| Message ID | 20240206222203.347626-1-jfalempe@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/mgag200: Add a workaround for low-latency | expand |
Hi Am 06.02.24 um 23:21 schrieb Jocelyn Falempe: > We found a regression in v5.10 on real-time server, using the > rt-kernel and the mgag200 driver. It's some really specialized > workload, with <10us latency expectation on isolated core. > After the v5.10, the real time tasks missed their <10us latency > when something prints on the screen (fbcon or printk) > > The regression has been bisected to 2 commits: > commit 0b34d58b6c32 ("drm/mgag200: Enable caching for SHMEM pages") > commit 4862ffaec523 ("drm/mgag200: Move vmap out of commit tail") > > The first one changed the system memory framebuffer from Write-Combine > to the default caching. > Before the second commit, the mgag200 driver used to unmap the > framebuffer after each frame, which implicitly does a cache flush. > Both regressions are fixed by this commit, which restore WC mapping > for the framebuffer in system memory, and add a cache flush. > This is only needed on x86_64, for low-latency workload, > so the new kconfig DRM_MGAG200_LATENCY_WORKAROUND depends on > PREEMPT_RT and X86. > > For more context, the whole thread can be found here: > https://lore.kernel.org/dri-devel/20231019135655.313759-1-jfalempe@redhat.com/ This URL should be in a Link tag below the SoB line, like this: Link: https://lore.kernel.org/dri-devel/20231019135655.313759-1-jfalempe@redhat.com/ # 1 You can refer to it from within the text with [1]. > > Signed-off-by: Jocelyn Falempe <jfalempe@redhat.com> > --- > drivers/gpu/drm/mgag200/Kconfig | 10 ++++++++++ > drivers/gpu/drm/mgag200/mgag200_drv.c | 17 +++++++++++++++++ > drivers/gpu/drm/mgag200/mgag200_mode.c | 8 ++++++++ > 3 files changed, 35 insertions(+) > > diff --git a/drivers/gpu/drm/mgag200/Kconfig b/drivers/gpu/drm/mgag200/Kconfig > index b28c5e4828f4..73e4feba743a 100644 > --- a/drivers/gpu/drm/mgag200/Kconfig > +++ b/drivers/gpu/drm/mgag200/Kconfig > @@ -11,3 +11,13 @@ config DRM_MGAG200 > MGA G200 desktop chips and the server variants. It requires 0.3.0 > of the modesetting userspace driver, and a version of mga driver > that will fail on KMS enabled devices. > + > +config DRM_MGAG200_LATENCY_WORKAROUND Can we call this DRM_MGAG200_IOBURST_WORKAROUND? I know what you mean by latency, but that's not what is happening in the driver. The latency you refer to is the deterministic response time of your process, but the response time of the driver actually goes up (because of the disabled caching). > + bool "Enable workaround for low latency server" This is a 'low-latency server'. But I'd just say "Disabled buffer caching", so that users know what they are getting into. > + depends on DRM_MGAG200 && PREEMPT_RT && X86 > + help > + Enable a workaround to have better latency with mgag200 driver. Here I'd say "Enable a workaround to avoid I/O bursts within the mgag200 driver at the expense of overall display performance." > + It restores the <v5.10 behavior, by mapping the framebuffer in system > + RAM as Write-Combining, and flushing the cache after each write. > + This is only needed on x86_64 and if you want low-latency. Maybe "This is only useful on x86_64 if you want to run processes with deterministic latency." The code itself looks good to me. Best regards Thomas > + If unsure, say N. > diff --git a/drivers/gpu/drm/mgag200/mgag200_drv.c b/drivers/gpu/drm/mgag200/mgag200_drv.c > index 54fce00e2136..3fdef8b580cc 100644 > --- a/drivers/gpu/drm/mgag200/mgag200_drv.c > +++ b/drivers/gpu/drm/mgag200/mgag200_drv.c > @@ -84,6 +84,20 @@ resource_size_t mgag200_probe_vram(void __iomem *mem, resource_size_t size) > return offset - 65536; > } > > +#if defined(CONFIG_DRM_MGAG200_LATENCY_WORKAROUND) > +static struct drm_gem_object *mgag200_create_object(struct drm_device *dev, size_t size) > +{ > + struct drm_gem_shmem_object *shmem; > + > + shmem = kzalloc(sizeof(*shmem), GFP_KERNEL); > + if (!shmem) > + return NULL; > + > + shmem->map_wc = true; > + return &shmem->base; > +} > +#endif > + > /* > * DRM driver > */ > @@ -99,6 +113,9 @@ static const struct drm_driver mgag200_driver = { > .major = DRIVER_MAJOR, > .minor = DRIVER_MINOR, > .patchlevel = DRIVER_PATCHLEVEL, > +#if defined(CONFIG_DRM_MGAG200_LATENCY_WORKAROUND) > + .gem_create_object = mgag200_create_object, > +#endif > DRM_GEM_SHMEM_DRIVER_OPS, > }; > > diff --git a/drivers/gpu/drm/mgag200/mgag200_mode.c b/drivers/gpu/drm/mgag200/mgag200_mode.c > index 0eb769dd76ce..34ef9fb6e96c 100644 > --- a/drivers/gpu/drm/mgag200/mgag200_mode.c > +++ b/drivers/gpu/drm/mgag200/mgag200_mode.c > @@ -13,6 +13,7 @@ > > #include <drm/drm_atomic.h> > #include <drm/drm_atomic_helper.h> > +#include <drm/drm_cache.h> > #include <drm/drm_damage_helper.h> > #include <drm/drm_edid.h> > #include <drm/drm_format_helper.h> > @@ -436,6 +437,13 @@ static void mgag200_handle_damage(struct mga_device *mdev, const struct iosys_ma > > iosys_map_incr(&dst, drm_fb_clip_offset(fb->pitches[0], fb->format, clip)); > drm_fb_memcpy(&dst, fb->pitches, vmap, fb, clip); > + > + /* Flushing the cache greatly improves latency on x86_64 */ > +#if defined(CONFIG_DRM_MGAG200_LATENCY_WORKAROUND) > + if (!vmap->is_iomem) > + drm_clflush_virt_range(vmap->vaddr + clip->y1 * fb->pitches[0], > + drm_rect_height(clip) * fb->pitches[0]); > +#endif > } > > /* > > base-commit: 1f36d634670d8001a45fe2f2dcae546819f9c7d8
On 07/02/2024 10:47, Thomas Zimmermann wrote: > Hi > > Am 06.02.24 um 23:21 schrieb Jocelyn Falempe: >> We found a regression in v5.10 on real-time server, using the >> rt-kernel and the mgag200 driver. It's some really specialized >> workload, with <10us latency expectation on isolated core. >> After the v5.10, the real time tasks missed their <10us latency >> when something prints on the screen (fbcon or printk) >> >> The regression has been bisected to 2 commits: >> commit 0b34d58b6c32 ("drm/mgag200: Enable caching for SHMEM pages") >> commit 4862ffaec523 ("drm/mgag200: Move vmap out of commit tail") >> >> The first one changed the system memory framebuffer from Write-Combine >> to the default caching. >> Before the second commit, the mgag200 driver used to unmap the >> framebuffer after each frame, which implicitly does a cache flush. >> Both regressions are fixed by this commit, which restore WC mapping >> for the framebuffer in system memory, and add a cache flush. >> This is only needed on x86_64, for low-latency workload, >> so the new kconfig DRM_MGAG200_LATENCY_WORKAROUND depends on >> PREEMPT_RT and X86. >> >> For more context, the whole thread can be found here: >> https://lore.kernel.org/dri-devel/20231019135655.313759-1-jfalempe@redhat.com/ > > This URL should be in a Link tag below the SoB line, like this: > > Link: > https://lore.kernel.org/dri-devel/20231019135655.313759-1-jfalempe@redhat.com/ # 1 You can refer to it from within the text with [1]. ok >> >> Signed-off-by: Jocelyn Falempe <jfalempe@redhat.com> >> --- >> drivers/gpu/drm/mgag200/Kconfig | 10 ++++++++++ >> drivers/gpu/drm/mgag200/mgag200_drv.c | 17 +++++++++++++++++ >> drivers/gpu/drm/mgag200/mgag200_mode.c | 8 ++++++++ >> 3 files changed, 35 insertions(+) >> >> diff --git a/drivers/gpu/drm/mgag200/Kconfig >> b/drivers/gpu/drm/mgag200/Kconfig >> index b28c5e4828f4..73e4feba743a 100644 >> --- a/drivers/gpu/drm/mgag200/Kconfig >> +++ b/drivers/gpu/drm/mgag200/Kconfig >> @@ -11,3 +11,13 @@ config DRM_MGAG200 >> MGA G200 desktop chips and the server variants. It requires 0.3.0 >> of the modesetting userspace driver, and a version of mga driver >> that will fail on KMS enabled devices. >> + >> +config DRM_MGAG200_LATENCY_WORKAROUND > > Can we call this DRM_MGAG200_IOBURST_WORKAROUND? I know what you mean by > latency, but that's not what is happening in the driver. The latency you > refer to is the deterministic response time of your process, but the > response time of the driver actually goes up (because of the disabled > caching). > >> + bool "Enable workaround for low latency server" > > This is a 'low-latency server'. But I'd just say "Disabled buffer > caching", so that users know what they are getting into. > >> + depends on DRM_MGAG200 && PREEMPT_RT && X86 >> + help >> + Enable a workaround to have better latency with mgag200 driver. > > Here I'd say "Enable a workaround to avoid I/O bursts within the mgag200 > driver at the expense of overall display performance." > >> + It restores the <v5.10 behavior, by mapping the framebuffer in >> system >> + RAM as Write-Combining, and flushing the cache after each write. >> + This is only needed on x86_64 and if you want low-latency. > > Maybe "This is only useful on x86_64 if you want to run processes with > deterministic latency." > > The code itself looks good to me. Thanks, I will send a v2 with all these changes.
diff --git a/drivers/gpu/drm/mgag200/Kconfig b/drivers/gpu/drm/mgag200/Kconfig index b28c5e4828f4..73e4feba743a 100644 --- a/drivers/gpu/drm/mgag200/Kconfig +++ b/drivers/gpu/drm/mgag200/Kconfig @@ -11,3 +11,13 @@ config DRM_MGAG200 MGA G200 desktop chips and the server variants. It requires 0.3.0 of the modesetting userspace driver, and a version of mga driver that will fail on KMS enabled devices. + +config DRM_MGAG200_LATENCY_WORKAROUND + bool "Enable workaround for low latency server" + depends on DRM_MGAG200 && PREEMPT_RT && X86 + help + Enable a workaround to have better latency with mgag200 driver. + It restores the <v5.10 behavior, by mapping the framebuffer in system + RAM as Write-Combining, and flushing the cache after each write. + This is only needed on x86_64 and if you want low-latency. + If unsure, say N. diff --git a/drivers/gpu/drm/mgag200/mgag200_drv.c b/drivers/gpu/drm/mgag200/mgag200_drv.c index 54fce00e2136..3fdef8b580cc 100644 --- a/drivers/gpu/drm/mgag200/mgag200_drv.c +++ b/drivers/gpu/drm/mgag200/mgag200_drv.c @@ -84,6 +84,20 @@ resource_size_t mgag200_probe_vram(void __iomem *mem, resource_size_t size) return offset - 65536; } +#if defined(CONFIG_DRM_MGAG200_LATENCY_WORKAROUND) +static struct drm_gem_object *mgag200_create_object(struct drm_device *dev, size_t size) +{ + struct drm_gem_shmem_object *shmem; + + shmem = kzalloc(sizeof(*shmem), GFP_KERNEL); + if (!shmem) + return NULL; + + shmem->map_wc = true; + return &shmem->base; +} +#endif + /* * DRM driver */ @@ -99,6 +113,9 @@ static const struct drm_driver mgag200_driver = { .major = DRIVER_MAJOR, .minor = DRIVER_MINOR, .patchlevel = DRIVER_PATCHLEVEL, +#if defined(CONFIG_DRM_MGAG200_LATENCY_WORKAROUND) + .gem_create_object = mgag200_create_object, +#endif DRM_GEM_SHMEM_DRIVER_OPS, }; diff --git a/drivers/gpu/drm/mgag200/mgag200_mode.c b/drivers/gpu/drm/mgag200/mgag200_mode.c index 0eb769dd76ce..34ef9fb6e96c 100644 --- a/drivers/gpu/drm/mgag200/mgag200_mode.c +++ b/drivers/gpu/drm/mgag200/mgag200_mode.c @@ -13,6 +13,7 @@ #include <drm/drm_atomic.h> #include <drm/drm_atomic_helper.h> +#include <drm/drm_cache.h> #include <drm/drm_damage_helper.h> #include <drm/drm_edid.h> #include <drm/drm_format_helper.h> @@ -436,6 +437,13 @@ static void mgag200_handle_damage(struct mga_device *mdev, const struct iosys_ma iosys_map_incr(&dst, drm_fb_clip_offset(fb->pitches[0], fb->format, clip)); drm_fb_memcpy(&dst, fb->pitches, vmap, fb, clip); + + /* Flushing the cache greatly improves latency on x86_64 */ +#if defined(CONFIG_DRM_MGAG200_LATENCY_WORKAROUND) + if (!vmap->is_iomem) + drm_clflush_virt_range(vmap->vaddr + clip->y1 * fb->pitches[0], + drm_rect_height(clip) * fb->pitches[0]); +#endif } /*
We found a regression in v5.10 on real-time server, using the rt-kernel and the mgag200 driver. It's some really specialized workload, with <10us latency expectation on isolated core. After the v5.10, the real time tasks missed their <10us latency when something prints on the screen (fbcon or printk) The regression has been bisected to 2 commits: commit 0b34d58b6c32 ("drm/mgag200: Enable caching for SHMEM pages") commit 4862ffaec523 ("drm/mgag200: Move vmap out of commit tail") The first one changed the system memory framebuffer from Write-Combine to the default caching. Before the second commit, the mgag200 driver used to unmap the framebuffer after each frame, which implicitly does a cache flush. Both regressions are fixed by this commit, which restore WC mapping for the framebuffer in system memory, and add a cache flush. This is only needed on x86_64, for low-latency workload, so the new kconfig DRM_MGAG200_LATENCY_WORKAROUND depends on PREEMPT_RT and X86. For more context, the whole thread can be found here: https://lore.kernel.org/dri-devel/20231019135655.313759-1-jfalempe@redhat.com/ Signed-off-by: Jocelyn Falempe <jfalempe@redhat.com> --- drivers/gpu/drm/mgag200/Kconfig | 10 ++++++++++ drivers/gpu/drm/mgag200/mgag200_drv.c | 17 +++++++++++++++++ drivers/gpu/drm/mgag200/mgag200_mode.c | 8 ++++++++ 3 files changed, 35 insertions(+) base-commit: 1f36d634670d8001a45fe2f2dcae546819f9c7d8