DRM_VMW_EVENT_FENCE_SIGNALED (struct drm_vmw_event_fence) and DRM_EVENT_VBLANK (struct drm_event_vblank) pass timestamps in 32-bit seconds/microseconds format.
As of commit c61eef726a78 ("drm: add support for monotonic vblank timestamps"), other DRM drivers use monotonic times for drm_event_vblank, but vmwgfx still uses CLOCK_REALTIME for both events, which suffers from the y2038/y2106 overflow as well as time jumps.
For consistency, this changes vmwgfx to use ktime_get_ts64 as well, which solves those problems and avoids the deprecated do_gettimeofday() function.
This should be transparent to to user space, as long as it doesn't compare the time against the result of gettimeofday().
Link: https://patchwork.kernel.org/patch/10076599/ Signed-off-by: Arnd Bergmann arnd@arndb.de --- Originally sent on Nov 27. Sinclair Yeh said he'd pick it up for the next pull request, but it's not in linux-next yet.
Resending the unchanged patch, please pick it up when you have time, or feel free to ignore this email in case it's already in some tree that just isn't part of linux-next but will be sent during the next merge window. --- drivers/gpu/drm/vmwgfx/vmwgfx_fence.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c index 6c5c75cf5e6c..9ed544f8958f 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c @@ -901,11 +901,12 @@ static void vmw_event_fence_action_seq_passed(struct vmw_fence_action *action) spin_lock_irq(&dev->event_lock);
if (likely(eaction->tv_sec != NULL)) { - struct timeval tv; + struct timespec64 ts;
- do_gettimeofday(&tv); - *eaction->tv_sec = tv.tv_sec; - *eaction->tv_usec = tv.tv_usec; + ktime_get_ts64(&ts); + /* monotonic time, so no y2038 overflow */ + *eaction->tv_sec = ts.tv_sec; + *eaction->tv_usec = ts.tv_nsec / NSEC_PER_USEC; }
drm_send_event_locked(dev, eaction->event);
Hi, Arnd,
Sinclair's on paternal leave and I thought this patch was already in drm-next. My bad. Dave, is it too late to pull this in for the next merge window?
/Thomas
On 01/16/2018 06:18 PM, Arnd Bergmann wrote:
DRM_VMW_EVENT_FENCE_SIGNALED (struct drm_vmw_event_fence) and DRM_EVENT_VBLANK (struct drm_event_vblank) pass timestamps in 32-bit seconds/microseconds format.
As of commit c61eef726a78 ("drm: add support for monotonic vblank timestamps"), other DRM drivers use monotonic times for drm_event_vblank, but vmwgfx still uses CLOCK_REALTIME for both events, which suffers from the y2038/y2106 overflow as well as time jumps.
For consistency, this changes vmwgfx to use ktime_get_ts64 as well, which solves those problems and avoids the deprecated do_gettimeofday() function.
This should be transparent to to user space, as long as it doesn't compare the time against the result of gettimeofday().
Link: https://urldefense.proofpoint.com/v2/url?u=https-3A__patchwork.kernel.org_pa... Signed-off-by: Arnd Bergmann arnd@arndb.de
Originally sent on Nov 27. Sinclair Yeh said he'd pick it up for the next pull request, but it's not in linux-next yet.
Resending the unchanged patch, please pick it up when you have time, or feel free to ignore this email in case it's already in some tree that just isn't part of linux-next but will be sent during the next merge window.
drivers/gpu/drm/vmwgfx/vmwgfx_fence.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c index 6c5c75cf5e6c..9ed544f8958f 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c @@ -901,11 +901,12 @@ static void vmw_event_fence_action_seq_passed(struct vmw_fence_action *action) spin_lock_irq(&dev->event_lock); if (likely(eaction->tv_sec != NULL)) {
struct timeval tv;
struct timespec64 ts;
do_gettimeofday(&tv);*eaction->tv_sec = tv.tv_sec;*eaction->tv_usec = tv.tv_usec;
ktime_get_ts64(&ts);/* monotonic time, so no y2038 overflow */*eaction->tv_sec = ts.tv_sec;*eaction->tv_usec = ts.tv_nsec / NSEC_PER_USEC;drm_send_event_locked(dev, eaction->event);
Thomas Hellstrom wrote:
Hi, Arnd,
Sinclair's on paternal leave and I thought this patch was already in drm-next. My bad. Dave, is it too late to pull this in for the next merge window?
/Thomas
On 01/16/2018 06:18 PM, Arnd Bergmann wrote:
DRM_VMW_EVENT_FENCE_SIGNALED (struct drm_vmw_event_fence) and DRM_EVENT_VBLANK (struct drm_event_vblank) pass timestamps in 32-bit seconds/microseconds format.
As of commit c61eef726a78 ("drm: add support for monotonic vblank timestamps"), other DRM drivers use monotonic times for drm_event_vblank, but vmwgfx still uses CLOCK_REALTIME for both events, which suffers from the y2038/y2106 overflow as well as time jumps.
For consistency, this changes vmwgfx to use ktime_get_ts64 as well, which solves those problems and avoids the deprecated do_gettimeofday() function.
This should be transparent to to user space, as long as it doesn't compare the time against the result of gettimeofday().
Link: https://urldefense.proofpoint.com/v2/url?u=https-3A__patchwork.kernel.org_pa... Signed-off-by: Arnd Bergmann arnd@arndb.de
Originally sent on Nov 27. Sinclair Yeh said he'd pick it up for the next pull request, but it's not in linux-next yet.
Resending the unchanged patch, please pick it up when you have time, or feel free to ignore this email in case it's already in some tree that just isn't part of linux-next but will be sent during the next merge window.
drivers/gpu/drm/vmwgfx/vmwgfx_fence.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c index 6c5c75cf5e6c..9ed544f8958f 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c @@ -901,11 +901,12 @@ static void vmw_event_fence_action_seq_passed(struct vmw_fence_action *action) spin_lock_irq(&dev->event_lock); if (likely(eaction->tv_sec != NULL)) { - struct timeval tv; + struct timespec64 ts; - do_gettimeofday(&tv); - *eaction->tv_sec = tv.tv_sec; - *eaction->tv_usec = tv.tv_usec; + ktime_get_ts64(&ts); + /* monotonic time, so no y2038 overflow */ + *eaction->tv_sec = ts.tv_sec; + *eaction->tv_usec = ts.tv_nsec / NSEC_PER_USEC; } drm_send_event_locked(dev, eaction->event);
Thomas, the same way my DRM patch has disappeared: Date
Tue, 19 Dec 2017 11:50:57 -0800
From Sinclair Yeh <>
Subject Re: [PATCH v.2] 4.15 vmgfx boot warning
This looks okay to me.
On Mon, Dec 18, 2017 at 07:26:03PM -0500, Woody Suwalski wrote:
The 4.15 drm_atomic_helper driver shows a warning during boot (both 32 and 64 bit x86) It is caused by a mismatch between the result of vmw_enable_vblank() and what the drm_atomic_helper expects: /... ret = drm_crtc_vblank_get(crtc); WARN_ONCE(ret != -EINVAL, "driver forgot to call drm_crtc_vblank_off()\n"); /...
Signed-off by: Woody Suwalski terraluna977@gmail.com
--- a/drivers/gpu/drm/drm_atomic_helper.c 2017-12-16 09:55:33.853374561 -0500 +++ b/drivers/gpu/drm/drm_atomic_helper.c 2017-12-16 10:55:56.089090752 -0500 @@ -889,7 +889,7 @@ disable_outputs(struct drm_device *dev, continue;
ret = drm_crtc_vblank_get(crtc); - WARN_ONCE(ret != -EINVAL, "driver forgot to call drm_crtc_vblank_off()\n"); + WARN_ONCE((ret != -EINVAL && ret != -ENOSYS), "driver forgot to call drm_crtc_vblank_off()\n"); if (ret == 0) drm_crtc_vblank_put(crtc); }
===============================================
The 4.15 drm_atomic_helper driver shows a warning during boot. It is caused by a mismatch between the result of vmw_enable_vblank() and what the drm_atomic_helper expects: /... ret = drm_crtc_vblank_get(crtc); WARN_ONCE(ret != -EINVAL, "driver forgot to call drm_crtc_vblank_off()\n"); /...
Signed-off by: Woody Suwalski terraluna977@gmail.com
--- a/drivers/gpu/drm/drm_atomic_helper.c 2017-12-16 09:55:33.853374561 -0500 +++ b/drivers/gpu/drm/drm_atomic_helper.c 2017-12-16 10:55:56.089090752 -0500 @@ -889,7 +889,7 @@ disable_outputs(struct drm_device *dev, continue;
ret = drm_crtc_vblank_get(crtc); - WARN_ONCE(ret != -EINVAL, "driver forgot to call drm_crtc_vblank_off()\n"); + WARN_ONCE((ret != -EINVAL && ret != -ENOSYS), "driver forgot to call drm_crtc_vblank_off()\n"); if (ret == 0) drm_crtc_vblank_put(crtc); }
On Tue, Jan 16, 2018 at 6:28 PM, Thomas Hellstrom thellstrom@vmware.com wrote:
Hi, Arnd,
Sinclair's on paternal leave and I thought this patch was already in drm-next. My bad.
No worries.
Dave, is it too late to pull this in for the next merge window?
The patch isn't urgent, it's fine to wait until after the merge window, I just want to avoid having to send it once more as I make my way through the remaining 'timeval' and 'timespec' users.
Arnd
Hi, Woody,
On 01/16/2018 10:39 PM, Woody Suwalski wrote:
Thomas, the same way my DRM patch has disappeared: Date
Tue, 19 Dec 2017 11:50:57 -0800 From Sinclair Yeh <> Subject Re: [PATCH v.2] 4.15 vmgfx boot warning
This looks okay to me.
Since this is a core DRM patch I think Sinclair was expecting it to be pulled in using the drm-misc tree. Did you get a comment from Daniel on this patch?
Thanks, Thomas
On 01/17/2018 07:33 AM, Thomas Hellstrom wrote:
Hi, Woody,
On 01/16/2018 10:39 PM, Woody Suwalski wrote:
Thomas, the same way my DRM patch has disappeared: Date
Tue, 19 Dec 2017 11:50:57 -0800 From Sinclair Yeh <> Subject Re: [PATCH v.2] 4.15 vmgfx boot warning
This looks okay to me.
Since this is a core DRM patch I think Sinclair was expecting it to be pulled in using the drm-misc tree. Did you get a comment from Daniel on this patch?
Thanks, Thomas
Actually, checkpatch.pl complains about -ENOSYS, so I'm sending a pull request for v1 of this patch instead, to make everybody happy. I've rephrased the commit log somewhat. Hope that's OK.
/Thomas
-
Arnd Bergmann -
Thomas Hellstrom -
Thomas Hellstrom -
Woody Suwalski