| Message ID | 1501260863-14687-2-git-send-email-will.deacon@arm.com |
|---|---|
| State | New |
| Headers | show |
| Series | [1/2] perf/aux: Make aux_{head,wakeup} ring_buffer members long | expand |
Will Deacon <will.deacon@arm.com> writes: > The aux_watermark member of struct ring_buffer represents the period (in > terms of bytes) at which wakeup events should be generated when data is > written to the aux buffer in non-snapshot mode. On hardware that cannot > generate an interrupt when the aux_head reaches an arbitrary wakeup index Curious: how do you support non-snapshot trace collection on such hardware? > (such as ARM SPE), the aux_head sampled from handle->head in > perf_aux_output_{skip,end} may in fact be past the wakeup index. This I think this is also true of hw where the interrupt is not precise. Thanks for looking at this. > can lead to wakeup slowly falling behind the head. For example, consider > the case where hardware can only generate an interrupt on a page-boundary > and the aux buffer is initialised as follows: > > // Buffer size is 2 * PAGE_SIZE > rb->aux_head = rb->aux_wakeup = 0 > rb->aux_watermark = PAGE_SIZE / 2 > > following the first perf_aux_output_begin call, the handle is > initialised with: > > handle->head = 0 > handle->size = 2 * PAGE_SIZE > handle->wakeup = PAGE_SIZE / 2 > > and the hardware will be programmed to generate an interrupt at > PAGE_SIZE. > > When the interrupt is raised, the hardware head will be at PAGE_SIZE, > so calling perf_aux_output_end(handle, PAGE_SIZE) puts the ring buffer > into the following state: > > rb->aux_head = PAGE_SIZE > rb->aux_wakeup = PAGE_SIZE / 2 > rb->aux_watermark = PAGE_SIZE / 2 > > and then the next call to perf_aux_output_begin will result in: > > handle->head = handle->wakeup = PAGE_SIZE > > for which the semantics are unclear and, for a smaller aux_watermark > (e.g. PAGE_SIZE / 4), then the wakeup would in fact be behind head at > this point. > > This patch fixes the problem by rounding down the aux_head (as sampled > from the handle) to the nearest aux_watermark boundary when updating > rb->aux_wakeup, therefore taking into account any overruns by the > hardware. Let's add a small comment to the @aux_wakeup field definition? Other than that, Acked-by: Alexander Shishkin <alexander.shishkin@linux.intel.com> > > Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> > Cc: Peter Zijlstra <peterz@infradead.org> > Reported-by: Mark Rutland <mark.rutland@arm.com> > Signed-off-by: Will Deacon <will.deacon@arm.com> > --- > kernel/events/ring_buffer.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c > index 330df5a7f762..8e511e52fc1b 100644 > --- a/kernel/events/ring_buffer.c > +++ b/kernel/events/ring_buffer.c > @@ -455,7 +455,7 @@ void perf_aux_output_end(struct perf_output_handle *handle, unsigned long size) > > if (aux_head - rb->aux_wakeup >= rb->aux_watermark) { > wakeup = true; > - rb->aux_wakeup += rb->aux_watermark; > + rb->aux_wakeup = rounddown(aux_head, rb->aux_watermark); > } > > if (wakeup) { > @@ -490,7 +490,7 @@ int perf_aux_output_skip(struct perf_output_handle *handle, unsigned long size) > aux_head = rb->user_page->aux_head = rb->aux_head; > if (aux_head - rb->aux_wakeup >= rb->aux_watermark) { > perf_output_wakeup(handle); > - rb->aux_wakeup += rb->aux_watermark; > + rb->aux_wakeup = rounddown(aux_head, rb->aux_watermark); > handle->wakeup = rb->aux_wakeup + rb->aux_watermark; > } > > -- > 2.1.4
Hi Alexander, Thanks for having a look at these. On Mon, Jul 31, 2017 at 01:02:16PM +0300, Alexander Shishkin wrote: > Will Deacon <will.deacon@arm.com> writes: > > > The aux_watermark member of struct ring_buffer represents the period (in > > terms of bytes) at which wakeup events should be generated when data is > > written to the aux buffer in non-snapshot mode. On hardware that cannot > > generate an interrupt when the aux_head reaches an arbitrary wakeup index > > Curious: how do you support non-snapshot trace collection on such > hardware? The watermark is constrained to lie on a page boundary, so as long as the buffer is at least a page (which it is!), we end up rounding up to the next page boundary, with lots of fun and games to avoid going past the head. > > (such as ARM SPE), the aux_head sampled from handle->head in > > perf_aux_output_{skip,end} may in fact be past the wakeup index. This > > I think this is also true of hw where the interrupt is not > precise. Thanks for looking at this. Yes, it all looks like "skid" to userspace. > > can lead to wakeup slowly falling behind the head. For example, consider > > the case where hardware can only generate an interrupt on a page-boundary > > and the aux buffer is initialised as follows: > > > > // Buffer size is 2 * PAGE_SIZE > > rb->aux_head = rb->aux_wakeup = 0 > > rb->aux_watermark = PAGE_SIZE / 2 > > > > following the first perf_aux_output_begin call, the handle is > > initialised with: > > > > handle->head = 0 > > handle->size = 2 * PAGE_SIZE > > handle->wakeup = PAGE_SIZE / 2 > > > > and the hardware will be programmed to generate an interrupt at > > PAGE_SIZE. > > > > When the interrupt is raised, the hardware head will be at PAGE_SIZE, > > so calling perf_aux_output_end(handle, PAGE_SIZE) puts the ring buffer > > into the following state: > > > > rb->aux_head = PAGE_SIZE > > rb->aux_wakeup = PAGE_SIZE / 2 > > rb->aux_watermark = PAGE_SIZE / 2 > > > > and then the next call to perf_aux_output_begin will result in: > > > > handle->head = handle->wakeup = PAGE_SIZE > > > > for which the semantics are unclear and, for a smaller aux_watermark > > (e.g. PAGE_SIZE / 4), then the wakeup would in fact be behind head at > > this point. > > > > This patch fixes the problem by rounding down the aux_head (as sampled > > from the handle) to the nearest aux_watermark boundary when updating > > rb->aux_wakeup, therefore taking into account any overruns by the > > hardware. > > Let's add a small comment to the @aux_wakeup field definition? Other > than that, Good thinking; I'll do that. > Acked-by: Alexander Shishkin <alexander.shishkin@linux.intel.com> Thanks! Will
diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c index 330df5a7f762..8e511e52fc1b 100644 --- a/kernel/events/ring_buffer.c +++ b/kernel/events/ring_buffer.c @@ -455,7 +455,7 @@ void perf_aux_output_end(struct perf_output_handle *handle, unsigned long size) if (aux_head - rb->aux_wakeup >= rb->aux_watermark) { wakeup = true; - rb->aux_wakeup += rb->aux_watermark; + rb->aux_wakeup = rounddown(aux_head, rb->aux_watermark); } if (wakeup) { @@ -490,7 +490,7 @@ int perf_aux_output_skip(struct perf_output_handle *handle, unsigned long size) aux_head = rb->user_page->aux_head = rb->aux_head; if (aux_head - rb->aux_wakeup >= rb->aux_watermark) { perf_output_wakeup(handle); - rb->aux_wakeup += rb->aux_watermark; + rb->aux_wakeup = rounddown(aux_head, rb->aux_watermark); handle->wakeup = rb->aux_wakeup + rb->aux_watermark; }
The aux_watermark member of struct ring_buffer represents the period (in terms of bytes) at which wakeup events should be generated when data is written to the aux buffer in non-snapshot mode. On hardware that cannot generate an interrupt when the aux_head reaches an arbitrary wakeup index (such as ARM SPE), the aux_head sampled from handle->head in perf_aux_output_{skip,end} may in fact be past the wakeup index. This can lead to wakeup slowly falling behind the head. For example, consider the case where hardware can only generate an interrupt on a page-boundary and the aux buffer is initialised as follows: // Buffer size is 2 * PAGE_SIZE rb->aux_head = rb->aux_wakeup = 0 rb->aux_watermark = PAGE_SIZE / 2 following the first perf_aux_output_begin call, the handle is initialised with: handle->head = 0 handle->size = 2 * PAGE_SIZE handle->wakeup = PAGE_SIZE / 2 and the hardware will be programmed to generate an interrupt at PAGE_SIZE. When the interrupt is raised, the hardware head will be at PAGE_SIZE, so calling perf_aux_output_end(handle, PAGE_SIZE) puts the ring buffer into the following state: rb->aux_head = PAGE_SIZE rb->aux_wakeup = PAGE_SIZE / 2 rb->aux_watermark = PAGE_SIZE / 2 and then the next call to perf_aux_output_begin will result in: handle->head = handle->wakeup = PAGE_SIZE for which the semantics are unclear and, for a smaller aux_watermark (e.g. PAGE_SIZE / 4), then the wakeup would in fact be behind head at this point. This patch fixes the problem by rounding down the aux_head (as sampled from the handle) to the nearest aux_watermark boundary when updating rb->aux_wakeup, therefore taking into account any overruns by the hardware. Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Reported-by: Mark Rutland <mark.rutland@arm.com> Signed-off-by: Will Deacon <will.deacon@arm.com> --- kernel/events/ring_buffer.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) -- 2.1.4