Linux IOMMU Development
 help / color / mirror / Atom feed
* [PATCH 0/2] Two fixes to vevent path
@ 2025-03-24 12:00 Yi Liu
  2025-03-24 12:00 ` [PATCH 1/2] iommufd: Initialize the flags of vevent in iommufd_viommu_report_event() Yi Liu
                   ` (2 more replies)
  0 siblings, 3 replies; 6+ messages in thread
From: Yi Liu @ 2025-03-24 12:00 UTC (permalink / raw)
  To: jgg, kevin.tian; +Cc: nicolinc, yi.l.liu, baolu.lu, iommu

iommufd selftest occassionally fail when reading vevent. After debugging,
it is due to vevent->header.flags. It is not set properly per allocation.
This can be fixed by using kzalloc(). There is another related issue. The
veventq->num_events inc/dec is not balanced. This is problemtic. Make a
fix for it as well.

Regards,
	Yi Liu

Yi Liu (2):
  iommufd: Initialize the flags of vevent in
    iommufd_viommu_report_event()
  iommufd: Balance veventq->num_events inc/dec

 drivers/iommu/iommufd/driver.c | 2 +-
 drivers/iommu/iommufd/eventq.c | 3 ++-
 2 files changed, 3 insertions(+), 2 deletions(-)

-- 
2.34.1


^ permalink raw reply	[flat|nested] 6+ messages in thread

* [PATCH 1/2] iommufd: Initialize the flags of vevent in iommufd_viommu_report_event()
  2025-03-24 12:00 [PATCH 0/2] Two fixes to vevent path Yi Liu
@ 2025-03-24 12:00 ` Yi Liu
  2025-03-24 14:04   ` Nicolin Chen
  2025-03-24 12:00 ` [PATCH 2/2] iommufd: Balance veventq->num_events inc/dec Yi Liu
  2025-03-25 13:35 ` [PATCH 0/2] Two fixes to vevent path Jason Gunthorpe
  2 siblings, 1 reply; 6+ messages in thread
From: Yi Liu @ 2025-03-24 12:00 UTC (permalink / raw)
  To: jgg, kevin.tian; +Cc: nicolinc, yi.l.liu, baolu.lu, iommu

The vevent->header.flags is not initialized per allocation, hence the
vevent read path may treat the vevent as lost_events_header wrongly.
Use kzalloc() to alloc memory for new vevent.

Fixes: e8e1ef9b77a7 ("iommufd/viommu: Add iommufd_viommu_report_event helper")
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
---
 drivers/iommu/iommufd/driver.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/iommu/iommufd/driver.c b/drivers/iommu/iommufd/driver.c
index 352513974154..81e01d85b7d4 100644
--- a/drivers/iommu/iommufd/driver.c
+++ b/drivers/iommu/iommufd/driver.c
@@ -102,7 +102,7 @@ int iommufd_viommu_report_event(struct iommufd_viommu *viommu,
 		goto out_set_header;
 	}
 
-	vevent = kmalloc(struct_size(vevent, event_data, data_len), GFP_ATOMIC);
+	vevent = kzalloc(struct_size(vevent, event_data, data_len), GFP_ATOMIC);
 	if (!vevent) {
 		rc = -ENOMEM;
 		vevent = &veventq->lost_events_header;
-- 
2.34.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread

* [PATCH 2/2] iommufd: Balance veventq->num_events inc/dec
  2025-03-24 12:00 [PATCH 0/2] Two fixes to vevent path Yi Liu
  2025-03-24 12:00 ` [PATCH 1/2] iommufd: Initialize the flags of vevent in iommufd_viommu_report_event() Yi Liu
@ 2025-03-24 12:00 ` Yi Liu
  2025-03-24 14:07   ` Nicolin Chen
  2025-03-25 13:35 ` [PATCH 0/2] Two fixes to vevent path Jason Gunthorpe
  2 siblings, 1 reply; 6+ messages in thread
From: Yi Liu @ 2025-03-24 12:00 UTC (permalink / raw)
  To: jgg, kevin.tian; +Cc: nicolinc, yi.l.liu, baolu.lu, iommu

iommufd_veventq_fops_read() decrements veventq->num_events when a vevent
is read out. However, the report path ony increments veventq->num_events
for normal events. To be balanced, make the read path decrement num_events
only for normal vevents.

Fixes: e36ba5ab808e ("iommufd: Add IOMMUFD_OBJ_VEVENTQ and IOMMUFD_CMD_VEVENTQ_ALLOC")
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
---
 drivers/iommu/iommufd/eventq.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/iommu/iommufd/eventq.c b/drivers/iommu/iommufd/eventq.c
index 4c43ace8c725..f39cf0797347 100644
--- a/drivers/iommu/iommufd/eventq.c
+++ b/drivers/iommu/iommufd/eventq.c
@@ -385,7 +385,8 @@ static ssize_t iommufd_veventq_fops_read(struct file *filep, char __user *buf,
 			break;
 		}
 		spin_lock(&eventq->lock);
-		veventq->num_events--;
+		if (!vevent_for_lost_events_header(cur))
+			veventq->num_events--;
 		spin_unlock(&eventq->lock);
 		done += cur->data_len;
 		kfree(cur);
-- 
2.34.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread

* Re: [PATCH 1/2] iommufd: Initialize the flags of vevent in iommufd_viommu_report_event()
  2025-03-24 12:00 ` [PATCH 1/2] iommufd: Initialize the flags of vevent in iommufd_viommu_report_event() Yi Liu
@ 2025-03-24 14:04   ` Nicolin Chen
  0 siblings, 0 replies; 6+ messages in thread
From: Nicolin Chen @ 2025-03-24 14:04 UTC (permalink / raw)
  To: Yi Liu; +Cc: jgg, kevin.tian, baolu.lu, iommu

On Mon, Mar 24, 2025 at 05:00:33AM -0700, Yi Liu wrote:
> The vevent->header.flags is not initialized per allocation, hence the
> vevent read path may treat the vevent as lost_events_header wrongly.
> Use kzalloc() to alloc memory for new vevent.
> 
> Fixes: e8e1ef9b77a7 ("iommufd/viommu: Add iommufd_viommu_report_event helper")
> Signed-off-by: Yi Liu <yi.l.liu@intel.com>

Reviewed-by: Nicolin Chen <nicolinc@nvidia.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>

Thanks!

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH 2/2] iommufd: Balance veventq->num_events inc/dec
  2025-03-24 12:00 ` [PATCH 2/2] iommufd: Balance veventq->num_events inc/dec Yi Liu
@ 2025-03-24 14:07   ` Nicolin Chen
  0 siblings, 0 replies; 6+ messages in thread
From: Nicolin Chen @ 2025-03-24 14:07 UTC (permalink / raw)
  To: Yi Liu; +Cc: jgg, kevin.tian, baolu.lu, iommu

On Mon, Mar 24, 2025 at 05:00:34AM -0700, Yi Liu wrote:
> iommufd_veventq_fops_read() decrements veventq->num_events when a vevent
> is read out. However, the report path ony increments veventq->num_events
> for normal events. To be balanced, make the read path decrement num_events
> only for normal vevents.
> 
> Fixes: e36ba5ab808e ("iommufd: Add IOMMUFD_OBJ_VEVENTQ and IOMMUFD_CMD_VEVENTQ_ALLOC")
> Signed-off-by: Yi Liu <yi.l.liu@intel.com>
 
Reviewed-by: Nicolin Chen <nicolinc@nvidia.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>

Thanks!

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH 0/2] Two fixes to vevent path
  2025-03-24 12:00 [PATCH 0/2] Two fixes to vevent path Yi Liu
  2025-03-24 12:00 ` [PATCH 1/2] iommufd: Initialize the flags of vevent in iommufd_viommu_report_event() Yi Liu
  2025-03-24 12:00 ` [PATCH 2/2] iommufd: Balance veventq->num_events inc/dec Yi Liu
@ 2025-03-25 13:35 ` Jason Gunthorpe
  2 siblings, 0 replies; 6+ messages in thread
From: Jason Gunthorpe @ 2025-03-25 13:35 UTC (permalink / raw)
  To: Yi Liu; +Cc: kevin.tian, nicolinc, baolu.lu, iommu

On Mon, Mar 24, 2025 at 05:00:32AM -0700, Yi Liu wrote:

> Yi Liu (2):
>   iommufd: Initialize the flags of vevent in
>     iommufd_viommu_report_event()
>   iommufd: Balance veventq->num_events inc/dec

Applied, thanks

Jason

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2025-03-25 13:35 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-03-24 12:00 [PATCH 0/2] Two fixes to vevent path Yi Liu
2025-03-24 12:00 ` [PATCH 1/2] iommufd: Initialize the flags of vevent in iommufd_viommu_report_event() Yi Liu
2025-03-24 14:04   ` Nicolin Chen
2025-03-24 12:00 ` [PATCH 2/2] iommufd: Balance veventq->num_events inc/dec Yi Liu
2025-03-24 14:07   ` Nicolin Chen
2025-03-25 13:35 ` [PATCH 0/2] Two fixes to vevent path Jason Gunthorpe

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox