Re: [PATCH v9 8/8] drm/gpusvm: Use hmm_range_fault_unlocked_timeout() for range faults

[Stanislav Kinsburskii]

On Mon, Jul 20, 2026 at 05:49:28PM -0700, Matthew Brost wrote:
> On Wed, Jul 15, 2026 at 11:16:52AM -0700, Stanislav Kinsburskii wrote:
> > Several GPU SVM paths take mmap_read_lock() only to call hmm_range_fault()
> > and open-code mmu interval sequence setup before each HMM walk. They also
> > retry -EBUSY until HMM_RANGE_DEFAULT_TIMEOUT expires.
> > 
> > Use hmm_range_fault_unlocked_timeout() for those faults. The HMM helper now
> > owns mmap_lock acquisition and refreshes range->notifier_seq for its
> > internal retries, while GPU SVM keeps its existing driver-lock validation
> > with mmu_interval_read_retry() after a successful fault.
> > 
> > Pass HMM_RANGE_DEFAULT_TIMEOUT as the helper retry budget for each HMM
> > fault attempt. This scopes the timeout to repeated HMM notifier retries
> > while preserving the outer retry loops that restart when the interval is
> > invalidated before GPU SVM updates or consumes the mapping state.
> > 
> 
> This part doesn't seem right for get_pages(), see below.
> 
> > Leave drm_gpusvm_check_pages() on hmm_range_fault() because that path is
> > called with the mmap lock already held by its caller.
> > 
> > Reviewed-by: Jason Gunthorpe <jgg@xxxxxxxxxx>
> > Signed-off-by: Stanislav Kinsburskii <skinsburskii@xxxxxxxxx>
> > ---
> >  drivers/gpu/drm/drm_gpusvm.c |   61 +++++-------------------------------------
> >  1 file changed, 7 insertions(+), 54 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/drm_gpusvm.c b/drivers/gpu/drm/drm_gpusvm.c
> > index 958cb605aedd..de5bbfe58ee9 100644
> > --- a/drivers/gpu/drm/drm_gpusvm.c
> > +++ b/drivers/gpu/drm/drm_gpusvm.c
> > @@ -773,8 +773,7 @@ enum drm_gpusvm_scan_result drm_gpusvm_scan_mm(struct drm_gpusvm_range *range,
> >  		.end = end,
> >  		.dev_private_owner = dev_private_owner,
> >  	};
> > -	unsigned long timeout =
> > -		jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
> > +	unsigned long timeout = msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
> >  	enum drm_gpusvm_scan_result state = DRM_GPUSVM_SCAN_UNPOPULATED, new_state;
> >  	unsigned long *pfns;
> >  	unsigned long npages = npages_in_range(start, end);
> > @@ -788,22 +787,7 @@ enum drm_gpusvm_scan_result drm_gpusvm_scan_mm(struct drm_gpusvm_range *range,
> >  	hmm_range.hmm_pfns = pfns;
> >  
> >  retry:
> > -	hmm_range.notifier_seq = mmu_interval_read_begin(notifier);
> > -	mmap_read_lock(range->gpusvm->mm);
> > -
> > -	while (true) {
> > -		err = hmm_range_fault(&hmm_range);
> > -		if (err == -EBUSY) {
> > -			if (time_after(jiffies, timeout))
> > -				break;
> > -
> > -			hmm_range.notifier_seq =
> > -				mmu_interval_read_begin(notifier);
> > -			continue;
> > -		}
> > -		break;
> > -	}
> > -	mmap_read_unlock(range->gpusvm->mm);
> > +	err = hmm_range_fault_unlocked_timeout(&hmm_range, timeout);
> >  	if (err)
> >  		goto err_free;
> >  
> > @@ -1406,8 +1390,7 @@ int drm_gpusvm_get_pages(struct drm_gpusvm *gpusvm,
> >  		.dev_private_owner = ctx->device_private_page_owner,
> >  	};
> >  	void *zdd;
> > -	unsigned long timeout =
> > -		jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
> > +	unsigned long timeout = msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
> >  	unsigned long i, j;
> >  	unsigned long npages = npages_in_range(pages_start, pages_end);
> >  	unsigned long num_dma_mapped;
> > @@ -1422,9 +1405,6 @@ int drm_gpusvm_get_pages(struct drm_gpusvm *gpusvm,
> >  	struct dma_iova_state *state = &svm_pages->state;
> >  
> >  retry:
> > -	if (time_after(jiffies, timeout))
> > -		return -EBUSY;
> > -
> 
> I think that by deleting the code above, you have changed this function's
> semantics by removing the hard cap of HMM_RANGE_DEFAULT_TIMEOUT. This
> code was added because, on some non-production platforms, the timing in
> this function could cause it to livelock.
> 
> Is there any reason this was remove aside from timeout variable not
> being a deadline now? You likely should add the deadline back in.
> 

Indeed, this one can be called from a kernel thread context as well.
I'll revert the change in the next revision.

Thanks,
Stanislav


> Matt
> 
> >  	hmm_range.notifier_seq = mmu_interval_read_begin(notifier);
> >  	if (drm_gpusvm_pages_valid_unlocked(gpusvm, svm_pages))
> >  		goto set_seqno;
> > @@ -1439,21 +1419,7 @@ int drm_gpusvm_get_pages(struct drm_gpusvm *gpusvm,
> >  	}
> >  
> >  	hmm_range.hmm_pfns = pfns;
> > -	while (true) {
> > -		mmap_read_lock(mm);
> > -		err = hmm_range_fault(&hmm_range);
> > -		mmap_read_unlock(mm);
> > -
> > -		if (err == -EBUSY) {
> > -			if (time_after(jiffies, timeout))
> > -				break;
> > -
> > -			hmm_range.notifier_seq =
> > -				mmu_interval_read_begin(notifier);
> > -			continue;
> > -		}
> > -		break;
> > -	}
> > +	err = hmm_range_fault_unlocked_timeout(&hmm_range, timeout);
> >  	mmput(mm);
> >  	if (err)
> >  		goto err_free;
> > @@ -1720,8 +1686,7 @@ int drm_gpusvm_range_evict(struct drm_gpusvm *gpusvm,
> >  		.end = drm_gpusvm_range_end(range),
> >  		.dev_private_owner = NULL,
> >  	};
> > -	unsigned long timeout =
> > -		jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
> > +	unsigned long timeout = msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
> >  	unsigned long *pfns;
> >  	unsigned long npages = npages_in_range(drm_gpusvm_range_start(range),
> >  					       drm_gpusvm_range_end(range));
> > @@ -1736,24 +1701,12 @@ int drm_gpusvm_range_evict(struct drm_gpusvm *gpusvm,
> >  		return -ENOMEM;
> >  
> >  	hmm_range.hmm_pfns = pfns;
> > -	while (!time_after(jiffies, timeout)) {
> > -		hmm_range.notifier_seq = mmu_interval_read_begin(notifier);
> > -		if (time_after(jiffies, timeout)) {
> > -			err = -ETIME;
> > -			break;
> > -		}
> > -
> > -		mmap_read_lock(mm);
> > -		err = hmm_range_fault(&hmm_range);
> > -		mmap_read_unlock(mm);
> > -		if (err != -EBUSY)
> > -			break;
> > -	}
> > +	err = hmm_range_fault_unlocked_timeout(&hmm_range, timeout);
> >  
> >  	kvfree(pfns);
> >  	mmput(mm);
> >  
> > -	return err;
> > +	return err == -EBUSY ? -ETIME : err;
> >  }
> >  EXPORT_SYMBOL_GPL(drm_gpusvm_range_evict);
> >  
> > 
> >