy_to_user(stream, buf, report_offset, report_size)) { return -EFAULT; } @@ -347,7 +360,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, size_t count, size_t *offset) { int report_size = stream->oa_buffer.format->size; - u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); size_t start_offset = *offset; unsigned long flags; @@ -364,26 +376,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, for (; xe_oa_circ_diff(stream, tail, head); head = xe_oa_circ_incr(stream, head, report_size)) { - u8 *report = oa_buf_base + head; - - ret = xe_oa_append_report(stream, buf, count, offset, report); + ret = xe_oa_append_report(stream, buf, count, offset, head); if (ret) break; if (!(stream->oa_buffer.circ_size % report_size)) { /* Clear out report id and timestamp to detect unlanded reports */ - oa_report_id_clear(stream, (void *)report); - oa_timestamp_clear(stream, (void *)report); + oa_report_id_clear(stream, head); + oa_timestamp_clear(stream, head); } else { - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - u32 part = oa_buf_end - report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + u32 part = stream->oa_buffer.circ_size - head; /* Zero out the entire report */ if (report_size <= part) { - memset(report, 0, report_size); + xe_map_memset(stream->oa->xe, map, head, 0, report_size); } else { - memset(report, 0, part); - memset(oa_buf_base, 0, report_size - part); + xe_map_memset(stream->oa->xe, map, head, 0, part); + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); } } } @@ -436,7 +446,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, + xe_bo_size(stream->oa_buffer.bo)); } static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) @@ -699,6 +710,7 @@ static int num_lri_dwords(int num_regs) static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) { xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + kfree(stream->oa_buffer.bounce); } static void xe_oa_free_configs(struct xe_oa_stream *stream) @@ -889,9 +901,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) return PTR_ERR(bo); stream->oa_buffer.bo = bo; + /* mmap implementation requires OA buffer to be in system memory */ xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); - stream->oa_buffer.vaddr = bo->vmap.vaddr; + + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); + if (!stream->oa_buffer.bounce) { + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + return -ENOMEM; + } + return 0; } diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h index 8906c3084b5f8..3d9ec8490899c 100644 --- a/drivers/gpu/drm/xe/xe_oa_types.h +++ b/drivers/gpu/drm/xe/xe_oa_types.h @@ -164,8 +164,8 @@ struct xe_oa_buffer { /** @bo: xe_bo backing the OA buffer */ struct xe_bo *bo; - /** @vaddr: mapped vaddr of the OA buffer */ - u8 *vaddr; + /** @bounce: bounce buffer used with xe_map layer */ + void *bounce; /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ spinlock_t ptr_lock; -- 2.54.0 ^ permalink raw reply related [flat|nested] 30+ messages in thread
* [PATCH v8 0/3] drm/xe/oa: Wa_14026633728 @ 2026-04-27 22:11 Ashutosh Dixit 2026-04-27 22:11 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 0 siblings, 1 reply; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-27 22:11 UTC (permalink / raw) To: intel-xe v7 though v2: Change to Patch 1 (see changelog in Patch 1) v8: Change to Patch 3 (see changelog in Patch 3) Ashutosh Dixit (3): drm/xe/oa: Use xe_map layer drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap drm/xe/oa: Implement Wa_14026633728 drivers/gpu/drm/xe/xe_device_wa_oob.rules | 1 + drivers/gpu/drm/xe/xe_oa.c | 126 ++++++++++++---------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- 3 files changed, 73 insertions(+), 58 deletions(-) -- 2.54.0 ^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 1/3] drm/xe/oa: Use xe_map layer 2026-04-27 22:11 [PATCH v8 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit @ 2026-04-27 22:11 ` Ashutosh Dixit 0 siblings, 0 replies; 30+ messages in thread From: Ashutosh Dixit @ 2026-04-27 22:11 UTC (permalink / raw) To: intel-xe OA code should have used xe_map layer to begin with. In CRI, the OA buffer can be located both in system and device memory. For these reasons, move OA code to use the xe_map layer when accessing the OA buffer. v2: Use xe_map layer and put_user in xe_oa_copy_to_user (Umesh) v3: To avoid performance impact in v2, revert to v1 but move xe_map_copy_to_user() to xe_map.h v4: Use bounce buffer and copy_to_user in xe_oa_copy_to_user v5: Fix offsets in oa_timestamp() and oa_timestamp_clear() (Umesh) v6: Rename head/tail in helper args to report_offset (Umesh) v7: Add xe_assert in xe_oa_copy_to_user (Umesh) Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com> --- drivers/gpu/drm/xe/xe_oa.c | 97 +++++++++++++++++++------------- drivers/gpu/drm/xe/xe_oa_types.h | 4 +- 2 files changed, 60 insertions(+), 41 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c index 6337e671c97ae..6ded3c2fce34e 100644 --- a/drivers/gpu/drm/xe/xe_oa.c +++ b/drivers/gpu/drm/xe/xe_oa.c @@ -213,32 +213,40 @@ static u32 xe_oa_hw_tail_read(struct xe_oa_stream *stream) #define oa_report_header_64bit(__s) \ ((__s)->oa_buffer.format->header == HDR_64_BIT) -static u64 oa_report_id(struct xe_oa_stream *stream, void *report) +static u64 oa_report_id(struct xe_oa_stream *stream, u32 report_offset) { - return oa_report_header_64bit(stream) ? *(u64 *)report : *(u32 *)report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + return oa_report_header_64bit(stream) ? + xe_map_rd(stream->oa->xe, map, report_offset, u64) : + xe_map_rd(stream->oa->xe, map, report_offset, u32); } -static void oa_report_id_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_report_id_clear(struct xe_oa_stream *stream, u32 report_offset) { - if (oa_report_header_64bit(stream)) - *(u64 *)report = 0; - else - *report = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, report_offset, u64, 0) : + xe_map_wr(stream->oa->xe, map, report_offset, u32, 0); } -static u64 oa_timestamp(struct xe_oa_stream *stream, void *report) +static u64 oa_timestamp(struct xe_oa_stream *stream, u32 report_offset) { + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + return oa_report_header_64bit(stream) ? - *((u64 *)report + 1) : - *((u32 *)report + 1); + xe_map_rd(stream->oa->xe, map, report_offset + 8, u64) : + xe_map_rd(stream->oa->xe, map, report_offset + 4, u32); } -static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 *report) +static void oa_timestamp_clear(struct xe_oa_stream *stream, u32 report_offset) { - if (oa_report_header_64bit(stream)) - *(u64 *)&report[2] = 0; - else - report[1] = 0; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + + oa_report_header_64bit(stream) ? + xe_map_wr(stream->oa->xe, map, report_offset + 8, u64, 0) : + xe_map_wr(stream->oa->xe, map, report_offset + 4, u32, 0); } static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) @@ -275,9 +283,7 @@ static bool xe_oa_buffer_check_unlocked(struct xe_oa_stream *stream) * they were written. If not : (╯°□°）╯︵ ┻━┻ */ while (xe_oa_circ_diff(stream, tail, stream->oa_buffer.tail) >= report_size) { - void *report = stream->oa_buffer.vaddr + tail; - - if (oa_report_id(stream, report) || oa_timestamp(stream, report)) + if (oa_report_id(stream, tail) || oa_timestamp(stream, tail)) break; tail = xe_oa_circ_diff(stream, tail, report_size); @@ -311,30 +317,37 @@ static enum hrtimer_restart xe_oa_poll_check_timer_cb(struct hrtimer *hrtimer) return HRTIMER_RESTART; } +static unsigned long +xe_oa_copy_to_user(struct xe_oa_stream *stream, void __user *dst, u32 report_offset, u32 len) +{ + xe_assert(stream->oa->xe, len <= stream->oa_buffer.format->size); + + xe_map_memcpy_from(stream->oa->xe, stream->oa_buffer.bounce, + &stream->oa_buffer.bo->vmap, report_offset, len); + return copy_to_user(dst, stream->oa_buffer.bounce, len); +} + static int xe_oa_append_report(struct xe_oa_stream *stream, char __user *buf, - size_t count, size_t *offset, const u8 *report) + size_t count, size_t *offset, u32 report_offset) { int report_size = stream->oa_buffer.format->size; int report_size_partial; - u8 *oa_buf_end; if ((count - *offset) < report_size) return -ENOSPC; buf += *offset; - oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - report_size_partial = oa_buf_end - report; + report_size_partial = stream->oa_buffer.circ_size - report_offset; if (report_size_partial < report_size) { - if (copy_to_user(buf, report, report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, report_offset, report_size_partial)) return -EFAULT; buf += report_size_partial; - if (copy_to_user(buf, stream->oa_buffer.vaddr, - report_size - report_size_partial)) + if (xe_oa_copy_to_user(stream, buf, 0, report_size - report_size_partial)) return -EFAULT; - } else if (copy_to_user(buf, report, report_size)) { + } else if (xe_oa_copy_to_user(stream, buf, report_offset, report_size)) { return -EFAULT; } @@ -347,7 +360,6 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, size_t count, size_t *offset) { int report_size = stream->oa_buffer.format->size; - u8 *oa_buf_base = stream->oa_buffer.vaddr; u32 gtt_offset = xe_bo_ggtt_addr(stream->oa_buffer.bo); size_t start_offset = *offset; unsigned long flags; @@ -364,26 +376,24 @@ static int xe_oa_append_reports(struct xe_oa_stream *stream, char __user *buf, for (; xe_oa_circ_diff(stream, tail, head); head = xe_oa_circ_incr(stream, head, report_size)) { - u8 *report = oa_buf_base + head; - - ret = xe_oa_append_report(stream, buf, count, offset, report); + ret = xe_oa_append_report(stream, buf, count, offset, head); if (ret) break; if (!(stream->oa_buffer.circ_size % report_size)) { /* Clear out report id and timestamp to detect unlanded reports */ - oa_report_id_clear(stream, (void *)report); - oa_timestamp_clear(stream, (void *)report); + oa_report_id_clear(stream, head); + oa_timestamp_clear(stream, head); } else { - u8 *oa_buf_end = stream->oa_buffer.vaddr + stream->oa_buffer.circ_size; - u32 part = oa_buf_end - report; + struct iosys_map *map = &stream->oa_buffer.bo->vmap; + u32 part = stream->oa_buffer.circ_size - head; /* Zero out the entire report */ if (report_size <= part) { - memset(report, 0, report_size); + xe_map_memset(stream->oa->xe, map, head, 0, report_size); } else { - memset(report, 0, part); - memset(oa_buf_base, 0, report_size - part); + xe_map_memset(stream->oa->xe, map, head, 0, part); + xe_map_memset(stream->oa->xe, map, 0, 0, report_size - part); } } } @@ -436,7 +446,8 @@ static void xe_oa_init_oa_buffer(struct xe_oa_stream *stream) spin_unlock_irqrestore(&stream->oa_buffer.ptr_lock, flags); /* Zero out the OA buffer since we rely on zero report id and timestamp fields */ - memset(stream->oa_buffer.vaddr, 0, xe_bo_size(stream->oa_buffer.bo)); + xe_map_memset(stream->oa->xe, &stream->oa_buffer.bo->vmap, 0, 0, + xe_bo_size(stream->oa_buffer.bo)); } static u32 __format_to_oactrl(const struct xe_oa_format *format, int counter_sel_mask) @@ -699,6 +710,7 @@ static int num_lri_dwords(int num_regs) static void xe_oa_free_oa_buffer(struct xe_oa_stream *stream) { xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + kfree(stream->oa_buffer.bounce); } static void xe_oa_free_configs(struct xe_oa_stream *stream) @@ -889,9 +901,16 @@ static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream, size_t size) return PTR_ERR(bo); stream->oa_buffer.bo = bo; + /* mmap implementation requires OA buffer to be in system memory */ xe_assert(stream->oa->xe, bo->vmap.is_iomem == 0); - stream->oa_buffer.vaddr = bo->vmap.vaddr; + + stream->oa_buffer.bounce = kmalloc(stream->oa_buffer.format->size, GFP_KERNEL); + if (!stream->oa_buffer.bounce) { + xe_bo_unpin_map_no_vm(stream->oa_buffer.bo); + return -ENOMEM; + } + return 0; } diff --git a/drivers/gpu/drm/xe/xe_oa_types.h b/drivers/gpu/drm/xe/xe_oa_types.h index 8906c3084b5f8..3d9ec8490899c 100644 --- a/drivers/gpu/drm/xe/xe_oa_types.h +++ b/drivers/gpu/drm/xe/xe_oa_types.h @@ -164,8 +164,8 @@ struct xe_oa_buffer { /** @bo: xe_bo backing the OA buffer */ struct xe_bo *bo; - /** @vaddr: mapped vaddr of the OA buffer */ - u8 *vaddr; + /** @bounce: bounce buffer used with xe_map layer */ + void *bounce; /** @ptr_lock: Lock protecting reads/writes to head/tail pointers */ spinlock_t ptr_lock; -- 2.54.0 ^ permalink raw reply related [flat|nested] 30+ messages in thread
end of thread, other threads:[~2026-04-27 22:11 UTC | newest] Thread overview: 30+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2026-04-07 3:02 [PATCH 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit 2026-04-07 3:02 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 2026-04-07 22:49 ` Umesh Nerlige Ramappa 2026-04-08 15:32 ` Dixit, Ashutosh 2026-04-08 18:21 ` Umesh Nerlige Ramappa 2026-04-07 3:02 ` [PATCH 2/3] drm/xe/oa: Use drm_gem_mmap_obj for OA buffer mmap Ashutosh Dixit 2026-04-07 22:52 ` Umesh Nerlige Ramappa 2026-04-07 3:02 ` [PATCH 3/3] drm/xe/oa: Implement Wa_14026633728 Ashutosh Dixit 2026-04-07 23:17 ` Umesh Nerlige Ramappa 2026-04-07 3:10 ` ✓ CI.KUnit: success for drm/xe/oa: Wa_14026633728 Patchwork 2026-04-07 3:51 ` ✓ Xe.CI.BAT: " Patchwork 2026-04-07 5:02 ` ✓ Xe.CI.FULL: " Patchwork -- strict thread matches above, loose matches on Subject: below -- 2026-04-09 23:17 [PATCH v2 0/3] " Ashutosh Dixit 2026-04-09 23:17 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 2026-04-11 0:49 ` Dixit, Ashutosh 2026-04-13 18:05 ` Umesh Nerlige Ramappa 2026-04-13 22:06 ` Dixit, Ashutosh 2026-04-11 0:48 [PATCH v3 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit 2026-04-11 0:48 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 2026-04-15 2:03 [PATCH v4 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit 2026-04-15 2:03 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 2026-04-24 21:05 ` Umesh Nerlige Ramappa 2026-04-24 23:56 ` Dixit, Ashutosh 2026-04-25 0:16 ` Dixit, Ashutosh 2026-04-27 18:27 ` Umesh Nerlige Ramappa 2026-04-27 19:04 ` Dixit, Ashutosh 2026-04-25 0:14 [PATCH v5 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit 2026-04-25 0:14 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 2026-04-27 19:23 ` Umesh Nerlige Ramappa 2026-04-27 19:02 [PATCH v6 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit 2026-04-27 19:02 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 2026-04-27 19:34 ` Umesh Nerlige Ramappa 2026-04-27 20:30 ` Dixit, Ashutosh 2026-04-27 20:26 [PATCH v7 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit 2026-04-27 20:26 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit 2026-04-27 22:11 [PATCH v8 0/3] drm/xe/oa: Wa_14026633728 Ashutosh Dixit 2026-04-27 22:11 ` [PATCH 1/3] drm/xe/oa: Use xe_map layer Ashutosh Dixit
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.