* [PATCH v3 0/3] Flat-CCS eviction enhancements
@ 2022-05-02 14:26 Ramalingam C
2022-05-02 14:26 ` [PATCH v3 1/3] drm/i915/gt: BUG_ON unexpected NULL at scatterlist walking Ramalingam C
` (2 more replies)
0 siblings, 3 replies; 5+ messages in thread
From: Ramalingam C @ 2022-05-02 14:26 UTC (permalink / raw)
To: intel-gfx, dri-devel; +Cc: Matthew Auld
Flat-CCS eviction enhancements
v3:
Incorporated the review suggestions [Matt]
Ramalingam C (3):
drm/i915/gt: BUG_ON unexpected NULL at scatterlist walking
drm/i915/gt: optimize the ccs_sz calculation per chunk
drm/i915/gt: Document the eviction of the Flat-CCS objects
drivers/gpu/drm/i915/gt/intel_migrate.c | 73 ++++++++++++++-----------
1 file changed, 40 insertions(+), 33 deletions(-)
--
2.20.1
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH v3 1/3] drm/i915/gt: BUG_ON unexpected NULL at scatterlist walking
2022-05-02 14:26 [PATCH v3 0/3] Flat-CCS eviction enhancements Ramalingam C
@ 2022-05-02 14:26 ` Ramalingam C
2022-05-02 16:45 ` Jani Nikula
2022-05-02 14:26 ` [PATCH v3 2/3] drm/i915/gt: optimize the ccs_sz calculation per chunk Ramalingam C
2022-05-02 14:26 ` [PATCH v3 3/3] drm/i915/gt: Document the eviction of the Flat-CCS objects Ramalingam C
2 siblings, 1 reply; 5+ messages in thread
From: Ramalingam C @ 2022-05-02 14:26 UTC (permalink / raw)
To: intel-gfx, dri-devel; +Cc: Matthew Auld
While locating the start of ccs scatterlist in smem scatterlist, that has
to be the size of lmem obj size + corresponding ccs data size, report bug
if scatterlist terminate before that length.
v2:
s/GEM_BUG_ON/BUG_ON with more commenting [Matt]
v3:
Converted GEM_BUG_ON into BUG_ON with more documentation [Matt]
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com> (v1)
---
drivers/gpu/drm/i915/gt/intel_migrate.c | 14 +++++++++++++-
1 file changed, 13 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_migrate.c b/drivers/gpu/drm/i915/gt/intel_migrate.c
index 9d552f30b627..168d17b6f48a 100644
--- a/drivers/gpu/drm/i915/gt/intel_migrate.c
+++ b/drivers/gpu/drm/i915/gt/intel_migrate.c
@@ -687,6 +687,16 @@ static void get_ccs_sg_sgt(struct sgt_dma *it, u32 bytes_to_cpy)
bytes_to_cpy -= len;
it->sg = __sg_next(it->sg);
+
+ /*
+ * On Flat-CCS capable platform when we back the lmem pages with
+ * smem pages we add extra pages at the end of the smem
+ * scatterlist, to store the ccs data corresponding to the lmem
+ * pages. get_ccs_sg_sgt() is called to get the pointer for the
+ * start of the extra pages added at the end of smem scatterlist.
+ * So scatterlist can't end at or before bytes_to_cpy.
+ */
+ BUG_ON(!it->sg);
it->dma = sg_dma_address(it->sg);
it->max = it->dma + sg_dma_len(it->sg);
} while (bytes_to_cpy);
@@ -748,8 +758,10 @@ intel_context_migrate_copy(struct intel_context *ce,
* Need to fix it.
*/
ccs_bytes_to_cpy = src_sz != dst_sz ? GET_CCS_BYTES(i915, bytes_to_cpy) : 0;
- if (ccs_bytes_to_cpy)
+ if (ccs_bytes_to_cpy) {
+ WARN_ON(abs(src_sz - dst_sz) < ccs_bytes_to_cpy);
get_ccs_sg_sgt(&it_ccs, bytes_to_cpy);
+ }
}
src_offset = 0;
--
2.20.1
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH v3 2/3] drm/i915/gt: optimize the ccs_sz calculation per chunk
2022-05-02 14:26 [PATCH v3 0/3] Flat-CCS eviction enhancements Ramalingam C
2022-05-02 14:26 ` [PATCH v3 1/3] drm/i915/gt: BUG_ON unexpected NULL at scatterlist walking Ramalingam C
@ 2022-05-02 14:26 ` Ramalingam C
2022-05-02 14:26 ` [PATCH v3 3/3] drm/i915/gt: Document the eviction of the Flat-CCS objects Ramalingam C
2 siblings, 0 replies; 5+ messages in thread
From: Ramalingam C @ 2022-05-02 14:26 UTC (permalink / raw)
To: intel-gfx, dri-devel; +Cc: Matthew Auld
Calculate the ccs_sz that needs to be emitted based on the src
and dst pages emitted per chunk. And handle the return value of emit_pte
for the ccs pages.
v2:
ccs_sz moved to the reduced scope [Matt]
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/gt/intel_migrate.c | 36 +++++++++----------------
1 file changed, 13 insertions(+), 23 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_migrate.c b/drivers/gpu/drm/i915/gt/intel_migrate.c
index 168d17b6f48a..fc6975e55fae 100644
--- a/drivers/gpu/drm/i915/gt/intel_migrate.c
+++ b/drivers/gpu/drm/i915/gt/intel_migrate.c
@@ -647,17 +647,9 @@ static int scatter_list_length(struct scatterlist *sg)
static void
calculate_chunk_sz(struct drm_i915_private *i915, bool src_is_lmem,
- int *src_sz, int *ccs_sz, u32 bytes_to_cpy,
- u32 ccs_bytes_to_cpy)
+ int *src_sz, u32 bytes_to_cpy, u32 ccs_bytes_to_cpy)
{
if (ccs_bytes_to_cpy) {
- /*
- * We can only copy the ccs data corresponding to
- * the CHUNK_SZ of lmem which is
- * GET_CCS_BYTES(i915, CHUNK_SZ))
- */
- *ccs_sz = min_t(int, ccs_bytes_to_cpy, GET_CCS_BYTES(i915, CHUNK_SZ));
-
if (!src_is_lmem)
/*
* When CHUNK_SZ is passed all the pages upto CHUNK_SZ
@@ -717,10 +709,10 @@ intel_context_migrate_copy(struct intel_context *ce,
struct drm_i915_private *i915 = ce->engine->i915;
u32 ccs_bytes_to_cpy = 0, bytes_to_cpy;
enum i915_cache_level ccs_cache_level;
- int src_sz, dst_sz, ccs_sz;
u32 src_offset, dst_offset;
u8 src_access, dst_access;
struct i915_request *rq;
+ int src_sz, dst_sz;
bool ccs_is_src;
int err;
@@ -803,7 +795,7 @@ intel_context_migrate_copy(struct intel_context *ce,
if (err)
goto out_rq;
- calculate_chunk_sz(i915, src_is_lmem, &src_sz, &ccs_sz,
+ calculate_chunk_sz(i915, src_is_lmem, &src_sz,
bytes_to_cpy, ccs_bytes_to_cpy);
len = emit_pte(rq, &it_src, src_cache_level, src_is_lmem,
@@ -837,37 +829,35 @@ intel_context_migrate_copy(struct intel_context *ce,
bytes_to_cpy -= len;
if (ccs_bytes_to_cpy) {
+ int ccs_sz;
+
err = rq->engine->emit_flush(rq, EMIT_INVALIDATE);
if (err)
goto out_rq;
+ ccs_sz = GET_CCS_BYTES(i915, len);
err = emit_pte(rq, &it_ccs, ccs_cache_level, false,
ccs_is_src ? src_offset : dst_offset,
ccs_sz);
+ if (err < 0)
+ goto out_rq;
+ if (err < ccs_sz) {
+ err = -EINVAL;
+ goto out_rq;
+ }
err = rq->engine->emit_flush(rq, EMIT_INVALIDATE);
if (err)
goto out_rq;
- /*
- * Using max of src_sz and dst_sz, as we need to
- * pass the lmem size corresponding to the ccs
- * blocks we need to handle.
- */
- ccs_sz = max_t(int, ccs_is_src ? ccs_sz : src_sz,
- ccs_is_src ? dst_sz : ccs_sz);
-
err = emit_copy_ccs(rq, dst_offset, dst_access,
- src_offset, src_access, ccs_sz);
+ src_offset, src_access, len);
if (err)
goto out_rq;
err = rq->engine->emit_flush(rq, EMIT_INVALIDATE);
if (err)
goto out_rq;
-
- /* Converting back to ccs bytes */
- ccs_sz = GET_CCS_BYTES(rq->engine->i915, ccs_sz);
ccs_bytes_to_cpy -= ccs_sz;
}
--
2.20.1
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH v3 3/3] drm/i915/gt: Document the eviction of the Flat-CCS objects
2022-05-02 14:26 [PATCH v3 0/3] Flat-CCS eviction enhancements Ramalingam C
2022-05-02 14:26 ` [PATCH v3 1/3] drm/i915/gt: BUG_ON unexpected NULL at scatterlist walking Ramalingam C
2022-05-02 14:26 ` [PATCH v3 2/3] drm/i915/gt: optimize the ccs_sz calculation per chunk Ramalingam C
@ 2022-05-02 14:26 ` Ramalingam C
2 siblings, 0 replies; 5+ messages in thread
From: Ramalingam C @ 2022-05-02 14:26 UTC (permalink / raw)
To: intel-gfx, dri-devel; +Cc: Matthew Auld, Thomas Hellstrom
Capture the eviction details for Flat-CCS capable, lmem objects.
v2:
Fix the Flat-ccs capbility of lmem obj with smem residency
possibility [Thomas]
v3:
Fixed the suggestions [Matt]
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
cc: Thomas Hellstrom <thomas.hellstrom@linux.intel.com>
cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/gt/intel_migrate.c | 23 ++++++++++++++---------
1 file changed, 14 insertions(+), 9 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_migrate.c b/drivers/gpu/drm/i915/gt/intel_migrate.c
index fc6975e55fae..509955885b93 100644
--- a/drivers/gpu/drm/i915/gt/intel_migrate.c
+++ b/drivers/gpu/drm/i915/gt/intel_migrate.c
@@ -485,16 +485,21 @@ static bool wa_1209644611_applies(int ver, u32 size)
* And CCS data can be copied in and out of CCS region through
* XY_CTRL_SURF_COPY_BLT. CPU can't access the CCS data directly.
*
- * When we exhaust the lmem, if the object's placements support smem, then we can
- * directly decompress the compressed lmem object into smem and start using it
- * from smem itself.
+ * I915 supports Flat-CCS on lmem only objects. When an objects has smem in
+ * its preference list, on memory pressure, i915 needs to migrate the lmem
+ * content into smem. If the lmem object is Flat-CCS compressed by userspace,
+ * then i915 needs to decompress it. But I915 lack the required information
+ * for such decompression. Hence I915 supports Flat-CCS only on lmem only objects.
*
- * But when we need to swapout the compressed lmem object into a smem region
- * though objects' placement doesn't support smem, then we copy the lmem content
- * as it is into smem region along with ccs data (using XY_CTRL_SURF_COPY_BLT).
- * When the object is referred, lmem content will be swaped in along with
- * restoration of the CCS data (using XY_CTRL_SURF_COPY_BLT) at corresponding
- * location.
+ * When we exhaust the lmem, Flat-CCS capable objects' lmem backing memory can
+ * be temporarily evicted to smem, along with the auxiliary CCS state, where
+ * it can be potentially swapped-out at a later point, if required.
+ * If userspace later touches the evicted pages, then we always move
+ * the backing memory back to lmem, which includes restoring the saved CCS state,
+ * and potentially performing any required swap-in.
+ *
+ * For the migration of the lmem objects with smem in placement list, such as
+ * {lmem, smem}, objects are treated as non Flat-CCS capable objects.
*/
static inline u32 *i915_flush_dw(u32 *cmd, u32 flags)
--
2.20.1
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH v3 1/3] drm/i915/gt: BUG_ON unexpected NULL at scatterlist walking
2022-05-02 14:26 ` [PATCH v3 1/3] drm/i915/gt: BUG_ON unexpected NULL at scatterlist walking Ramalingam C
@ 2022-05-02 16:45 ` Jani Nikula
0 siblings, 0 replies; 5+ messages in thread
From: Jani Nikula @ 2022-05-02 16:45 UTC (permalink / raw)
To: Ramalingam C, intel-gfx, dri-devel; +Cc: Matthew Auld
On Mon, 02 May 2022, Ramalingam C <ramalingam.c@intel.com> wrote:
> While locating the start of ccs scatterlist in smem scatterlist, that has
> to be the size of lmem obj size + corresponding ccs data size, report bug
> if scatterlist terminate before that length.
>
> v2:
> s/GEM_BUG_ON/BUG_ON with more commenting [Matt]
> v3:
> Converted GEM_BUG_ON into BUG_ON with more documentation [Matt]
>
> Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
> Reviewed-by: Matthew Auld <matthew.auld@intel.com> (v1)
> ---
> drivers/gpu/drm/i915/gt/intel_migrate.c | 14 +++++++++++++-
> 1 file changed, 13 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_migrate.c b/drivers/gpu/drm/i915/gt/intel_migrate.c
> index 9d552f30b627..168d17b6f48a 100644
> --- a/drivers/gpu/drm/i915/gt/intel_migrate.c
> +++ b/drivers/gpu/drm/i915/gt/intel_migrate.c
> @@ -687,6 +687,16 @@ static void get_ccs_sg_sgt(struct sgt_dma *it, u32 bytes_to_cpy)
> bytes_to_cpy -= len;
>
> it->sg = __sg_next(it->sg);
> +
> + /*
> + * On Flat-CCS capable platform when we back the lmem pages with
> + * smem pages we add extra pages at the end of the smem
> + * scatterlist, to store the ccs data corresponding to the lmem
> + * pages. get_ccs_sg_sgt() is called to get the pointer for the
> + * start of the extra pages added at the end of smem scatterlist.
> + * So scatterlist can't end at or before bytes_to_cpy.
> + */
> + BUG_ON(!it->sg);
Why would you have to bring the entire kernel down in this case? Why not
just let it oops on the NULL pointer dereference?
I'd prefer nuking *all* of the current BUG/BUG_ON in the driver, and not
add any single one back.
BR,
Jani.
> it->dma = sg_dma_address(it->sg);
> it->max = it->dma + sg_dma_len(it->sg);
> } while (bytes_to_cpy);
> @@ -748,8 +758,10 @@ intel_context_migrate_copy(struct intel_context *ce,
> * Need to fix it.
> */
> ccs_bytes_to_cpy = src_sz != dst_sz ? GET_CCS_BYTES(i915, bytes_to_cpy) : 0;
> - if (ccs_bytes_to_cpy)
> + if (ccs_bytes_to_cpy) {
> + WARN_ON(abs(src_sz - dst_sz) < ccs_bytes_to_cpy);
> get_ccs_sg_sgt(&it_ccs, bytes_to_cpy);
> + }
> }
>
> src_offset = 0;
--
Jani Nikula, Intel Open Source Graphics Center
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2022-05-02 16:45 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-05-02 14:26 [PATCH v3 0/3] Flat-CCS eviction enhancements Ramalingam C
2022-05-02 14:26 ` [PATCH v3 1/3] drm/i915/gt: BUG_ON unexpected NULL at scatterlist walking Ramalingam C
2022-05-02 16:45 ` Jani Nikula
2022-05-02 14:26 ` [PATCH v3 2/3] drm/i915/gt: optimize the ccs_sz calculation per chunk Ramalingam C
2022-05-02 14:26 ` [PATCH v3 3/3] drm/i915/gt: Document the eviction of the Flat-CCS objects Ramalingam C
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).