Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
To: Matthew Brost <matthew.brost@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@kernel.org>, intel-xe@lists.freedesktop.org
Subject: Re: [Intel-xe] [PATCH v2 28/31] drm/xe: Allow dma-fences as in-syncs for compute / faulting VM
Date: Mon, 8 May 2023 17:29:40 -0400	[thread overview]
Message-ID: <ZFlpxCMh4S32Rlmm@intel.com> (raw)
In-Reply-To: <ZFhONAGmO8X+qD1f@DUT025-TGLU.fm.intel.com>

On Mon, May 08, 2023 at 01:19:48AM +0000, Matthew Brost wrote:
> On Fri, May 05, 2023 at 03:43:15PM -0400, Rodrigo Vivi wrote:
> > On Mon, May 01, 2023 at 05:17:24PM -0700, Matthew Brost wrote:
> > > This is allowed per the dma-fencing rules.
> > 
> > it would be good a word saying 'why' we are doing this.
> > only because we can it doesn't mean we should...
> > 
> 
> This is allowed and encouraged the dma-fencing rule. This along with
> allowing compute VMs to export dma-fences on binds will result in a
> simpler compute UMD.
> 
> Sound ok?

makes sense. with that

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>


> 
> Matt
>   
> > > 
> > > Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> > > ---
> > >  drivers/gpu/drm/xe/xe_sync.c | 12 +++++++-----
> > >  1 file changed, 7 insertions(+), 5 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/xe/xe_sync.c b/drivers/gpu/drm/xe/xe_sync.c
> > > index 99f1ed87196d..1e4e4acb2c4a 100644
> > > --- a/drivers/gpu/drm/xe/xe_sync.c
> > > +++ b/drivers/gpu/drm/xe/xe_sync.c
> > > @@ -105,6 +105,7 @@ int xe_sync_entry_parse(struct xe_device *xe, struct xe_file *xef,
> > >  {
> > >  	struct drm_xe_sync sync_in;
> > >  	int err;
> > > +	bool signal;
> > >  
> > >  	if (copy_from_user(&sync_in, sync_user, sizeof(*sync_user)))
> > >  		return -EFAULT;
> > > @@ -113,9 +114,10 @@ int xe_sync_entry_parse(struct xe_device *xe, struct xe_file *xef,
> > >  			 ~(SYNC_FLAGS_TYPE_MASK | DRM_XE_SYNC_SIGNAL)))
> > >  		return -EINVAL;
> > >  
> > > +	signal = sync_in.flags & DRM_XE_SYNC_SIGNAL;
> > >  	switch (sync_in.flags & SYNC_FLAGS_TYPE_MASK) {
> > >  	case DRM_XE_SYNC_SYNCOBJ:
> > > -		if (XE_IOCTL_ERR(xe, no_dma_fences))
> > > +		if (XE_IOCTL_ERR(xe, no_dma_fences && signal))
> > >  			return -ENOTSUPP;
> > >  
> > >  		if (XE_IOCTL_ERR(xe, upper_32_bits(sync_in.addr)))
> > > @@ -125,7 +127,7 @@ int xe_sync_entry_parse(struct xe_device *xe, struct xe_file *xef,
> > >  		if (XE_IOCTL_ERR(xe, !sync->syncobj))
> > >  			return -ENOENT;
> > >  
> > > -		if (!(sync_in.flags & DRM_XE_SYNC_SIGNAL)) {
> > > +		if (!signal) {
> > >  			sync->fence = drm_syncobj_fence_get(sync->syncobj);
> > >  			if (XE_IOCTL_ERR(xe, !sync->fence))
> > >  				return -EINVAL;
> > > @@ -133,7 +135,7 @@ int xe_sync_entry_parse(struct xe_device *xe, struct xe_file *xef,
> > >  		break;
> > >  
> > >  	case DRM_XE_SYNC_TIMELINE_SYNCOBJ:
> > > -		if (XE_IOCTL_ERR(xe, no_dma_fences))
> > > +		if (XE_IOCTL_ERR(xe, no_dma_fences && signal))
> > >  			return -ENOTSUPP;
> > >  
> > >  		if (XE_IOCTL_ERR(xe, upper_32_bits(sync_in.addr)))
> > > @@ -146,7 +148,7 @@ int xe_sync_entry_parse(struct xe_device *xe, struct xe_file *xef,
> > >  		if (XE_IOCTL_ERR(xe, !sync->syncobj))
> > >  			return -ENOENT;
> > >  
> > > -		if (sync_in.flags & DRM_XE_SYNC_SIGNAL) {
> > > +		if (signal) {
> > >  			sync->chain_fence = dma_fence_chain_alloc();
> > >  			if (!sync->chain_fence)
> > >  				return -ENOMEM;
> > > @@ -168,7 +170,7 @@ int xe_sync_entry_parse(struct xe_device *xe, struct xe_file *xef,
> > >  		break;
> > >  
> > >  	case DRM_XE_SYNC_USER_FENCE:
> > > -		if (XE_IOCTL_ERR(xe, !(sync_in.flags & DRM_XE_SYNC_SIGNAL)))
> > > +		if (XE_IOCTL_ERR(xe, !signal))
> > >  			return -ENOTSUPP;
> > >  
> > >  		if (XE_IOCTL_ERR(xe, sync_in.addr & 0x7))
> > > -- 
> > > 2.34.1
> > > 

  reply	other threads:[~2023-05-08 21:29 UTC|newest]

Thread overview: 126+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-05-02  0:16 [Intel-xe] [PATCH v2 00/31] Upstreaming prep / all of mbrosts patches Matthew Brost
2023-05-02  0:16 ` [Intel-xe] [PATCH v2 01/31] drm/sched: Add run_wq argument to drm_sched_init Matthew Brost
2023-05-03 12:03   ` Thomas Hellström
2023-05-03 15:06     ` Matthew Brost
2023-05-05 18:24       ` Rodrigo Vivi
2023-05-02  0:16 ` [Intel-xe] [PATCH v2 02/31] drm/sched: Move schedule policy to scheduler Matthew Brost
2023-05-03 12:13   ` Thomas Hellström
2023-05-03 15:11     ` Matthew Brost
2023-05-02  0:16 ` [Intel-xe] [PATCH v2 03/31] drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy Matthew Brost
2023-05-08 12:40   ` Thomas Hellström
2023-05-22  1:16     ` Matthew Brost
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 04/31] drm/xe: Use DRM_SCHED_POLICY_SINGLE_ENTITY mode Matthew Brost
2023-05-08 12:41   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 05/31] drm/xe: Long running job update Matthew Brost
2023-05-05 18:36   ` Rodrigo Vivi
2023-05-08  1:14     ` Matthew Brost
2023-05-08 13:14   ` Thomas Hellström
2023-05-09 14:56     ` Matthew Brost
2023-05-09 15:21       ` Thomas Hellström
2023-05-09 22:16         ` Matthew Brost
2023-05-10  8:15           ` Thomas Hellström
2023-05-09 22:21     ` Matthew Brost
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 06/31] drm/xe: Ensure LR engines are not persistent Matthew Brost
2023-05-05 18:38   ` Rodrigo Vivi
2023-05-08  1:03     ` Matthew Brost
2023-05-09 12:21   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 07/31] drm/xe: Only try to lock external BOs in VM bind Matthew Brost
2023-05-05 18:40   ` Rodrigo Vivi
2023-05-08  1:08     ` Matthew Brost
2023-05-08  1:15       ` Christopher Snowhill
2023-05-08 21:34       ` Rodrigo Vivi
2023-05-09 12:29         ` Thomas Hellström
2023-05-10 23:25           ` Matthew Brost
2023-05-11  7:43             ` Thomas Hellström
2023-05-08  1:17   ` Christopher Snowhill
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 08/31] drm/xe: VM LRU bulk move Matthew Brost
2023-05-08 21:39   ` Rodrigo Vivi
2023-05-09 22:09     ` Matthew Brost
2023-05-10  1:37       ` Rodrigo Vivi
2023-05-09 12:47   ` Thomas Hellström
2023-05-09 22:05     ` Matthew Brost
2023-05-10  8:14       ` Thomas Hellström
2023-05-10 18:40         ` Matthew Brost
2023-05-11  7:24           ` Thomas Hellström
2023-05-11 14:11             ` Matthew Brost
2023-05-12  9:03               ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 09/31] drm/xe/guc: Read HXG fields from DW1 of G2H response Matthew Brost
2023-05-05 18:50   ` Rodrigo Vivi
2023-05-09 12:49   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 10/31] drm/xe/guc: Return the lower part of blocking H2G message Matthew Brost
2023-05-05 18:52   ` Rodrigo Vivi
2023-05-08  1:10     ` Matthew Brost
2023-05-08  9:20       ` Michal Wajdeczko
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 11/31] drm/xe/guc: Use doorbells for submission if possible Matthew Brost
2023-05-08 21:42   ` Rodrigo Vivi
2023-05-10  0:49     ` Matthew Brost
2023-05-09 13:00   ` Thomas Hellström
2023-05-10  0:51     ` Matthew Brost
2023-05-21 12:32   ` Oded Gabbay
2023-06-08 19:30     ` Matthew Brost
2023-06-12 13:01       ` Oded Gabbay
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 12/31] drm/xe/guc: Print doorbell ID in GuC engine debugfs entry Matthew Brost
2023-05-05 18:55   ` Rodrigo Vivi
2023-05-09 13:01     ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 13/31] maple_tree: split up MA_STATE() macro Matthew Brost
2023-05-09 13:21   ` Thomas Hellström
2023-05-10  0:29     ` Matthew Brost
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 14/31] maple_tree: Export mas_preallocate Matthew Brost
2023-05-09 13:33   ` Thomas Hellström
2023-05-10  0:31     ` Matthew Brost
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 15/31] drm: manager to keep track of GPUs VA mappings Matthew Brost
2023-05-09 13:49   ` Thomas Hellström
2023-05-10  0:55     ` Matthew Brost
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 16/31] drm/xe: Port Xe to GPUVA Matthew Brost
2023-05-09 13:52   ` Thomas Hellström
2023-05-11  2:41     ` Matthew Brost
2023-05-11  7:39       ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 17/31] drm/xe: NULL binding implementation Matthew Brost
2023-05-09 14:34   ` Rodrigo Vivi
2023-05-11  2:52     ` Matthew Brost
2023-05-09 15:17   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 18/31] drm/xe: Avoid doing rebinds Matthew Brost
2023-05-09 14:48   ` Rodrigo Vivi
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 19/31] drm/xe: Reduce the number list links in xe_vma Matthew Brost
2023-05-08 21:43   ` Rodrigo Vivi
2023-05-11  8:38   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 20/31] drm/xe: Optimize size of xe_vma allocation Matthew Brost
2023-05-05 19:37   ` Rodrigo Vivi
2023-05-08  1:21     ` Matthew Brost
2023-05-11  9:05   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 21/31] drm/gpuva: Add drm device to GPUVA manager Matthew Brost
2023-05-05 19:39   ` Rodrigo Vivi
2023-05-11  9:06     ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 22/31] drm/gpuva: Move dma-resv " Matthew Brost
2023-05-11  9:10   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 23/31] drm/gpuva: Add support for extobj Matthew Brost
2023-05-11  9:35   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 24/31] drm/xe: Userptr refactor Matthew Brost
2023-05-05 19:41   ` Rodrigo Vivi
2023-05-11  9:46   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 25/31] drm: execution context for GEM buffers v3 Matthew Brost
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 26/31] drm/exec: Always compile drm_exec Matthew Brost
2023-05-09 14:45   ` Rodrigo Vivi
2023-05-10  0:37     ` Matthew Brost
2023-05-10  0:38     ` Matthew Brost
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 27/31] drm/xe: Use drm_exec for locking rather than TTM exec helpers Matthew Brost
2023-05-05 19:42   ` Rodrigo Vivi
2023-05-11 10:01   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 28/31] drm/xe: Allow dma-fences as in-syncs for compute / faulting VM Matthew Brost
2023-05-05 19:43   ` Rodrigo Vivi
2023-05-08  1:19     ` Matthew Brost
2023-05-08 21:29       ` Rodrigo Vivi [this message]
2023-05-11 10:03   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 29/31] drm/xe: Allow compute VMs to output dma-fences on binds Matthew Brost
2023-05-09 14:50   ` Rodrigo Vivi
2023-05-11 10:04   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 30/31] drm/xe: remove async worker, sync binds, new error handling Matthew Brost
2023-05-17 16:53   ` Thomas Hellström
2023-05-02  0:17 ` [Intel-xe] [PATCH v2 31/31] drm/xe/uapi: Add some VM bind kernel doc Matthew Brost
2023-05-05 19:45   ` Rodrigo Vivi
2023-05-11 10:14     ` Thomas Hellström
2023-05-02  0:20 ` [Intel-xe] ✗ CI.Patch_applied: failure for Upstreaming prep / all of mbrosts patches (rev2) Patchwork
2023-05-02  1:54   ` Christopher Snowhill (kode54)
2023-05-02  1:59   ` Christopher Snowhill (kode54)
2023-05-03 12:37 ` [Intel-xe] [PATCH v2 00/31] Upstreaming prep / all of mbrosts patches Thomas Hellström
2023-05-03 15:27   ` Matthew Brost

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=ZFlpxCMh4S32Rlmm@intel.com \
    --to=rodrigo.vivi@intel.com \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=matthew.brost@intel.com \
    --cc=rodrigo.vivi@kernel.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