> -----Original Message-----
> From: Auld, Matthew <matthew.auld@xxxxxxxxx>
> Sent: Wednesday, July 27, 2022 10:14 PM
> To: intel-gfx@xxxxxxxxxxxxxxxxxxxxx
> Cc: dri-devel@xxxxxxxxxxxxxxxxxxxxx; Thomas Hellström <thomas.hellstrom@xxxxxxxxxxxxxxx>; C,
> Ramalingam <ramalingam.c@xxxxxxxxx>
> Subject: [PATCH] drm/i915/ttm: don't leak the ccs state
>
> The kernel only manages the ccs state with lmem-only objects, however the kernel should still take
> care not to leak the CCS state from the previous user.
>
> Fixes: 48760ffe923a ("drm/i915/gt: Clear compress metadata for Flat-ccs objects")
> Signed-off-by: Matthew Auld <matthew.auld@xxxxxxxxx>
> Cc: Thomas Hellström <thomas.hellstrom@xxxxxxxxxxxxxxx>
> Cc: Ramalingam C <ramalingam.c@xxxxxxxxx>
> ---
> drivers/gpu/drm/i915/gt/intel_migrate.c | 23 ++++++++++++++++++++++-
> 1 file changed, 22 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_migrate.c b/drivers/gpu/drm/i915/gt/intel_migrate.c
> index a69b244f14d0..9a0814422ba4 100644
> --- a/drivers/gpu/drm/i915/gt/intel_migrate.c
> +++ b/drivers/gpu/drm/i915/gt/intel_migrate.c
> @@ -708,7 +708,7 @@ intel_context_migrate_copy(struct intel_context *ce,
> u8 src_access, dst_access;
> struct i915_request *rq;
> int src_sz, dst_sz;
> - bool ccs_is_src;
> + bool ccs_is_src, overwrite_ccs;
> int err;
>
> GEM_BUG_ON(ce->vm != ce->engine->gt->migrate.context->vm);
> @@ -749,6 +749,8 @@ intel_context_migrate_copy(struct intel_context *ce,
> get_ccs_sg_sgt(&it_ccs, bytes_to_cpy);
> }
>
> + overwrite_ccs = HAS_FLAT_CCS(i915) && !ccs_bytes_to_cpy &&
> +dst_is_lmem;
> +
> src_offset = 0;
> dst_offset = CHUNK_SZ;
> if (HAS_64K_PAGES(ce->engine->i915)) { @@ -852,6 +854,25 @@
> intel_context_migrate_copy(struct intel_context *ce,
> if (err)
> goto out_rq;
> ccs_bytes_to_cpy -= ccs_sz;
> + } else if (overwrite_ccs) {
> + err = rq->engine->emit_flush(rq, EMIT_INVALIDATE);
> + if (err)
> + goto out_rq;
> +
> + /*
> + * While we can't always restore/manage the CCS state,
> + * we still need to ensure we don't leak the CCS state
> + * from the previous user, so make sure we overwrite it
> + * with something.
> + */
> + err = emit_copy_ccs(rq, dst_offset, INDIRECT_ACCESS,
> + dst_offset, DIRECT_ACCESS, len);
> + if (err)
> + goto out_rq;
> +
> + err = rq->engine->emit_flush(rq, EMIT_INVALIDATE);
> + if (err)
> + goto out_rq;
The change is looking good to the purpose. But shouldn't this be the part of lmem allocation itself?
Ram.
> }
>
> /* Arbitration is re-enabled between requests. */
> --
> 2.37.1
