Skip to contentSkip to conversation

fbarray: fix attach deadlock

PATCH Superseded Darek Stojaczyk 4 participants 5 messages Mar 29, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fbarray: fix attach deadlock

rte_fbarray_attach() currently locks its internal spinlock, but never releases it. Secondary processes won't even start if there is more than one fbarray to be attached to - the second rte_fbarray_attach() would be just stuck.

Fix it by releasing the lock at the end of rte_fbarray_attach(). I believe this was the original intention.

Fixes 5b61c62cfd76 ("fbarray: add internal tailq for mapped areas")
---
 lib/librte_eal/common/eal_common_fbarray.c | 2 ++
 1 file changed, 2 insertions(+)
lib/librte_eal/common/eal_common_fbarray.c +2 -0
index 0e7366e5e..5ca8d6f0e 100644
@@ -906,6 +906,7 @@ rte_fbarray_attach(struct rte_fbarray *arr)
 
 	/* we're done */
 
+	rte_spinlock_unlock(&mem_area_lock);
 	return 0;
 fail:
 	if (data)
@@ -913,6 +914,7 @@ rte_fbarray_attach(struct rte_fbarray *arr)
 	if (fd >= 0)
 		close(fd);
 	free(ma);
+	rte_spinlock_unlock(&mem_area_lock);
 	return -1;
 }
 
Signature · 1 line
2.17.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 44 lines

-----Original Message-----
From: dev <[email protected]> On Behalf Of Darek Stojaczyk
Sent: Friday, March 29, 2019 1:10 PM
To: [email protected]
Cc: [email protected]; [email protected]; Darek Stojaczyk <[email protected]>; [email protected]; [email protected]
Subject: [dpdk-dev] [PATCH] fbarray: fix attach deadlock

rte_fbarray_attach() currently locks its internal spinlock, but never releases it. Secondary processes won't even start if there is more than one fbarray to be attached to - the second rte_fbarray_attach() would be just stuck.

Fix it by releasing the lock at the end of rte_fbarray_attach(). I believe this was the original intention.

Fixes 5b61c62cfd76 ("fbarray: add internal tailq for mapped areas")
---
 lib/librte_eal/common/eal_common_fbarray.c | 2 ++
 1 file changed, 2 insertions(+)
lib/librte_eal/common/eal_common_fbarray.c +1 -0
index 0e7366e5e..5ca8d6f0e 100644
@@ -906,6 +906,7 @@ rte_fbarray_attach(struct rte_fbarray *arr)
 
 /* we're done */
 
+rte_spinlock_unlock(&mem_area_lock);
 return 0;
 fail:
 if (data)

lib/librte_eal/common/eal_common_fbarray.c rte_fbarray_attach(struct rte_fbarray *arr) · line 919

Quoted 6 lines
lib/librte_eal/common/eal_common_fbarray.c +1 -0
@@ -913,6 +914,7 @@ rte_fbarray_attach(struct rte_fbarray *arr)
 if (fd >= 0)
 close(fd);
 free(ma);
+rte_spinlock_unlock(&mem_area_lock);
 return -1;
 }

This is an obvious issue, good catch!

Quoted 2 lines
lib/librte_eal/common/eal_common_fbarray.c +0 -1
--

2.17.1

IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.

#2 lore raw

PATCH v2 1/1 fbarray: fix attach deadlock

In reply to Darek Stojaczyk

rte_fbarray_attach() currently locks its internal spinlock, but never releases it. Secondary processes won't even start if there is more than one fbarray to be attached to - the second rte_fbarray_attach() would be just stuck.

Fix it by releasing the lock at the end of rte_fbarray_attach(). I believe this was the original intention.

Fixes 5b61c62cfd76 ("fbarray: add internal tailq for mapped areas")

---
v2:
- fixed one more case where we could unlock the spinlock before locking it

 lib/librte_eal/common/eal_common_fbarray.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
lib/librte_eal/common/eal_common_fbarray.c +6 -2
index 0e7366e5e..31cce80a8 100644
@@ -859,8 +859,10 @@ rte_fbarray_attach(struct rte_fbarray *arr)
 	}
 
 	page_sz = sysconf(_SC_PAGESIZE);
-	if (page_sz == (size_t)-1)
-		goto fail;
+	if (page_sz == (size_t)-1) {
+		free(ma);
+		return -1;
+	}
 
 	mmap_len = calc_data_size(page_sz, arr->elt_sz, arr->len);
 
@@ -906,6 +908,7 @@ rte_fbarray_attach(struct rte_fbarray *arr)
 
 	/* we're done */
 
+	rte_spinlock_unlock(&mem_area_lock);
 	return 0;
 fail:
 	if (data)
@@ -913,6 +916,7 @@ rte_fbarray_attach(struct rte_fbarray *arr)
 	if (fd >= 0)
 		close(fd);
 	free(ma);
+	rte_spinlock_unlock(&mem_area_lock);
 	return -1;
 }
 
Signature · 1 line
2.17.1
Acked-by on v2 1/1
#3 lore raw
Quoted 14 lines · Darek Stojaczyk

rte_fbarray_attach() currently locks its internal spinlock, but never releases it. Secondary processes won't even start if there is more than one fbarray to be attached to - the second rte_fbarray_attach() would be just stuck.

Fix it by releasing the lock at the end of rte_fbarray_attach(). I believe this was the original intention.

Fixes 5b61c62cfd76 ("fbarray: add internal tailq for mapped areas")
Quoted 6 lines

---
v2:
- fixed one more case where we could unlock the spinlock before locking it

Thanks for catching this!

There is one more case where we do unlock on init without locking, i'll submit a patch separately (and will check other functions with a microscope just in case).

Signature · 2 lines
Thanks,
Anatoly
#4 lore raw

29/03/2019 11:42, Burakov, Anatoly:

Quoted 22 lines · Darek Stojaczyk
Quoted 14 lines

rte_fbarray_attach() currently locks its internal spinlock, but never releases it. Secondary processes won't even start if there is more than one fbarray to be attached to - the second rte_fbarray_attach() would be just stuck.

Fix it by releasing the lock at the end of rte_fbarray_attach(). I believe this was the original intention.

Fixes 5b61c62cfd76 ("fbarray: add internal tailq for mapped areas")
Quoted 6 lines

---
v2:
- fixed one more case where we could unlock the spinlock before locking it

Quoted 3 lines

Thanks for catching this!

Applied, thanks

Quoted 3 lines

There is one more case where we do unlock on init without locking, i'll submit a patch separately (and will check other functions with a microscope just in case).

We'll take this one too.