Some of the newer OA formats are not powers of 2. For those formats,
hw_tail is adjusted accordingly when checking for new reports.
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.rama...@intel.com>
---
drivers/gpu/drm/i915/i915_perf.c | 50 ++++++++++++++++++--------------
1 file changed, 28 insertions(+), 22 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 030f3a598229..2e1b305032c0 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -542,6 +542,7 @@ static bool oa_buffer_check_unlocked(struct
i915_perf_stream *stream)
bool pollin;
u32 hw_tail;
u64 now;
+ u32 partial_report_size;
/* We have to consider the (unlikely) possibility that read() errors
* could result in an OA buffer reset which might reset the head and
@@ -551,10 +552,16 @@ static bool oa_buffer_check_unlocked(struct
i915_perf_stream *stream)
hw_tail = stream->perf->ops.oa_hw_tail_read(stream);
- /* The tail pointer increases in 64 byte increments,
- * not in report_size steps...
+ /* The tail pointer increases in 64 byte increments, whereas report
+ * sizes need not be integral multiples or 64 or powers of 2.
+ * Compute potentially partially landed report in the OA buffer
*/
- hw_tail &= ~(report_size - 1);
+ partial_report_size = OA_TAKEN(hw_tail, stream->oa_buffer.tail);
+ partial_report_size %= report_size;
+
+ /* Subtract partial amount off the tail */
+ hw_tail = gtt_offset + ((hw_tail - partial_report_size) &
+ (stream->oa_buffer.vma->size - 1));
now = ktime_get_mono_fast_ns();
@@ -677,6 +684,8 @@ static int append_oa_sample(struct i915_perf_stream *stream,
{
int report_size = stream->oa_buffer.format->size;
struct drm_i915_perf_record_header header;
+ int report_size_partial;
+ u8 *oa_buf_end;
header.type = DRM_I915_PERF_RECORD_SAMPLE;
header.pad = 0;
@@ -690,8 +699,21 @@ static int append_oa_sample(struct i915_perf_stream
*stream,
return -EFAULT;
buf += sizeof(header);
- if (copy_to_user(buf, report, report_size))
+ oa_buf_end = stream->oa_buffer.vaddr +
+ stream->oa_buffer.vma->size;
+ report_size_partial = oa_buf_end - report;
+
+ if (report_size_partial < report_size) {
+ if (copy_to_user(buf, report, report_size_partial))
+ return -EFAULT;
+ buf += report_size_partial;
+
+ if (copy_to_user(buf, stream->oa_buffer.vaddr,
+ report_size - report_size_partial))
+ return -EFAULT;
+ } else if (copy_to_user(buf, report, report_size)) {
return -EFAULT;
+ }
(*offset) += header.size;
@@ -759,8 +781,8 @@ static int gen8_append_oa_reports(struct i915_perf_stream
*stream,
* all a power of two).
*/
if (drm_WARN_ONCE(&uncore->i915->drm,
- head > OA_BUFFER_SIZE || head % report_size ||
- tail > OA_BUFFER_SIZE || tail % report_size,
+ head > OA_BUFFER_SIZE ||
+ tail > OA_BUFFER_SIZE,
"Inconsistent OA buffer pointers: head = %u, tail =
%u\n",
head, tail))
return -EIO;
@@ -774,22 +796,6 @@ static int gen8_append_oa_reports(struct i915_perf_stream
*stream,
u32 ctx_id;
u64 reason;
- /*
- * All the report sizes factor neatly into the buffer
- * size so we never expect to see a report split
- * between the beginning and end of the buffer.
- *
- * Given the initial alignment check a misalignment
- * here would imply a driver bug that would result
- * in an overrun.
- */
- if (drm_WARN_ON(&uncore->i915->drm,
- (OA_BUFFER_SIZE - head) < report_size)) {
- drm_err(&uncore->i915->drm,
- "Spurious OA head ptr: non-integral report
offset\n");
- break;
- }
-
/*
* The reason field includes flags identifying what
* triggered this specific report (mostly timer
--
2.36.1
