From: Daniel Vetter <daniel@ffwll.ch>
To: Chris Wilson <chris@chris-wilson.co.uk>
Cc: intel-gfx@lists.freedesktop.org,
David Laight <David.Laight@aculab.com>,
dri-devel@lists.freedesktop.org
Subject: Re: [Intel-gfx] [PATCH] drm: Inject a cond_resched() into long drm_clflush_sg()
Date: Thu, 16 Jan 2020 07:52:42 +0100 [thread overview]
Message-ID: <20200116065242.GC8400@dvetter-linux.ger.corp.intel.com> (raw)
In-Reply-To: <20200115205245.2772800-1-chris@chris-wilson.co.uk>
On Wed, Jan 15, 2020 at 08:52:45PM +0000, Chris Wilson wrote:
> Since we may try and flush the cachelines associated with large buffers
> (an 8K framebuffer is about 128MiB, even before we try HDR), this leads
> to unacceptably long latencies (when using a voluntary CONFIG_PREEMPT).
> If we call cond_resched() between each sg chunk, that it about every 128
> pages, we have a natural break point in which to check if the process
> needs to be rescheduled. Naturally, this means that drm_clflush_sg() can
> only be called from process context -- which is true at the moment. The
> other clflush routines remain usable from atomic context.
>
> Even though flushing large objects takes a demonstrable amount to time
> to flush all the cachelines, clflush is still preferred over a
> system-wide wbinvd as the latter has unpredictable latencies affecting
> the whole system not just the local task.
>
> Reported-by: David Laight <David.Laight@ACULAB.COM>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: David Laight <David.Laight@ACULAB.COM>
The original bug report is complaining about latencies for SCHED_RT
threads, on a system that doesn't even use CONFIG_PREEMPT. I'm not sure
it's terribly valid to cater to that use-case - all the desktop distros
seem a lot more reasonable. So firmly *shrug* from my side ...
Patch itself looks correct, just not seeing the point.
-Daniel
> ---
> drivers/gpu/drm/drm_cache.c | 49 ++++++++++++++++++++++++++++++++++---
> 1 file changed, 45 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_cache.c b/drivers/gpu/drm/drm_cache.c
> index 03e01b000f7a..fbd2bb644544 100644
> --- a/drivers/gpu/drm/drm_cache.c
> +++ b/drivers/gpu/drm/drm_cache.c
> @@ -112,23 +112,64 @@ drm_clflush_pages(struct page *pages[], unsigned long num_pages)
> }
> EXPORT_SYMBOL(drm_clflush_pages);
>
> +static __always_inline struct sgt_iter {
> + struct scatterlist *sgp;
> + unsigned long pfn;
> + unsigned int curr;
> + unsigned int max;
> +} __sgt_iter(struct scatterlist *sgl) {
> + struct sgt_iter s = { .sgp = sgl };
> +
> + if (s.sgp) {
> + s.max = s.curr = s.sgp->offset;
> + s.max += s.sgp->length;
> + s.pfn = page_to_pfn(sg_page(s.sgp));
> + }
> +
> + return s;
> +}
> +
> +static inline struct scatterlist *__sg_next_resched(struct scatterlist *sg)
> +{
> + if (sg_is_last(sg))
> + return NULL;
> +
> + ++sg;
> + if (unlikely(sg_is_chain(sg))) {
> + sg = sg_chain_ptr(sg);
> + cond_resched();
> + }
> + return sg;
> +}
> +
> +#define for_each_sgt_page(__pp, __iter, __sgt) \
> + for ((__iter) = __sgt_iter((__sgt)->sgl); \
> + ((__pp) = (__iter).pfn == 0 ? NULL : \
> + pfn_to_page((__iter).pfn + ((__iter).curr >> PAGE_SHIFT))); \
> + (((__iter).curr += PAGE_SIZE) >= (__iter).max) ? \
> + (__iter) = __sgt_iter(__sg_next_resched((__iter).sgp)), 0 : 0)
> +
> /**
> * drm_clflush_sg - Flush dcache lines pointing to a scather-gather.
> * @st: struct sg_table.
> *
> * Flush every data cache line entry that points to an address in the
> - * sg.
> + * sg. This may schedule between scatterlist chunks, in order to keep
> + * the system preemption-latency down for large buffers.
> */
> void
> drm_clflush_sg(struct sg_table *st)
> {
> + might_sleep();
> +
> #if defined(CONFIG_X86)
> if (static_cpu_has(X86_FEATURE_CLFLUSH)) {
> - struct sg_page_iter sg_iter;
> + struct sgt_iter sg_iter;
> + struct page *page;
>
> mb(); /*CLFLUSH is ordered only by using memory barriers*/
> - for_each_sg_page(st->sgl, &sg_iter, st->nents, 0)
> - drm_clflush_page(sg_page_iter_page(&sg_iter));
> + for_each_sgt_page(page, sg_iter, st)
> + drm_clflush_page(page);
> mb(); /*Make sure that all cache line entry is flushed*/
>
> return;
> --
> 2.25.0
>
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
next prev parent reply other threads:[~2020-01-16 6:52 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-01-15 20:52 [Intel-gfx] [PATCH] drm: Inject a cond_resched() into long drm_clflush_sg() Chris Wilson
2020-01-15 21:28 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for " Patchwork
2020-01-15 21:57 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2020-01-15 21:57 ` [Intel-gfx] ✗ Fi.CI.BUILD: warning " Patchwork
2020-01-16 6:52 ` Daniel Vetter [this message]
2020-01-16 7:40 ` [Intel-gfx] [PATCH] " Chris Wilson
2020-01-16 12:26 ` David Laight
2020-01-16 12:28 ` Chris Wilson
2020-01-16 13:58 ` David Laight
2020-01-16 14:01 ` Chris Wilson
2020-01-16 14:40 ` David Laight
2020-01-16 15:04 ` David Laight
2020-01-18 10:45 ` [Intel-gfx] ✓ Fi.CI.IGT: success for " Patchwork
2020-01-23 14:36 ` [Intel-gfx] [PATCH] " David Laight
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20200116065242.GC8400@dvetter-linux.ger.corp.intel.com \
--to=daniel@ffwll.ch \
--cc=David.Laight@aculab.com \
--cc=chris@chris-wilson.co.uk \
--cc=dri-devel@lists.freedesktop.org \
--cc=intel-gfx@lists.freedesktop.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).