Skip to contentSkip to conversation

xen-blkback: free requests on disconnection

PATCH In review Roger Pau Monne 4 participants 19 messages Sep 4, 2015 – Sep 25, 2015

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

PATCH 1/1 xen-blkback: free requests on disconnection

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>

---
This patch should be backported to 4.2.

---
 drivers/block/xen-blkback/xenbus.c | 38 ++++++++++++++++++++------------------
 1 file changed, 20 insertions(+), 18 deletions(-)
drivers/block/xen-blkback/xenbus.c +20 -18
index deb3f00..7676575 100644
@@ -212,6 +212,9 @@ static int xen_blkif_map(struct xen_blkif *blkif, grant_ref_t *gref,
 
 static int xen_blkif_disconnect(struct xen_blkif *blkif)
 {
+	struct pending_req *req, *n;
+	int i = 0, j;
+
 	if (blkif->xenblkd) {
 		kthread_stop(blkif->xenblkd);
 		wake_up(&blkif->shutdown_wq);
@@ -238,13 +241,28 @@ static int xen_blkif_disconnect(struct xen_blkif *blkif)
 	/* Remove all persistent grants and the cache of ballooned pages. */
 	xen_blkbk_free_caches(blkif);
 
+	/* Check that there is no request in use */
+	list_for_each_entry_safe(req, n, &blkif->pending_free, free_list) {
+		list_del(&req->free_list);
+
+		for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++)
+			kfree(req->segments[j]);
+
+		for (j = 0; j < MAX_INDIRECT_PAGES; j++)
+			kfree(req->indirect_pages[j]);
+
+		kfree(req);
+		i++;
+	}
+
+	WARN_ON(i != (XEN_BLKIF_REQS_PER_PAGE * blkif->nr_ring_pages));
+	blkif->nr_ring_pages = 0;
+
 	return 0;
 }
 
 static void xen_blkif_free(struct xen_blkif *blkif)
 {
-	struct pending_req *req, *n;
-	int i = 0, j;
 
 	xen_blkif_disconnect(blkif);
 	xen_vbd_free(&blkif->vbd);
@@ -257,22 +275,6 @@ static void xen_blkif_free(struct xen_blkif *blkif)
 	BUG_ON(!list_empty(&blkif->free_pages));
 	BUG_ON(!RB_EMPTY_ROOT(&blkif->persistent_gnts));
 
-	/* Check that there is no request in use */
-	list_for_each_entry_safe(req, n, &blkif->pending_free, free_list) {
-		list_del(&req->free_list);
-
-		for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++)
-			kfree(req->segments[j]);
-
-		for (j = 0; j < MAX_INDIRECT_PAGES; j++)
-			kfree(req->indirect_pages[j]);
-
-		kfree(req);
-		i++;
-	}
-
-	WARN_ON(i != (XEN_BLKIF_REQS_PER_PAGE * blkif->nr_ring_pages));
-
 	kmem_cache_free(xen_blkif_cachep, blkif);
 }
 
Signature · 1 line
1.9.5 (Apple Git-50.3)
Tested-by on v1 1/1
#1 lore raw

Hi Roger,

Quoted 10 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Quoted 7 lines
Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>

The patch is fixing my problem when using UEFI in the guest. Thank you!

Regards,

Signature · 6 lines
Julien Grall

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel
Tested-by on v1 1/1
#2 lore raw
In reply to Roger Pau Monne

Hi Roger,

Quoted 10 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Quoted 7 lines
Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>

The patch is fixing my problem when using UEFI in the guest. Thank you!

Regards,

Signature · 1 line
Julien Grall
#3 lore raw
Quoted 21 lines · Julien Grall

Hi Roger,

Quoted 10 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Quoted 7 lines
Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 3 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Testing it now for regressions and should be sending an git pull next week for it.

Thanks!

Quoted 10 lines

Regards,

Signature · 6 lines
Julien Grall

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel
#4 lore raw
In reply to Julien Grall
Quoted 21 lines · Julien Grall

Hi Roger,

Quoted 10 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Quoted 7 lines
Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 3 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Testing it now for regressions and should be sending an git pull next week for it.

Thanks!

Quoted 10 lines

Regards,

Signature · 6 lines
Julien Grall

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel
#5 lore raw
In reply to Julien Grall

Hi Julien,

Quoted 21 lines · Julien Grall

Hi Roger,

Quoted 10 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Quoted 7 lines
Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

Thanks,
-Bob

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel

#6 lore raw
In reply to Julien Grall

Hi Julien,

Quoted 21 lines · Julien Grall

Hi Roger,

Quoted 10 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Quoted 7 lines
Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

Thanks,
-Bob

#7 lore raw
Quoted 1 line · Bob Liu

Hi Julien,

Hi Bob,

Quoted 22 lines · Julien Grall
Quoted 21 lines

Hi Roger,

Quoted 17 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 5 lines
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

This is related to commit 86839c56dee28c315a4c19b7bfee450ccd84cd25 "xen/block: add multi-page ring support" (Roger, it may be worth to indicate the offending commit in you commit message).

When starting a guest using UEFI. After the domain is destroyed I get the following warning from blkback:

------------[ cut here ]------------
WARNING: CPU: 2 PID: 95 at
/home/julien/works/linux/drivers/block/xen-blkback/xenbus.c:274
xen_blkif_deferred_free+0x1f4/0x1f8()
Modules linked in:
CPU: 2 PID: 95 Comm: kworker/2:1 Tainted: G        W       4.2.0 #85
Hardware name: APM X-Gene Mustang board (DT)
Workqueue: events xen_blkif_deferred_free
Call trace:
[<ffff8000000890a8>] dump_backtrace+0x0/0x124
[<ffff8000000891dc>] show_stack+0x10/0x1c
[<ffff8000007653bc>] dump_stack+0x78/0x98
[<ffff800000097e88>] warn_slowpath_common+0x9c/0xd4
[<ffff800000097f80>] warn_slowpath_null+0x14/0x20
[<ffff800000557a0c>] xen_blkif_deferred_free+0x1f0/0x1f8
[<ffff8000000ad020>] process_one_work+0x160/0x3b4
[<ffff8000000ad3b4>] worker_thread+0x140/0x494
[<ffff8000000b2e34>] kthread+0xd8/0xf0
---[ end trace 6f859b7883c88cdd ]---

This is because the allocation of the requests are done during the connection but the free is done when the domain is destroyed. Therefore if the domain is re-initializing the connection (because UEFI or PV Grub is used), the request won't be free and kept until the end.

So I think this should be backported in Linux 4.2 where the patch has been introduced.

Regards,

Signature · 6 lines
Julien Grall

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel
#8 lore raw
In reply to Bob Liu
Quoted 1 line · Bob Liu

Hi Julien,

Hi Bob,

Quoted 22 lines · Julien Grall
Quoted 21 lines

Hi Roger,

Quoted 17 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 5 lines
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

This is related to commit 86839c56dee28c315a4c19b7bfee450ccd84cd25 "xen/block: add multi-page ring support" (Roger, it may be worth to indicate the offending commit in you commit message).

When starting a guest using UEFI. After the domain is destroyed I get the following warning from blkback:

------------[ cut here ]------------
WARNING: CPU: 2 PID: 95 at
/home/julien/works/linux/drivers/block/xen-blkback/xenbus.c:274
xen_blkif_deferred_free+0x1f4/0x1f8()
Modules linked in:
CPU: 2 PID: 95 Comm: kworker/2:1 Tainted: G        W       4.2.0 #85
Hardware name: APM X-Gene Mustang board (DT)
Workqueue: events xen_blkif_deferred_free
Call trace:
[<ffff8000000890a8>] dump_backtrace+0x0/0x124
[<ffff8000000891dc>] show_stack+0x10/0x1c
[<ffff8000007653bc>] dump_stack+0x78/0x98
[<ffff800000097e88>] warn_slowpath_common+0x9c/0xd4
[<ffff800000097f80>] warn_slowpath_null+0x14/0x20
[<ffff800000557a0c>] xen_blkif_deferred_free+0x1f0/0x1f8
[<ffff8000000ad020>] process_one_work+0x160/0x3b4
[<ffff8000000ad3b4>] worker_thread+0x140/0x494
[<ffff8000000b2e34>] kthread+0xd8/0xf0
---[ end trace 6f859b7883c88cdd ]---

This is because the allocation of the requests are done during the connection but the free is done when the domain is destroyed. Therefore if the domain is re-initializing the connection (because UEFI or PV Grub is used), the request won't be free and kept until the end.

So I think this should be backported in Linux 4.2 where the patch has been introduced.

Regards,

Signature · 1 line
Julien Grall
#9 lore raw
Quoted 61 lines · Julien Grall
Quoted 1 line · Bob Liu

Hi Julien,

Hi Bob,

Quoted 22 lines · Julien Grall
Quoted 21 lines

Hi Roger,

Quoted 17 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 5 lines
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

This is related to commit 86839c56dee28c315a4c19b7bfee450ccd84cd25 "xen/block: add multi-page ring support" (Roger, it may be worth to indicate the offending commit in you commit message).

When starting a guest using UEFI. After the domain is destroyed I get the following warning from blkback:

------------[ cut here ]------------
WARNING: CPU: 2 PID: 95 at
/home/julien/works/linux/drivers/block/xen-blkback/xenbus.c:274
xen_blkif_deferred_free+0x1f4/0x1f8()
Modules linked in:
CPU: 2 PID: 95 Comm: kworker/2:1 Tainted: G        W       4.2.0 #85
Hardware name: APM X-Gene Mustang board (DT)
Workqueue: events xen_blkif_deferred_free
Call trace:
[<ffff8000000890a8>] dump_backtrace+0x0/0x124
[<ffff8000000891dc>] show_stack+0x10/0x1c
[<ffff8000007653bc>] dump_stack+0x78/0x98
[<ffff800000097e88>] warn_slowpath_common+0x9c/0xd4
[<ffff800000097f80>] warn_slowpath_null+0x14/0x20
[<ffff800000557a0c>] xen_blkif_deferred_free+0x1f0/0x1f8
[<ffff8000000ad020>] process_one_work+0x160/0x3b4
[<ffff8000000ad3b4>] worker_thread+0x140/0x494
[<ffff8000000b2e34>] kthread+0xd8/0xf0
---[ end trace 6f859b7883c88cdd ]---
Quoted 5 lines

This is because the allocation of the requests are done during the connection but the free is done when the domain is destroyed. Therefore if the domain is re-initializing the connection (because UEFI or PV Grub is used), the request won't be free and kept until the end.

Thank you!
Roger, I think it's better to have this information in your commit message too.

Signature · 2 lines
Regards,
-Bob
#10 lore raw

El 07/09/15 a les 13.35, Bob Liu ha escrit:

Quoted 63 lines · Julien Grall
Quoted 61 lines · Bob Liu
Quoted 1 line

Hi Julien,

Hi Bob,

Quoted 22 lines · Julien Grall
Quoted 21 lines

Hi Roger,

Quoted 17 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 5 lines
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

This is related to commit 86839c56dee28c315a4c19b7bfee450ccd84cd25 "xen/block: add multi-page ring support" (Roger, it may be worth to indicate the offending commit in you commit message).

When starting a guest using UEFI. After the domain is destroyed I get the following warning from blkback:

------------[ cut here ]------------
WARNING: CPU: 2 PID: 95 at
/home/julien/works/linux/drivers/block/xen-blkback/xenbus.c:274
xen_blkif_deferred_free+0x1f4/0x1f8()
Modules linked in:
CPU: 2 PID: 95 Comm: kworker/2:1 Tainted: G        W       4.2.0 #85
Hardware name: APM X-Gene Mustang board (DT)
Workqueue: events xen_blkif_deferred_free
Call trace:
[<ffff8000000890a8>] dump_backtrace+0x0/0x124
[<ffff8000000891dc>] show_stack+0x10/0x1c
[<ffff8000007653bc>] dump_stack+0x78/0x98
[<ffff800000097e88>] warn_slowpath_common+0x9c/0xd4
[<ffff800000097f80>] warn_slowpath_null+0x14/0x20
[<ffff800000557a0c>] xen_blkif_deferred_free+0x1f0/0x1f8
[<ffff8000000ad020>] process_one_work+0x160/0x3b4
[<ffff8000000ad3b4>] worker_thread+0x140/0x494
[<ffff8000000b2e34>] kthread+0xd8/0xf0
---[ end trace 6f859b7883c88cdd ]---
Quoted 8 lines
Quoted 5 lines

This is because the allocation of the requests are done during the connection but the free is done when the domain is destroyed. Therefore if the domain is re-initializing the connection (because UEFI or PV Grub is used), the request won't be free and kept until the end.

Thank you!
Roger, I think it's better to have this information in your commit message too.

Konrad, would you like me to resend the patch with the modified commit message, or do you plan to amend it yourself while committing?

Roger.

#11 lore raw
Quoted 70 lines · Roger Pau Monné

El 07/09/15 a les 13.35, Bob Liu ha escrit:

Quoted 69 lines · Julien Grall
Quoted 66 lines · Bob Liu
Quoted 1 line

Hi Julien,

Hi Bob,

Quoted 22 lines · Julien Grall
Quoted 21 lines

Hi Roger,

Quoted 17 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 5 lines
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

This is related to commit 86839c56dee28c315a4c19b7bfee450ccd84cd25 "xen/block: add multi-page ring support" (Roger, it may be worth to indicate the offending commit in you commit message).

When starting a guest using UEFI. After the domain is destroyed I get the following warning from blkback:

------------[ cut here ]------------
WARNING: CPU: 2 PID: 95 at
/home/julien/works/linux/drivers/block/xen-blkback/xenbus.c:274
xen_blkif_deferred_free+0x1f4/0x1f8()
Modules linked in:
CPU: 2 PID: 95 Comm: kworker/2:1 Tainted: G        W       4.2.0 #85
Hardware name: APM X-Gene Mustang board (DT)
Workqueue: events xen_blkif_deferred_free
Call trace:
[<ffff8000000890a8>] dump_backtrace+0x0/0x124
[<ffff8000000891dc>] show_stack+0x10/0x1c
[<ffff8000007653bc>] dump_stack+0x78/0x98
[<ffff800000097e88>] warn_slowpath_common+0x9c/0xd4
[<ffff800000097f80>] warn_slowpath_null+0x14/0x20
[<ffff800000557a0c>] xen_blkif_deferred_free+0x1f0/0x1f8
[<ffff8000000ad020>] process_one_work+0x160/0x3b4
[<ffff8000000ad3b4>] worker_thread+0x140/0x494
[<ffff8000000b2e34>] kthread+0xd8/0xf0
---[ end trace 6f859b7883c88cdd ]---

This is because the allocation of the requests are done during the connection but the free is done when the domain is destroyed. Therefore if the domain is re-initializing the connection (because UEFI or PV Grub is used), the request won't be free and kept until the end.

Quoted 5 lines
Quoted 2 lines

Thank you!
Roger, I think it's better to have this information in your commit message too.

Konrad, would you like me to resend the patch with the modified commit message, or do you plan to amend it yourself while committing?

I will amend it. Thanks!

Quoted 3 lines

Roger.

#12 lore raw
In reply to Roger Pau Monné
Quoted 70 lines · Roger Pau Monné

El 07/09/15 a les 13.35, Bob Liu ha escrit:

Quoted 69 lines · Julien Grall
Quoted 66 lines · Bob Liu
Quoted 1 line

Hi Julien,

Hi Bob,

Quoted 22 lines · Julien Grall
Quoted 21 lines

Hi Roger,

Quoted 17 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 5 lines
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

This is related to commit 86839c56dee28c315a4c19b7bfee450ccd84cd25 "xen/block: add multi-page ring support" (Roger, it may be worth to indicate the offending commit in you commit message).

When starting a guest using UEFI. After the domain is destroyed I get the following warning from blkback:

------------[ cut here ]------------
WARNING: CPU: 2 PID: 95 at
/home/julien/works/linux/drivers/block/xen-blkback/xenbus.c:274
xen_blkif_deferred_free+0x1f4/0x1f8()
Modules linked in:
CPU: 2 PID: 95 Comm: kworker/2:1 Tainted: G        W       4.2.0 #85
Hardware name: APM X-Gene Mustang board (DT)
Workqueue: events xen_blkif_deferred_free
Call trace:
[<ffff8000000890a8>] dump_backtrace+0x0/0x124
[<ffff8000000891dc>] show_stack+0x10/0x1c
[<ffff8000007653bc>] dump_stack+0x78/0x98
[<ffff800000097e88>] warn_slowpath_common+0x9c/0xd4
[<ffff800000097f80>] warn_slowpath_null+0x14/0x20
[<ffff800000557a0c>] xen_blkif_deferred_free+0x1f0/0x1f8
[<ffff8000000ad020>] process_one_work+0x160/0x3b4
[<ffff8000000ad3b4>] worker_thread+0x140/0x494
[<ffff8000000b2e34>] kthread+0xd8/0xf0
---[ end trace 6f859b7883c88cdd ]---

This is because the allocation of the requests are done during the connection but the free is done when the domain is destroyed. Therefore if the domain is re-initializing the connection (because UEFI or PV Grub is used), the request won't be free and kept until the end.

Quoted 5 lines
Quoted 2 lines

Thank you!
Roger, I think it's better to have this information in your commit message too.

Konrad, would you like me to resend the patch with the modified commit message, or do you plan to amend it yourself while committing?

I will amend it. Thanks!

Quoted 3 lines

Roger.

#13 lore raw

Hi Konrad,

Quoted 4 lines · Konrad Rzeszutek Wilk
Quoted 2 lines

Konrad, would you like me to resend the patch with the modified commit message, or do you plan to amend it yourself while committing?

I will amend it. Thanks!

What the status for this patch?

Regards,

Signature · 1 line
Julien Grall
#14 lore raw
Quoted 9 lines · Julien Grall

Hi Konrad,

Quoted 4 lines · Konrad Rzeszutek Wilk
Quoted 2 lines

Konrad, would you like me to resend the patch with the modified commit message, or do you plan to amend it yourself while committing?

I will amend it. Thanks!

What the status for this patch?

Applied and sent the git pull.

Quoted 5 lines

Regards,

Signature · 1 line
Julien Grall
#15 lore raw
In reply to Julien Grall
Quoted 9 lines · Julien Grall

Hi Konrad,

Quoted 4 lines · Konrad Rzeszutek Wilk
Quoted 2 lines

Konrad, would you like me to resend the patch with the modified commit message, or do you plan to amend it yourself while committing?

I will amend it. Thanks!

What the status for this patch?

Applied and sent the git pull.

Quoted 5 lines

Regards,

Signature · 1 line
Julien Grall
#16 lore raw
In reply to Konrad Rzeszutek Wilk

Hi Konrad,

Quoted 4 lines · Konrad Rzeszutek Wilk
Quoted 2 lines

Konrad, would you like me to resend the patch with the modified commit message, or do you plan to amend it yourself while committing?

I will amend it. Thanks!

What the status for this patch?

Regards,

Signature · 1 line
Julien Grall
#17 lore raw
In reply to Bob Liu

El 07/09/15 a les 13.35, Bob Liu ha escrit:

Quoted 63 lines · Julien Grall
Quoted 61 lines · Bob Liu
Quoted 1 line

Hi Julien,

Hi Bob,

Quoted 22 lines · Julien Grall
Quoted 21 lines

Hi Roger,

Quoted 17 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 5 lines
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

This is related to commit 86839c56dee28c315a4c19b7bfee450ccd84cd25 "xen/block: add multi-page ring support" (Roger, it may be worth to indicate the offending commit in you commit message).

When starting a guest using UEFI. After the domain is destroyed I get the following warning from blkback:

------------[ cut here ]------------
WARNING: CPU: 2 PID: 95 at
/home/julien/works/linux/drivers/block/xen-blkback/xenbus.c:274
xen_blkif_deferred_free+0x1f4/0x1f8()
Modules linked in:
CPU: 2 PID: 95 Comm: kworker/2:1 Tainted: G        W       4.2.0 #85
Hardware name: APM X-Gene Mustang board (DT)
Workqueue: events xen_blkif_deferred_free
Call trace:
[<ffff8000000890a8>] dump_backtrace+0x0/0x124
[<ffff8000000891dc>] show_stack+0x10/0x1c
[<ffff8000007653bc>] dump_stack+0x78/0x98
[<ffff800000097e88>] warn_slowpath_common+0x9c/0xd4
[<ffff800000097f80>] warn_slowpath_null+0x14/0x20
[<ffff800000557a0c>] xen_blkif_deferred_free+0x1f0/0x1f8
[<ffff8000000ad020>] process_one_work+0x160/0x3b4
[<ffff8000000ad3b4>] worker_thread+0x140/0x494
[<ffff8000000b2e34>] kthread+0xd8/0xf0
---[ end trace 6f859b7883c88cdd ]---
Quoted 8 lines
Quoted 5 lines

This is because the allocation of the requests are done during the connection but the free is done when the domain is destroyed. Therefore if the domain is re-initializing the connection (because UEFI or PV Grub is used), the request won't be free and kept until the end.

Thank you!
Roger, I think it's better to have this information in your commit message too.

Konrad, would you like me to resend the patch with the modified commit message, or do you plan to amend it yourself while committing?

Roger.

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel

#18 lore raw
In reply to Julien Grall
Quoted 61 lines · Julien Grall
Quoted 1 line · Bob Liu

Hi Julien,

Hi Bob,

Quoted 22 lines · Julien Grall
Quoted 21 lines

Hi Roger,

Quoted 17 lines · Roger Pau Monne

Request allocation has been moved to connect_ring, which is called every time blkback connects to the frontend (this can happen multiple times during a blkback instance life cycle). On the other hand, request freeing has not been moved, so it's only called when destroying the backend instance. Due to this mismatch, blkback can allocate the request pool multiple times, without freeing it.

In order to fix it, move the freeing of requests to xen_blkif_disconnect to restore the symmetry between request allocation and freeing.

Cc Julien Grall <[email protected]>
Cc Konrad Rzeszutek Wilk <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc David Vrabel <[email protected]>
Quoted 5 lines
Quoted 2 lines

The patch is fixing my problem when using UEFI in the guest. Thank you!

Could you please explain the problem you met a bit more? So that I can know back port this patch if met similar issue.

This is related to commit 86839c56dee28c315a4c19b7bfee450ccd84cd25 "xen/block: add multi-page ring support" (Roger, it may be worth to indicate the offending commit in you commit message).

When starting a guest using UEFI. After the domain is destroyed I get the following warning from blkback:

------------[ cut here ]------------
WARNING: CPU: 2 PID: 95 at
/home/julien/works/linux/drivers/block/xen-blkback/xenbus.c:274
xen_blkif_deferred_free+0x1f4/0x1f8()
Modules linked in:
CPU: 2 PID: 95 Comm: kworker/2:1 Tainted: G        W       4.2.0 #85
Hardware name: APM X-Gene Mustang board (DT)
Workqueue: events xen_blkif_deferred_free
Call trace:
[<ffff8000000890a8>] dump_backtrace+0x0/0x124
[<ffff8000000891dc>] show_stack+0x10/0x1c
[<ffff8000007653bc>] dump_stack+0x78/0x98
[<ffff800000097e88>] warn_slowpath_common+0x9c/0xd4
[<ffff800000097f80>] warn_slowpath_null+0x14/0x20
[<ffff800000557a0c>] xen_blkif_deferred_free+0x1f0/0x1f8
[<ffff8000000ad020>] process_one_work+0x160/0x3b4
[<ffff8000000ad3b4>] worker_thread+0x140/0x494
[<ffff8000000b2e34>] kthread+0xd8/0xf0
---[ end trace 6f859b7883c88cdd ]---
Quoted 5 lines

This is because the allocation of the requests are done during the connection but the free is done when the domain is destroyed. Therefore if the domain is re-initializing the connection (because UEFI or PV Grub is used), the request won't be free and kept until the end.

Thank you!
Roger, I think it's better to have this information in your commit message too.

Signature · 7 lines
Regards,
-Bob

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel