Skip to contentSkip to conversation

device-dax: avoid hang on error before devm_memremap_pages()

PATCH Applied Stefan Hajnoczi 3 participants 3 messages Jul 31, 2018 – Sep 11, 2018

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 device-dax: avoid hang on error before devm_memremap_pages()

dax_pmem_percpu_exit() waits for dax_pmem_percpu_release() to invoke the dax_pmem->cmp completion. Unfortunately this approach to cleaning up the percpu_ref only works after devm_memremap_pages() was successful.

If devm_add_action_or_reset() or devm_memremap_pages() fails, dax_pmem_percpu_release() is not invoked. Therefore dax_pmem_percpu_exit() hangs waiting for the completion:

rc = devm_add_action_or_reset(dev, dax_pmem_percpu_exit,
&dax_pmem->ref);
if (rc)
return rc;

  dax_pmem->pgmap.ref = &dax_pmem->ref;
  addr = devm_memremap_pages(dev, &dax_pmem->pgmap);

Avoid the hang by calling percpu_ref_exit() in the error paths instead of going through dax_pmem_percpu_exit().

---
Found by code inspection. Compile-tested only.

---
 drivers/dax/pmem.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/dax/pmem.c +8 -4
index fd49b24fd6af..99e2aace8078 100644
@@ -105,15 +105,19 @@ static int dax_pmem_probe(struct device *dev)
 	if (rc)
 		return rc;
 
-	rc = devm_add_action_or_reset(dev, dax_pmem_percpu_exit,
-							&dax_pmem->ref);
-	if (rc)
+	rc = devm_add_action(dev, dax_pmem_percpu_exit, &dax_pmem->ref);
+	if (rc) {
+		percpu_ref_exit(&dax_pmem->ref);
 		return rc;
+	}
 
 	dax_pmem->pgmap.ref = &dax_pmem->ref;
 	addr = devm_memremap_pages(dev, &dax_pmem->pgmap);
-	if (IS_ERR(addr))
+	if (IS_ERR(addr)) {
+		devm_remove_action(dev, dax_pmem_percpu_exit, &dax_pmem->ref);
+		percpu_ref_exit(&dax_pmem->ref);
 		return PTR_ERR(addr);
+	}
 
 	rc = devm_add_action_or_reset(dev, dax_pmem_percpu_kill,
 							&dax_pmem->ref);
Signature · 6 lines
2.17.1

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm
maintainer Applied
#1 lore raw
Quoted 16 lines · Stefan Hajnoczi

dax_pmem_percpu_exit() waits for dax_pmem_percpu_release() to invoke the dax_pmem->cmp completion. Unfortunately this approach to cleaning up the percpu_ref only works after devm_memremap_pages() was successful.

If devm_add_action_or_reset() or devm_memremap_pages() fails, dax_pmem_percpu_release() is not invoked. Therefore dax_pmem_percpu_exit() hangs waiting for the completion:

   rc = devm_add_action_or_reset(dev, dax_pmem_percpu_exit,
   				&dax_pmem->ref);
   if (rc)
   	return rc;

   dax_pmem->pgmap.ref = &dax_pmem->ref;
   addr = devm_memremap_pages(dev, &dax_pmem->pgmap);
Quoted 4 lines

Avoid the hang by calling percpu_ref_exit() in the error paths instead of going through dax_pmem_percpu_exit().

Applied

Quoted 24 lines

---
Found by code inspection. Compile-tested only.

---
  drivers/dax/pmem.c | 12 ++++++++----
  1 file changed, 8 insertions(+), 4 deletions(-)
drivers/dax/pmem.c +4 -3
index fd49b24fd6af..99e2aace8078 100644
@@ -105,15 +105,19 @@ static int dax_pmem_probe(struct device *dev)
  	if (rc)
  		return rc;
 
-	rc = devm_add_action_or_reset(dev, dax_pmem_percpu_exit,
-							&dax_pmem->ref);
-	if (rc)
+	rc = devm_add_action(dev, dax_pmem_percpu_exit, &dax_pmem->ref);
+	if (rc) {
+		percpu_ref_exit(&dax_pmem->ref);
  		return rc;
+	}
 
  	dax_pmem->pgmap.ref = &dax_pmem->ref;

drivers/dax/pmem.c static int dax_pmem_probe(struct device *dev) · line 123

Quoted 10 lines
drivers/dax/pmem.c +4 -1
  	addr = devm_memremap_pages(dev, &dax_pmem->pgmap);
-	if (IS_ERR(addr))
+	if (IS_ERR(addr)) {
+		devm_remove_action(dev, dax_pmem_percpu_exit, &dax_pmem->ref);
+		percpu_ref_exit(&dax_pmem->ref);
  		return PTR_ERR(addr);
+	}
 
  	rc = devm_add_action_or_reset(dev, dax_pmem_percpu_kill,
  							&dax_pmem->ref);

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm

#2 lore raw
Quoted 20 lines · Dave Jiang
Quoted 17 lines · Stefan Hajnoczi

dax_pmem_percpu_exit() waits for dax_pmem_percpu_release() to invoke the dax_pmem->cmp completion. Unfortunately this approach to cleaning up the percpu_ref only works after devm_memremap_pages() was successful.

If devm_add_action_or_reset() or devm_memremap_pages() fails, dax_pmem_percpu_release() is not invoked. Therefore dax_pmem_percpu_exit() hangs waiting for the completion:

   rc = devm_add_action_or_reset(dev, dax_pmem_percpu_exit,
                                &dax_pmem->ref);
   if (rc)
        return rc;

   dax_pmem->pgmap.ref = &dax_pmem->ref;
   addr = devm_memremap_pages(dev, &dax_pmem->pgmap);
Quoted 8 lines
Quoted 4 lines

Avoid the hang by calling percpu_ref_exit() in the error paths instead of going through dax_pmem_percpu_exit().

Applied

A similar problem exists in other devm_memremap_pages() users. I had a more comprehensive fix in-flight that I will rebase on top of this change.

    https://lore.kernel.org/patchwork/patch/961576/

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm