* [PATCH] cxl/pmem: Fix reference counting for delayed work
@ 2021-10-29 19:55 Dan Williams
2021-10-31 18:58 ` Ben Widawsky
2021-11-01 11:30 ` Jonathan Cameron
0 siblings, 2 replies; 5+ messages in thread
From: Dan Williams @ 2021-10-29 19:55 UTC (permalink / raw)
To: linux-cxl; +Cc: ira.weiny, ben.widawsky, vishal.l.verma, alison.schofield
There is a potential race between queue_work() returning and the
queued-work running that could result in put_device() running before
get_device(). Introduce the cxl_nvdimm_bridge_state_work() helper that
takes the reference unconditionally, but drops it if no new work was
queued, to keep the references balanced.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
drivers/cxl/pmem.c | 17 +++++++++++++----
1 file changed, 13 insertions(+), 4 deletions(-)
diff --git a/drivers/cxl/pmem.c b/drivers/cxl/pmem.c
index ceb2115981e5..38bcbb4e9409 100644
--- a/drivers/cxl/pmem.c
+++ b/drivers/cxl/pmem.c
@@ -266,14 +266,24 @@ static void cxl_nvb_update_state(struct work_struct *work)
put_device(&cxl_nvb->dev);
}
+static void cxl_nvdimm_bridge_state_work(struct cxl_nvdimm_bridge *cxl_nvb)
+{
+ /*
+ * Take a reference that the workqueue will drop if new work
+ * gets queued.
+ */
+ get_device(&cxl_nvb->dev);
+ if (!queue_work(cxl_pmem_wq, &cxl_nvb->state_work))
+ put_device(&cxl_nvb->dev);
+}
+
static void cxl_nvdimm_bridge_remove(struct device *dev)
{
struct cxl_nvdimm_bridge *cxl_nvb = to_cxl_nvdimm_bridge(dev);
if (cxl_nvb->state == CXL_NVB_ONLINE)
cxl_nvb->state = CXL_NVB_OFFLINE;
- if (queue_work(cxl_pmem_wq, &cxl_nvb->state_work))
- get_device(&cxl_nvb->dev);
+ cxl_nvdimm_bridge_state_work(cxl_nvb);
}
static int cxl_nvdimm_bridge_probe(struct device *dev)
@@ -294,8 +304,7 @@ static int cxl_nvdimm_bridge_probe(struct device *dev)
}
cxl_nvb->state = CXL_NVB_ONLINE;
- if (queue_work(cxl_pmem_wq, &cxl_nvb->state_work))
- get_device(&cxl_nvb->dev);
+ cxl_nvdimm_bridge_state_work(cxl_nvb);
return 0;
}
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] cxl/pmem: Fix reference counting for delayed work
2021-10-29 19:55 [PATCH] cxl/pmem: Fix reference counting for delayed work Dan Williams
@ 2021-10-31 18:58 ` Ben Widawsky
2021-10-31 19:27 ` Dan Williams
2021-11-01 11:30 ` Jonathan Cameron
1 sibling, 1 reply; 5+ messages in thread
From: Ben Widawsky @ 2021-10-31 18:58 UTC (permalink / raw)
To: Dan Williams; +Cc: linux-cxl, ira.weiny, vishal.l.verma, alison.schofield
On 21-10-29 12:55:47, Dan Williams wrote:
> There is a potential race between queue_work() returning and the
> queued-work running that could result in put_device() running before
> get_device(). Introduce the cxl_nvdimm_bridge_state_work() helper that
> takes the reference unconditionally, but drops it if no new work was
> queued, to keep the references balanced.
>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Arguably fixes/stable?
Reviewed-by: Ben Widawsky <ben.widawsky@intel.com>
> ---
> drivers/cxl/pmem.c | 17 +++++++++++++----
> 1 file changed, 13 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/cxl/pmem.c b/drivers/cxl/pmem.c
> index ceb2115981e5..38bcbb4e9409 100644
> --- a/drivers/cxl/pmem.c
> +++ b/drivers/cxl/pmem.c
> @@ -266,14 +266,24 @@ static void cxl_nvb_update_state(struct work_struct *work)
> put_device(&cxl_nvb->dev);
> }
>
> +static void cxl_nvdimm_bridge_state_work(struct cxl_nvdimm_bridge *cxl_nvb)
> +{
> + /*
> + * Take a reference that the workqueue will drop if new work
> + * gets queued.
> + */
> + get_device(&cxl_nvb->dev);
> + if (!queue_work(cxl_pmem_wq, &cxl_nvb->state_work))
> + put_device(&cxl_nvb->dev);
> +}
> +
> static void cxl_nvdimm_bridge_remove(struct device *dev)
> {
> struct cxl_nvdimm_bridge *cxl_nvb = to_cxl_nvdimm_bridge(dev);
>
> if (cxl_nvb->state == CXL_NVB_ONLINE)
> cxl_nvb->state = CXL_NVB_OFFLINE;
> - if (queue_work(cxl_pmem_wq, &cxl_nvb->state_work))
> - get_device(&cxl_nvb->dev);
> + cxl_nvdimm_bridge_state_work(cxl_nvb);
> }
>
> static int cxl_nvdimm_bridge_probe(struct device *dev)
> @@ -294,8 +304,7 @@ static int cxl_nvdimm_bridge_probe(struct device *dev)
> }
>
> cxl_nvb->state = CXL_NVB_ONLINE;
> - if (queue_work(cxl_pmem_wq, &cxl_nvb->state_work))
> - get_device(&cxl_nvb->dev);
> + cxl_nvdimm_bridge_state_work(cxl_nvb);
>
> return 0;
> }
>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] cxl/pmem: Fix reference counting for delayed work
2021-10-31 18:58 ` Ben Widawsky
@ 2021-10-31 19:27 ` Dan Williams
0 siblings, 0 replies; 5+ messages in thread
From: Dan Williams @ 2021-10-31 19:27 UTC (permalink / raw)
To: Ben Widawsky; +Cc: linux-cxl, Weiny, Ira, Vishal L Verma, Schofield, Alison
On Sun, Oct 31, 2021 at 11:58 AM Ben Widawsky <ben.widawsky@intel.com> wrote:
>
> On 21-10-29 12:55:47, Dan Williams wrote:
> > There is a potential race between queue_work() returning and the
> > queued-work running that could result in put_device() running before
> > get_device(). Introduce the cxl_nvdimm_bridge_state_work() helper that
> > takes the reference unconditionally, but drops it if no new work was
> > queued, to keep the references balanced.
> >
> > Signed-off-by: Dan Williams <dan.j.williams@intel.com>
>
> Arguably fixes/stable?
Whoops, yes, definitely needs a Fixes:. A stable cc is debatable, but
if autosel picked up this fix I wouldn't say no, so might as well cc
stable upfront.
> Reviewed-by: Ben Widawsky <ben.widawsky@intel.com>
Thanks!
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] cxl/pmem: Fix reference counting for delayed work
2021-10-29 19:55 [PATCH] cxl/pmem: Fix reference counting for delayed work Dan Williams
2021-10-31 18:58 ` Ben Widawsky
@ 2021-11-01 11:30 ` Jonathan Cameron
2021-11-03 0:57 ` Dan Williams
1 sibling, 1 reply; 5+ messages in thread
From: Jonathan Cameron @ 2021-11-01 11:30 UTC (permalink / raw)
To: Dan Williams
Cc: linux-cxl, ira.weiny, ben.widawsky, vishal.l.verma, alison.schofield
On Fri, 29 Oct 2021 12:55:47 -0700
Dan Williams <dan.j.williams@intel.com> wrote:
> There is a potential race between queue_work() returning and the
> queued-work running that could result in put_device() running before
> get_device(). Introduce the cxl_nvdimm_bridge_state_work() helper that
> takes the reference unconditionally, but drops it if no new work was
> queued, to keep the references balanced.
>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Good spot, I'm guessing this was an inspection thing rather than a problem
you've managed to trigger, but either way.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> ---
> drivers/cxl/pmem.c | 17 +++++++++++++----
> 1 file changed, 13 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/cxl/pmem.c b/drivers/cxl/pmem.c
> index ceb2115981e5..38bcbb4e9409 100644
> --- a/drivers/cxl/pmem.c
> +++ b/drivers/cxl/pmem.c
> @@ -266,14 +266,24 @@ static void cxl_nvb_update_state(struct work_struct *work)
> put_device(&cxl_nvb->dev);
> }
>
> +static void cxl_nvdimm_bridge_state_work(struct cxl_nvdimm_bridge *cxl_nvb)
> +{
> + /*
> + * Take a reference that the workqueue will drop if new work
> + * gets queued.
> + */
> + get_device(&cxl_nvb->dev);
> + if (!queue_work(cxl_pmem_wq, &cxl_nvb->state_work))
> + put_device(&cxl_nvb->dev);
> +}
> +
> static void cxl_nvdimm_bridge_remove(struct device *dev)
> {
> struct cxl_nvdimm_bridge *cxl_nvb = to_cxl_nvdimm_bridge(dev);
>
> if (cxl_nvb->state == CXL_NVB_ONLINE)
> cxl_nvb->state = CXL_NVB_OFFLINE;
> - if (queue_work(cxl_pmem_wq, &cxl_nvb->state_work))
> - get_device(&cxl_nvb->dev);
> + cxl_nvdimm_bridge_state_work(cxl_nvb);
> }
>
> static int cxl_nvdimm_bridge_probe(struct device *dev)
> @@ -294,8 +304,7 @@ static int cxl_nvdimm_bridge_probe(struct device *dev)
> }
>
> cxl_nvb->state = CXL_NVB_ONLINE;
> - if (queue_work(cxl_pmem_wq, &cxl_nvb->state_work))
> - get_device(&cxl_nvb->dev);
> + cxl_nvdimm_bridge_state_work(cxl_nvb);
>
> return 0;
> }
>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] cxl/pmem: Fix reference counting for delayed work
2021-11-01 11:30 ` Jonathan Cameron
@ 2021-11-03 0:57 ` Dan Williams
0 siblings, 0 replies; 5+ messages in thread
From: Dan Williams @ 2021-11-03 0:57 UTC (permalink / raw)
To: Jonathan Cameron
Cc: linux-cxl, Weiny, Ira, Ben Widawsky, Vishal L Verma, Schofield, Alison
On Mon, Nov 1, 2021 at 4:30 AM Jonathan Cameron
<Jonathan.Cameron@huawei.com> wrote:
>
> On Fri, 29 Oct 2021 12:55:47 -0700
> Dan Williams <dan.j.williams@intel.com> wrote:
>
> > There is a potential race between queue_work() returning and the
> > queued-work running that could result in put_device() running before
> > get_device(). Introduce the cxl_nvdimm_bridge_state_work() helper that
> > takes the reference unconditionally, but drops it if no new work was
> > queued, to keep the references balanced.
> >
> > Signed-off-by: Dan Williams <dan.j.williams@intel.com>
>
> Good spot, I'm guessing this was an inspection thing rather than a problem
> you've managed to trigger, but either way.
Correct, it was by inspection. I expect it would be extremely
difficult to hit this race in practice.
> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Thanks!
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2021-11-03 0:57 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-10-29 19:55 [PATCH] cxl/pmem: Fix reference counting for delayed work Dan Williams
2021-10-31 18:58 ` Ben Widawsky
2021-10-31 19:27 ` Dan Williams
2021-11-01 11:30 ` Jonathan Cameron
2021-11-03 0:57 ` Dan Williams
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).