Hi,
I messed-up with versions, my prev comment landed in v2, so I put it
here to clean things up.
On 02.12.2022 13:28, Matthew Auld wrote:
From: Chris Wilson <chris.p.wilson@xxxxxxxxx>
If the ring is nearly full when calling into emit_pte(), we might
incorrectly trample the reserved_space when constructing the packet to
emit the PTEs. This then triggers the GEM_BUG_ON(rq->reserved_space >
ring->space) when later submitting the request, since the request itself
doesn't have enough space left in the ring to emit things like
workarounds, breadcrumbs etc.
v2: Fix the whitespace errors
Testcase: igt@i915_selftests@live_emit_pte_full_ring
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/7535
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/6889
Fixes: cf586021642d ("drm/i915/gt: Pipelined page migration")
Signed-off-by: Chris Wilson <chris.p.wilson@xxxxxxxxx>
Signed-off-by: Matthew Auld <matthew.auld@xxxxxxxxx>
Cc: Andrzej Hajda <andrzej.hajda@xxxxxxxxx>
Cc: Andi Shyti <andi.shyti@xxxxxxxxxxxxxxx>
Cc: Nirmoy Das <nirmoy.das@xxxxxxxxx>
Cc: <stable@xxxxxxxxxxxxxxx> # v5.15+
Tested-by: Nirmoy Das <nirmoy.das@xxxxxxxxx>
Reviewed-by: Nirmoy Das <nirmoy.das@xxxxxxxxx>
---
drivers/gpu/drm/i915/gt/intel_migrate.c | 16 ++++++++++++----
1 file changed, 12 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_migrate.c b/drivers/gpu/drm/i915/gt/intel_migrate.c
index b405a04135ca..b783f6f740c8 100644
--- a/drivers/gpu/drm/i915/gt/intel_migrate.c
+++ b/drivers/gpu/drm/i915/gt/intel_migrate.c
@@ -342,6 +342,16 @@ static int emit_no_arbitration(struct i915_request *rq)
return 0;
}
+static int max_pte_pkt_size(struct i915_request *rq, int pkt)
+{
+ struct intel_ring *ring = rq->ring;
+
+ pkt = min_t(int, pkt, (ring->space - rq->reserved_space) / sizeof(u32) + 5);
+ pkt = min_t(int, pkt, (ring->size - ring->emit) / sizeof(u32) + 5);
+
+ return pkt;
+}
+
I guess, the assumption that subtractions of u32 values do not
overflows, is valid.
Then I guess more natural would be use u32 for all vars involved, this
way we can use min instead of min_t, minor nit.
Anyway:
Reviewed-by: Andrzej Hajda <andrzej.hajda@xxxxxxxxx>
Regards
Andrzej
static int emit_pte(struct i915_request *rq,
struct sgt_dma *it,
enum i915_cache_level cache_level,
@@ -388,8 +398,7 @@ static int emit_pte(struct i915_request *rq,
return PTR_ERR(cs);
/* Pack as many PTE updates as possible into a single MI command */
- pkt = min_t(int, dword_length, ring->space / sizeof(u32) + 5);
- pkt = min_t(int, pkt, (ring->size - ring->emit) / sizeof(u32) + 5);
+ pkt = max_pte_pkt_size(rq, dword_length);
hdr = cs;
*cs++ = MI_STORE_DATA_IMM | REG_BIT(21); /* as qword elements */
@@ -422,8 +431,7 @@ static int emit_pte(struct i915_request *rq,
}
}
- pkt = min_t(int, dword_rem, ring->space / sizeof(u32) + 5);
- pkt = min_t(int, pkt, (ring->size - ring->emit) / sizeof(u32) + 5);
+ pkt = max_pte_pkt_size(rq, dword_rem);
hdr = cs;
*cs++ = MI_STORE_DATA_IMM | REG_BIT(21);
