virtio_balloon: fix race by fill and leak
Subsystems (4)
PATCH 1/2 virtio_balloon: fix race by fill and leak
During my compaction-related stuff, I encountered a bug with ballooning.
With repeated inflating and deflating cycle, guest memory( ie, cat /proc/meminfo | grep MemTotal) is decreased and couldn't be recovered.
The reason is balloon_lock doesn't cover release_pages_balloon so struct virtio_balloon fields could be overwritten by race of fill_balloon(e,g, vb->*pfns could be critical).
This patch fixes it in my test.
--- drivers/virtio/virtio_balloon.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/virtio/virtio_balloon.c +1 -1
@@ -209,8 +209,8 @@ static unsigned leak_balloon(struct virtio_balloon *vb, size_t num)*/ if (vb->num_pfns !=0 )tell_host (vb, vb->deflate_vq); -mutex_unlock (&vb->balloon_lock);release_pages_balloon (vb); +mutex_unlock (&vb->balloon_lock);return num_freed_pages; }
Signature · 7 lines
1.9.1 -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
PATCH 2/2 virtio_balloon: fix race between migration and ballooning
In balloon_page_dequeue, pages_lock should cover the loop (ie, list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and then list_del by isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.
general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exception
Dumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabled
--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
mm/balloon_compaction.c +2 -2
@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around
@@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE ); -spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/*
Signature · 7 lines
1.9.1 -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
Quoted 84 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +2
-2@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE ); -spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/* Signature · 1 line
1.9.1
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Minchan KimQuoted 84 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +2
-2@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE ); -spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/* Signature · 1 line
1.9.1
Michael S. Tsirkin
maintainerRe: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Minchan KimQuoted 71 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +1
-1@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE );
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 89
Quoted 10 lines
mm/balloon_compaction.c +1
-1-spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/*
I think this will cause deadlocks.
pages_lock now nests within page lock, balloon_page_putback nests them in the reverse order.
Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
We really must take a page reference before touching the page.
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
mm/balloon_compaction.c +32 -2
@@ -56,12 +56,34 @@ EXPORT_SYMBOL_GPL(balloon_page_enqueue);*/ struct page *balloon_page_dequeue (struct balloon_dev_info *b_dev_info) { -struct page *page, *tmp; +struct page *page;unsigned long flags;bool dequeued_page; +LIST_HEAD (processed);/* protected by b_dev_info->pages_lock */ dequeued_page = false; -list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) { +/* +* We need to go over b_dev_info->pages and lock each page, +* but b_dev_info->pages_lock must nest within page lock. +* +* To make this safe, remove each page from b_dev_info->pages list +* under b_dev_info->pages_lock, then drop this lock. Once list is +* empty, re-add them also under b_dev_info->pages_lock. +*/ +spin_lock_irqsave (&b_dev_info->pages_lock, flags); +while (!list_empty (&b_dev_info->pages)) { + page =list_first_entry (&b_dev_info->pages,typeof (*page), lru); +/* move to processed list to avoid going over it another time */ +list_move (&page->lru, &processed); + +if (!get_page_unless_zero (page)) +continue ; +/* +* pages_lock nests within page lock, +* so drop it before trylock_page +*/ +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +/* * Block others from accessing the 'page' while we get around * establishing additional references and preparing the 'page'
@@ -72,6 +94,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)if (!PagePrivate (page)) {/* raced with isolation */ unlock_page (page); +put_page (page);continue ; }#endif
@@ -80,11 +103,18 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)__count_vm_event (BALLOON_DEFLATE );spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); +put_page (page); dequeued_page = true;break ; } +put_page (page); +spin_lock_irqsave (&b_dev_info->pages_lock, flags); } +/* re-add remaining entries */ +list_splice (&processed, &b_dev_info->pages); +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +if (!dequeued_page) {/* * If we are unable to dequeue a balloon page because the page
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
Quoted 80 lines · Michael S. Tsirkin
Quoted 79 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +2
-2@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE ); -spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);
Quoted 7 lines
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 119
Quoted 2 lines
mm/balloon_compaction.c +0
-0if (!dequeued_page) {/* I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.
In balloon_page_dequeu, we used trylock so I don't think it's deadlock.
Quoted 3 lines
Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
I did but I don't see any warning.
Quoted 3 lines
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
If a page was isolated, the page shouldn't stay b_dev_info->pages list so balloon_page_dequeue cannot see the page.
Am I missing something?
Quoted 7 lines
We really must take a page reference before touching the page.
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
If I missed something, I am happy to retest and report the result when I go to the office.
Thanks.
Quoted 70 lines
mm/balloon_compaction.c +32
-2@@ -56,12 +56,34 @@ EXPORT_SYMBOL_GPL(balloon_page_enqueue);*/ struct page *balloon_page_dequeue (struct balloon_dev_info *b_dev_info) { -struct page *page, *tmp; +struct page *page;unsigned long flags;bool dequeued_page; +LIST_HEAD (processed);/* protected by b_dev_info->pages_lock */ dequeued_page = false; -list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) { +/* +* We need to go over b_dev_info->pages and lock each page, +* but b_dev_info->pages_lock must nest within page lock. +* +* To make this safe, remove each page from b_dev_info->pages list +* under b_dev_info->pages_lock, then drop this lock. Once list is +* empty, re-add them also under b_dev_info->pages_lock. +*/ +spin_lock_irqsave (&b_dev_info->pages_lock, flags); +while (!list_empty (&b_dev_info->pages)) { + page =list_first_entry (&b_dev_info->pages,typeof (*page), lru); +/* move to processed list to avoid going over it another time */ +list_move (&page->lru, &processed); + +if (!get_page_unless_zero (page)) +continue ; +/* +* pages_lock nests within page lock, +* so drop it before trylock_page +*/ +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +/* * Block others from accessing the 'page' while we get around * establishing additional references and preparing the 'page' @@ -72,6 +94,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)if (!PagePrivate (page)) {/* raced with isolation */ unlock_page (page); +put_page (page);continue ; }#endif @@ -80,11 +103,18 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)__count_vm_event (BALLOON_DEFLATE );spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); +put_page (page); dequeued_page = true;break ; } +put_page (page); +spin_lock_irqsave (&b_dev_info->pages_lock, flags); } +/* re-add remaining entries */ +list_splice (&processed, &b_dev_info->pages); +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +if (!dequeued_page) {/* * If we are unable to dequeue a balloon page because the page
Signature · 8 lines
Kind regards, Minchan Kim -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Michael S. Tsirkin
maintainerRe: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
Quoted 7 lines · Minchan Kim
Quoted 4 lines
I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.In
balloon_page_dequeu, we used trylock so I don't think it's deadlock.
I went over this again and I don't see the issue anymore. I think I was mistaken, so I dropped my patch and picked up yours. Sorry about the noise.
Quoted 7 lines
Quoted 3 lines
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
If a page was isolated, the page shouldn't stay
b_dev_info->pageslist soballoon_page_dequeuecannot see the page.
Am I missing something?
I mean without locks, as it is now. With either your or my patch in place, it's fine.
Signature · 7 lines
MST -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
Quoted 6 lines · Michael S. Tsirkin
Quoted 5 lines · Minchan Kim
Quoted 4 lines
I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.
Quoted 6 lines
Quoted 2 lines
In
balloon_page_dequeu, we used trylock so I don't think it's deadlock.I went over this again and I don't see the issue anymore. I think I was mistaken, so I dropped my patch and picked up yours. Sorry about the noise.
No problem. Thanks for the review.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Michael S. TsirkinQuoted 6 lines · Michael S. Tsirkin
Quoted 5 lines · Minchan Kim
Quoted 4 lines
I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.
Quoted 6 lines
Quoted 2 lines
In
balloon_page_dequeu, we used trylock so I don't think it's deadlock.I went over this again and I don't see the issue anymore. I think I was mistaken, so I dropped my patch and picked up yours. Sorry about the noise.
No problem. Thanks for the review.
Michael S. Tsirkin
maintainerRe: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Minchan KimQuoted 7 lines · Minchan Kim
Quoted 4 lines
I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.In
balloon_page_dequeu, we used trylock so I don't think it's deadlock.
I went over this again and I don't see the issue anymore. I think I was mistaken, so I dropped my patch and picked up yours. Sorry about the noise.
Quoted 7 lines
Quoted 3 lines
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
If a page was isolated, the page shouldn't stay
b_dev_info->pageslist soballoon_page_dequeuecannot see the page.
Am I missing something?
I mean without locks, as it is now. With either your or my patch in place, it's fine.
Signature · 1 line
MST
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Michael S. TsirkinQuoted 94 lines · Michael S. Tsirkin
Quoted 71 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +1
-1@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE );mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 89
Quoted 10 lines
mm/balloon_compaction.c +1
-1-spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/* I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
Quoted 7 lines
We really must take a page reference before touching the page.
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
Nice catch! Thanks for spotting it. I just have one minor nit. See below
Quoted 51 lines
mm/balloon_compaction.c +25
-2@@ -56,12 +56,34 @@ EXPORT_SYMBOL_GPL(balloon_page_enqueue);*/ struct page *balloon_page_dequeue (struct balloon_dev_info *b_dev_info) { -struct page *page, *tmp; +struct page *page;unsigned long flags;bool dequeued_page; +LIST_HEAD (processed);/* protected by b_dev_info->pages_lock */ dequeued_page = false; -list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) { +/* +* We need to go over b_dev_info->pages and lock each page, +* but b_dev_info->pages_lock must nest within page lock. +* +* To make this safe, remove each page from b_dev_info->pages list +* under b_dev_info->pages_lock, then drop this lock. Once list is +* empty, re-add them also under b_dev_info->pages_lock. +*/ +spin_lock_irqsave (&b_dev_info->pages_lock, flags); +while (!list_empty (&b_dev_info->pages)) { + page =list_first_entry (&b_dev_info->pages,typeof (*page), lru); +/* move to processed list to avoid going over it another time */ +list_move (&page->lru, &processed); + +if (!get_page_unless_zero (page)) +continue ; +/* +* pages_lock nests within page lock, +* so drop it before trylock_page +*/ +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +/* * Block others from accessing the 'page' while we get around * establishing additional references and preparing the 'page' @@ -72,6 +94,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)if (!PagePrivate (page)) {/* raced with isolation */ unlock_page (page); +put_page (page);continue ; }#endif
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 108
Quoted 6 lines
mm/balloon_compaction.c +1
-0@@ -80,11 +103,18 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)__count_vm_event (BALLOON_DEFLATE );spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); +put_page (page); dequeued_page = true;break ;
^^^^[1]
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 115
Quoted 7 lines
mm/balloon_compaction.c +4
-0} +put_page (page); +spin_lock_irqsave (&b_dev_info->pages_lock, flags); } +/* re-add remaining entries */ +list_splice (&processed, &b_dev_info->pages);
By breaking the loop at its ordinary and expected way-out case [1] we'll hit list_splice without holding b_dev_info->pages_lock, won't we?
perhaps by adding the following on top of your patch we can address that pickle aforementioned:
Cheers!
Rafael
--
mm/balloon_compaction.c +0 -0
@@ -58,7 +58,7 @@ struct page *balloon_page_dequeue(struct
balloon_dev_info *b_dev_info)
{
struct page *page;
unsigned long flags;
- bool dequeued_page;
+ bool dequeued_page, locked;
LIST_HEAD(processed); /* protected by b_dev_info->pages_lock */
dequeued_page = false;
@@ -105,13 +105,17 @@ struct page *balloon_page_dequeue(struct
balloon_dev_info *b_dev_info)
unlock_page(page);
put_page(page);
dequeued_page = true;
+ locked = false;
break;
}
put_page(page);
spin_lock_irqsave(&b_dev_info->pages_lock, flags);
+ locked = true;
}
/* re-add remaining entries */
+ if (!locked)
+ spin_lock_irqsave(&b_dev_info->pages_lock, flags);
list_splice(&processed, &b_dev_info->pages);
spin_unlock_irqrestore(&b_dev_info->pages_lock, flags);--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
Quoted 98 lines · Rafael Aquini
Quoted 97 lines · Michael S. Tsirkin
Quoted 71 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +1
-1@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE );mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 89
Quoted 10 lines
mm/balloon_compaction.c +1
-1-spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/* I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
We really must take a page reference before touching the page.
Quoted 7 lines
Quoted 4 lines
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
Nice catch! Thanks for spotting it. I just have one minor nit. See below
Hmm, As I replied mst's mail, I really cannot understand what you gouys are pointing out.
If we use lock_page in balloon_page_dequeue, I agree it's deadlock but we used trylock_page so it's not a deadlock.
About the page refcount, we don't need take a page reference because if one of page in the list was isolated for migration, the page shouldn't stay in b_dev_info->pages list so balloon_page_dequeue cannot touch the page.
Could you elaborate it more detail if I have missed something? It's a stable patch so I want to be careful.
Thanks.
balloon_page_isolate
{
trylock_page(page)
spin_lock_irqsave(&b_dev_info->pages_lock)
list_del(&page->lru);
}balloon_page_dequeue
{
spin_lock_irqsave(&b_dev_info->pages_lock)
list_for_each_entry_safe(page, tmp, &b_dev_info->pages, lru) {
trylock_page(page)
}}
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Rafael AquiniQuoted 98 lines · Rafael Aquini
Quoted 97 lines · Michael S. Tsirkin
Quoted 71 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +1
-1@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE );mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 89
Quoted 10 lines
mm/balloon_compaction.c +1
-1-spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/* I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
We really must take a page reference before touching the page.
Quoted 7 lines
Quoted 4 lines
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
Nice catch! Thanks for spotting it. I just have one minor nit. See below
Hmm, As I replied mst's mail, I really cannot understand what you gouys are pointing out.
If we use lock_page in balloon_page_dequeue, I agree it's deadlock but we used trylock_page so it's not a deadlock.
About the page refcount, we don't need take a page reference because if one of page in the list was isolated for migration, the page shouldn't stay in b_dev_info->pages list so balloon_page_dequeue cannot touch the page.
Could you elaborate it more detail if I have missed something? It's a stable patch so I want to be careful.
Thanks.
balloon_page_isolate
{
trylock_page(page)
spin_lock_irqsave(&b_dev_info->pages_lock)
list_del(&page->lru);
}balloon_page_dequeue
{
spin_lock_irqsave(&b_dev_info->pages_lock)
list_for_each_entry_safe(page, tmp, &b_dev_info->pages, lru) {
trylock_page(page)
}}
Michael S. Tsirkin
maintainerRe: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Rafael AquiniQuoted 170 lines · Rafael Aquini
Quoted 94 lines · Michael S. Tsirkin
Quoted 71 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +1
-1@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE );mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 89
Quoted 10 lines
mm/balloon_compaction.c +1
-1-spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/* I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
Quoted 7 lines
We really must take a page reference before touching the page.
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
Nice catch! Thanks for spotting it. I just have one minor nit. See below
Quoted 51 lines
mm/balloon_compaction.c +25
-2@@ -56,12 +56,34 @@ EXPORT_SYMBOL_GPL(balloon_page_enqueue);*/ struct page *balloon_page_dequeue (struct balloon_dev_info *b_dev_info) { -struct page *page, *tmp; +struct page *page;unsigned long flags;bool dequeued_page; +LIST_HEAD (processed);/* protected by b_dev_info->pages_lock */ dequeued_page = false; -list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) { +/* +* We need to go over b_dev_info->pages and lock each page, +* but b_dev_info->pages_lock must nest within page lock. +* +* To make this safe, remove each page from b_dev_info->pages list +* under b_dev_info->pages_lock, then drop this lock. Once list is +* empty, re-add them also under b_dev_info->pages_lock. +*/ +spin_lock_irqsave (&b_dev_info->pages_lock, flags); +while (!list_empty (&b_dev_info->pages)) { + page =list_first_entry (&b_dev_info->pages,typeof (*page), lru); +/* move to processed list to avoid going over it another time */ +list_move (&page->lru, &processed); + +if (!get_page_unless_zero (page)) +continue ; +/* +* pages_lock nests within page lock, +* so drop it before trylock_page +*/ +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +/* * Block others from accessing the 'page' while we get around * establishing additional references and preparing the 'page' @@ -72,6 +94,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)if (!PagePrivate (page)) {/* raced with isolation */ unlock_page (page); +put_page (page);continue ; }#endif mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 108
Quoted 6 lines
mm/balloon_compaction.c +1
-0@@ -80,11 +103,18 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)__count_vm_event (BALLOON_DEFLATE );spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); +put_page (page); dequeued_page = true;break ;^^^^[1]Quoted 5 lines
mm/balloon_compaction.c +2
-0} +put_page (page); +spin_lock_irqsave (&b_dev_info->pages_lock, flags); }
Quoted 5 lines
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 115
Quoted 2 lines
mm/balloon_compaction.c +2
-0+/* re-add remaining entries */ +list_splice (&processed, &b_dev_info->pages);By breaking the loop at its ordinary and expected way-out case [1] we'll hit
list_splicewithout holdingb_dev_info->pages_lock, won't we?
Ouch. right.
Quoted 2 lines
perhaps by adding the following on top of your patch we can address that pickle aforementioned:
I'd rather just goto outside or return.
But maybe Minchan is right and the original patch is ok. I still need to go into this.
Quoted 26 lines
Cheers!
Rafael
--mm/balloon_compaction.c +1
-1@@ -58,7 +58,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) {struct page *page;unsigned long flags; -bool dequeued_page; +bool dequeued_page, locked;LIST_HEAD (processed);/* protected by b_dev_info->pages_lock */ dequeued_page = false;mm/balloon_compaction.c +1
-0@@ -105,13 +105,17 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)unlock_page (page );put_page (page); dequeued_page = true; + locked = false;break ; }
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct · line 121
Quoted 10 lines
mm/balloon_compaction.c +3
-0put_page (page);spin_lock_irqsave (&b_dev_info->pages_lock, flags); + locked = true; }/* re-add remaining entries */ +if (!locked) +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_splice (&processed, &b_dev_info->pages);spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
Quoted 177 lines · Michael S. Tsirkin
Quoted 170 lines · Rafael Aquini
Quoted 94 lines · Michael S. Tsirkin
Quoted 71 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +1
-1@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE );mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 89
Quoted 10 lines
mm/balloon_compaction.c +1
-1-spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/* I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
Quoted 7 lines
We really must take a page reference before touching the page.
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
Nice catch! Thanks for spotting it. I just have one minor nit. See below
Quoted 51 lines
mm/balloon_compaction.c +25
-2@@ -56,12 +56,34 @@ EXPORT_SYMBOL_GPL(balloon_page_enqueue);*/ struct page *balloon_page_dequeue (struct balloon_dev_info *b_dev_info) { -struct page *page, *tmp; +struct page *page;unsigned long flags;bool dequeued_page; +LIST_HEAD (processed);/* protected by b_dev_info->pages_lock */ dequeued_page = false; -list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) { +/* +* We need to go over b_dev_info->pages and lock each page, +* but b_dev_info->pages_lock must nest within page lock. +* +* To make this safe, remove each page from b_dev_info->pages list +* under b_dev_info->pages_lock, then drop this lock. Once list is +* empty, re-add them also under b_dev_info->pages_lock. +*/ +spin_lock_irqsave (&b_dev_info->pages_lock, flags); +while (!list_empty (&b_dev_info->pages)) { + page =list_first_entry (&b_dev_info->pages,typeof (*page), lru); +/* move to processed list to avoid going over it another time */ +list_move (&page->lru, &processed); + +if (!get_page_unless_zero (page)) +continue ; +/* +* pages_lock nests within page lock, +* so drop it before trylock_page +*/ +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +/* * Block others from accessing the 'page' while we get around * establishing additional references and preparing the 'page' @@ -72,6 +94,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)if (!PagePrivate (page)) {/* raced with isolation */ unlock_page (page); +put_page (page);continue ; }#endif mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 108
Quoted 6 lines
mm/balloon_compaction.c +1
-0@@ -80,11 +103,18 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)__count_vm_event (BALLOON_DEFLATE );spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); +put_page (page); dequeued_page = true;break ;^^^^[1]Quoted 5 lines
mm/balloon_compaction.c +2
-0} +put_page (page); +spin_lock_irqsave (&b_dev_info->pages_lock, flags); }Quoted 5 lines
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 115
Quoted 2 lines
mm/balloon_compaction.c +2
-0+/* re-add remaining entries */ +list_splice (&processed, &b_dev_info->pages);By breaking the loop at its ordinary and expected way-out case [1] we'll hit
list_splicewithout holdingb_dev_info->pages_lock, won't we?
Quoted 9 lines
Ouch. right.
Quoted 2 lines
perhaps by adding the following on top of your patch we can address that pickle aforementioned:
I'd rather just goto outside or return.
But maybe Minchan is right and the original patch is ok. I still need to go into this.
I went back and did follow up Minchan's argument and his old patch and, yes, I think we're fine there because, as he states, we're using trylock_page() in those points page lock will nest into b_dev_info->pages_lock. OTOH, I understood your work on making sure we would follow up lock nesting order with strict correctness.
As Minchan's approach keeps the code simpler, I'm voting for it.
Cheers!
-- Rafael
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Michael S. TsirkinQuoted 177 lines · Michael S. Tsirkin
Quoted 170 lines · Rafael Aquini
Quoted 94 lines · Michael S. Tsirkin
Quoted 71 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +1
-1@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE );mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 89
Quoted 10 lines
mm/balloon_compaction.c +1
-1-spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/* I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
Quoted 7 lines
We really must take a page reference before touching the page.
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
Nice catch! Thanks for spotting it. I just have one minor nit. See below
Quoted 51 lines
mm/balloon_compaction.c +25
-2@@ -56,12 +56,34 @@ EXPORT_SYMBOL_GPL(balloon_page_enqueue);*/ struct page *balloon_page_dequeue (struct balloon_dev_info *b_dev_info) { -struct page *page, *tmp; +struct page *page;unsigned long flags;bool dequeued_page; +LIST_HEAD (processed);/* protected by b_dev_info->pages_lock */ dequeued_page = false; -list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) { +/* +* We need to go over b_dev_info->pages and lock each page, +* but b_dev_info->pages_lock must nest within page lock. +* +* To make this safe, remove each page from b_dev_info->pages list +* under b_dev_info->pages_lock, then drop this lock. Once list is +* empty, re-add them also under b_dev_info->pages_lock. +*/ +spin_lock_irqsave (&b_dev_info->pages_lock, flags); +while (!list_empty (&b_dev_info->pages)) { + page =list_first_entry (&b_dev_info->pages,typeof (*page), lru); +/* move to processed list to avoid going over it another time */ +list_move (&page->lru, &processed); + +if (!get_page_unless_zero (page)) +continue ; +/* +* pages_lock nests within page lock, +* so drop it before trylock_page +*/ +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +/* * Block others from accessing the 'page' while we get around * establishing additional references and preparing the 'page' @@ -72,6 +94,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)if (!PagePrivate (page)) {/* raced with isolation */ unlock_page (page); +put_page (page);continue ; }#endif mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 108
Quoted 6 lines
mm/balloon_compaction.c +1
-0@@ -80,11 +103,18 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)__count_vm_event (BALLOON_DEFLATE );spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); +put_page (page); dequeued_page = true;break ;^^^^[1]Quoted 5 lines
mm/balloon_compaction.c +2
-0} +put_page (page); +spin_lock_irqsave (&b_dev_info->pages_lock, flags); }Quoted 5 lines
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 115
Quoted 2 lines
mm/balloon_compaction.c +2
-0+/* re-add remaining entries */ +list_splice (&processed, &b_dev_info->pages);By breaking the loop at its ordinary and expected way-out case [1] we'll hit
list_splicewithout holdingb_dev_info->pages_lock, won't we?
Quoted 9 lines
Ouch. right.
Quoted 2 lines
perhaps by adding the following on top of your patch we can address that pickle aforementioned:
I'd rather just goto outside or return.
But maybe Minchan is right and the original patch is ok. I still need to go into this.
I went back and did follow up Minchan's argument and his old patch and, yes, I think we're fine there because, as he states, we're using trylock_page() in those points page lock will nest into b_dev_info->pages_lock. OTOH, I understood your work on making sure we would follow up lock nesting order with strict correctness.
As Minchan's approach keeps the code simpler, I'm voting for it.
Cheers!
-- Rafael
Re: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Michael S. TsirkinQuoted 94 lines · Michael S. Tsirkin
Quoted 71 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +1
-1@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE );mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 89
Quoted 10 lines
mm/balloon_compaction.c +1
-1-spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/* I think this will cause deadlocks.
pages_locknow nests within page lock,balloon_page_putbacknests them in the reverse order.Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
Quoted 7 lines
We really must take a page reference before touching the page.
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
Nice catch! Thanks for spotting it. I just have one minor nit. See below
Quoted 51 lines
mm/balloon_compaction.c +25
-2@@ -56,12 +56,34 @@ EXPORT_SYMBOL_GPL(balloon_page_enqueue);*/ struct page *balloon_page_dequeue (struct balloon_dev_info *b_dev_info) { -struct page *page, *tmp; +struct page *page;unsigned long flags;bool dequeued_page; +LIST_HEAD (processed);/* protected by b_dev_info->pages_lock */ dequeued_page = false; -list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) { +/* +* We need to go over b_dev_info->pages and lock each page, +* but b_dev_info->pages_lock must nest within page lock. +* +* To make this safe, remove each page from b_dev_info->pages list +* under b_dev_info->pages_lock, then drop this lock. Once list is +* empty, re-add them also under b_dev_info->pages_lock. +*/ +spin_lock_irqsave (&b_dev_info->pages_lock, flags); +while (!list_empty (&b_dev_info->pages)) { + page =list_first_entry (&b_dev_info->pages,typeof (*page), lru); +/* move to processed list to avoid going over it another time */ +list_move (&page->lru, &processed); + +if (!get_page_unless_zero (page)) +continue ; +/* +* pages_lock nests within page lock, +* so drop it before trylock_page +*/ +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +/* * Block others from accessing the 'page' while we get around * establishing additional references and preparing the 'page' @@ -72,6 +94,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)if (!PagePrivate (page)) {/* raced with isolation */ unlock_page (page); +put_page (page);continue ; }#endif
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 108
Quoted 6 lines
mm/balloon_compaction.c +1
-0@@ -80,11 +103,18 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)__count_vm_event (BALLOON_DEFLATE );spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); +put_page (page); dequeued_page = true;break ;
^^^^[1]
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 115
Quoted 7 lines
mm/balloon_compaction.c +4
-0} +put_page (page); +spin_lock_irqsave (&b_dev_info->pages_lock, flags); } +/* re-add remaining entries */ +list_splice (&processed, &b_dev_info->pages);
By breaking the loop at its ordinary and expected way-out case [1] we'll hit list_splice without holding b_dev_info->pages_lock, won't we?
perhaps by adding the following on top of your patch we can address that pickle aforementioned:
Cheers!
Rafael
--
mm/balloon_compaction.c +0 -0
@@ -58,7 +58,7 @@ struct page *balloon_page_dequeue(struct
balloon_dev_info *b_dev_info)
{
struct page *page;
unsigned long flags;
- bool dequeued_page;
+ bool dequeued_page, locked;
LIST_HEAD(processed); /* protected by b_dev_info->pages_lock */
dequeued_page = false;
@@ -105,13 +105,17 @@ struct page *balloon_page_dequeue(struct
balloon_dev_info *b_dev_info)
unlock_page(page);
put_page(page);
dequeued_page = true;
+ locked = false;
break;
}
put_page(page);
spin_lock_irqsave(&b_dev_info->pages_lock, flags);
+ locked = true;
}
/* re-add remaining entries */
+ if (!locked)
+ spin_lock_irqsave(&b_dev_info->pages_lock, flags);
list_splice(&processed, &b_dev_info->pages);
spin_unlock_irqrestore(&b_dev_info->pages_lock, flags);Michael S. Tsirkin
maintainerRe: [PATCH 2/2] virtio_balloon: fix race between migration and ballooning
In reply to Minchan KimQuoted 71 lines · Minchan Kim
In
balloon_page_dequeue,pages_lockshould cover the loop (ie,list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and thenlist_delby isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exceptionDumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabledCc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)mm/balloon_compaction.c +1
-1@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around @@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE );
mm/balloon_compaction.c struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) · line 89
Quoted 10 lines
mm/balloon_compaction.c +1
-1-spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/*
I think this will cause deadlocks.
pages_lock now nests within page lock, balloon_page_putback nests them in the reverse order.
Did you test this with lockdep? You really should for locking changes, and I'd expect it to warn about this.
Also, there's another issue there I think: after isolation page could also get freed before we try to lock it.
We really must take a page reference before touching the page.
I think we need something like the below to fix this issue. Could you please try this out, and send Tested-by? I will repost as a proper patch if this works for you.
mm/balloon_compaction.c +32 -2
@@ -56,12 +56,34 @@ EXPORT_SYMBOL_GPL(balloon_page_enqueue);*/ struct page *balloon_page_dequeue (struct balloon_dev_info *b_dev_info) { -struct page *page, *tmp; +struct page *page;unsigned long flags;bool dequeued_page; +LIST_HEAD (processed);/* protected by b_dev_info->pages_lock */ dequeued_page = false; -list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) { +/* +* We need to go over b_dev_info->pages and lock each page, +* but b_dev_info->pages_lock must nest within page lock. +* +* To make this safe, remove each page from b_dev_info->pages list +* under b_dev_info->pages_lock, then drop this lock. Once list is +* empty, re-add them also under b_dev_info->pages_lock. +*/ +spin_lock_irqsave (&b_dev_info->pages_lock, flags); +while (!list_empty (&b_dev_info->pages)) { + page =list_first_entry (&b_dev_info->pages,typeof (*page), lru); +/* move to processed list to avoid going over it another time */ +list_move (&page->lru, &processed); + +if (!get_page_unless_zero (page)) +continue ; +/* +* pages_lock nests within page lock, +* so drop it before trylock_page +*/ +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +/* * Block others from accessing the 'page' while we get around * establishing additional references and preparing the 'page'
@@ -72,6 +94,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)if (!PagePrivate (page)) {/* raced with isolation */ unlock_page (page); +put_page (page);continue ; }#endif
@@ -80,11 +103,18 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)__count_vm_event (BALLOON_DEFLATE );spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); +put_page (page); dequeued_page = true;break ; } +put_page (page); +spin_lock_irqsave (&b_dev_info->pages_lock, flags); } +/* re-add remaining entries */ +list_splice (&processed, &b_dev_info->pages); +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags); +if (!dequeued_page) {/* * If we are unable to dequeue a balloon page because the page
PATCH 2/2 virtio_balloon: fix race between migration and ballooning
In reply to Minchan KimIn balloon_page_dequeue, pages_lock should cover the loop (ie, list_for_each_entry_safe). Otherwise, the cursor page could be isolated by compaction and then list_del by isolation could poison the page->lru.{prev,next} so the loop finally could access wrong address like this. This patch fixes the bug.
general protection fault: 0000 [#1] SMP Dumping ftrace buffer: (ftrace buffer empty) Modules linked in: CPU: 2 PID: 82 Comm: vballoon Not tainted 4.4.0-rc5-mm1-access_bit+ #1906 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011 task: ffff8800a7ff0000 ti: ffff8800a7fec000 task.ti: ffff8800a7fec000 RIP: 0010:[<ffffffff8115e754>] [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP: 0018:ffff8800a7fefdc0 EFLAGS: 00010246 RAX: ffff88013fff9a70 RBX: ffffea000056fe00 RCX: 0000000000002b7d RDX: ffff88013fff9a70 RSI: ffffea000056fe00 RDI: ffff88013fff9a68 RBP: ffff8800a7fefde8 R08: ffffea000056fda0 R09: 0000000000000000 R10: ffff8800a7fefd90 R11: 0000000000000001 R12: dead0000000000e0 R13: ffffea000056fe20 R14: ffff880138809070 R15: ffff880138809060 FS: 0000000000000000(0000) GS:ffff88013fc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007f229c10e000 CR3: 00000000b8b53000 CR4: 00000000000006a0 Stack: 0000000000000100 ffff880138809088 ffff880138809000 ffff880138809060 0000000000000046 ffff8800a7fefe28 ffffffff812c86d3 ffff880138809020 ffff880138809000 fffffffffff91900 0000000000000100 ffff880138809060 Call Trace: [<ffffffff812c86d3>] leak_balloon+0x93/0x1a0 [<ffffffff812c8bc7>] balloon+0x217/0x2a0 [<ffffffff8143739e>] ? __schedule+0x31e/0x8b0 [<ffffffff81078160>] ? abort_exclusive_wait+0xb0/0xb0 [<ffffffff812c89b0>] ? update_balloon_stats+0xf0/0xf0 [<ffffffff8105b6e9>] kthread+0xc9/0xe0 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 [<ffffffff8143b4af>] ret_from_fork+0x3f/0x70 [<ffffffff8105b620>] ? kthread_park+0x60/0x60 Code: 8d 60 e0 0f 84 af 00 00 00 48 8b 43 20 a8 01 75 3b 48 89 d8 f0 0f ba 28 00 72 10 48 8b 03 f6 c4 08 75 2f 48 89 df e8 8c 83 f9 ff <49> 8b 44 24 20 4d 8d 6c 24 20 48 83 e8 20 4d 39 f5 74 7a 4c 89 RIP [<ffffffff8115e754>] balloon_page_dequeue+0x54/0x130 RSP <ffff8800a7fefdc0> ---[ end trace 43cf28060d708d5f ]--- Kernel panic - not syncing: Fatal exception
Dumping ftrace buffer:
(ftrace buffer empty)
Kernel Offset: disabled
--- mm/balloon_compaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
mm/balloon_compaction.c +2 -2
@@ -61,6 +61,7 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info)bool dequeued_page; dequeued_page = false; +spin_lock_irqsave (&b_dev_info->pages_lock, flags);list_for_each_entry_safe (page, tmp, &b_dev_info->pages, lru) {/* * Block others from accessing the 'page' while we get around
@@ -75,15 +76,14 @@ struct page *balloon_page_dequeue(struct balloon_dev_info *b_dev_info) continue; }#endif -spin_lock_irqsave (&b_dev_info->pages_lock, flags);balloon_page_delete (page);__count_vm_event (BALLOON_DEFLATE ); -spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);unlock_page (page); dequeued_page = true;break ; } } +spin_unlock_irqrestore (&b_dev_info->pages_lock, flags);if (!dequeued_page) {/*
Signature · 1 line
1.9.1
Quoted 36 lines · Minchan Kim
During my compaction-related stuff, I encountered a bug with ballooning.
With repeated inflating and deflating cycle, guest
memory( ie, cat /proc/meminfo | grep MemTotal)is decreased and couldn't be recovered.The reason is
balloon_lockdoesn't coverrelease_pages_balloonso structvirtio_balloonfields could be overwritten by race offill_balloon(e,g, vb->*pfns could be critical).This patch fixes it in my test.
Cc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- drivers/virtio/virtio_balloon.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/virtio/virtio_balloon.c +1
-1@@ -209,8 +209,8 @@ static unsigned leak_balloon(struct virtio_balloon *vb, size_t num)*/ if (vb->num_pfns !=0 )tell_host (vb, vb->deflate_vq); -mutex_unlock (&vb->balloon_lock);release_pages_balloon (vb); +mutex_unlock (&vb->balloon_lock);return num_freed_pages; }Signature · 1 line
1.9.1
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 36 lines · Minchan Kim
During my compaction-related stuff, I encountered a bug with ballooning.
With repeated inflating and deflating cycle, guest
memory( ie, cat /proc/meminfo | grep MemTotal)is decreased and couldn't be recovered.The reason is
balloon_lockdoesn't coverrelease_pages_balloonso structvirtio_balloonfields could be overwritten by race offill_balloon(e,g, vb->*pfns could be critical).This patch fixes it in my test.
Cc <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- drivers/virtio/virtio_balloon.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/virtio/virtio_balloon.c +1
-1@@ -209,8 +209,8 @@ static unsigned leak_balloon(struct virtio_balloon *vb, size_t num)*/ if (vb->num_pfns !=0 )tell_host (vb, vb->deflate_vq); -mutex_unlock (&vb->balloon_lock);release_pages_balloon (vb); +mutex_unlock (&vb->balloon_lock);return num_freed_pages; }Signature · 1 line
1.9.1
Quoted 7 lines · Minchan Kim
During my compaction-related stuff, I encountered a bug with ballooning.
With repeated inflating and deflating cycle, guest
memory( ie, cat /proc/meminfo | grep MemTotal)is decreased and couldn't be recovered.
Quoted 8 lines
The reason is
balloon_lockdoesn't coverrelease_pages_balloonso structvirtio_balloonfields could be overwritten by race offill_balloon(e,g, vb->*pfns could be critical).This patch fixes it in my test.
Cc <[email protected]>Signed-off-by Minchan Kim <[email protected]>
Quoted 20 lines
--- drivers/virtio/virtio_balloon.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/virtio/virtio_balloon.c +1
-1@@ -209,8 +209,8 @@ static unsigned leak_balloon(struct virtio_balloon *vb, size_t num)*/ if (vb->num_pfns !=0 )tell_host (vb, vb->deflate_vq); -mutex_unlock (&vb->balloon_lock);release_pages_balloon (vb); +mutex_unlock (&vb->balloon_lock);return num_freed_pages; }Signature · 1 line
1.9.1
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 7 lines · Minchan Kim
During my compaction-related stuff, I encountered a bug with ballooning.
With repeated inflating and deflating cycle, guest
memory( ie, cat /proc/meminfo | grep MemTotal)is decreased and couldn't be recovered.
Quoted 8 lines
The reason is
balloon_lockdoesn't coverrelease_pages_balloonso structvirtio_balloonfields could be overwritten by race offill_balloon(e,g, vb->*pfns could be critical).This patch fixes it in my test.
Cc <[email protected]>Signed-off-by Minchan Kim <[email protected]>
Quoted 20 lines
--- drivers/virtio/virtio_balloon.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/virtio/virtio_balloon.c +1
-1@@ -209,8 +209,8 @@ static unsigned leak_balloon(struct virtio_balloon *vb, size_t num)*/ if (vb->num_pfns !=0 )tell_host (vb, vb->deflate_vq); -mutex_unlock (&vb->balloon_lock);release_pages_balloon (vb); +mutex_unlock (&vb->balloon_lock);return num_freed_pages; }Signature · 1 line
1.9.1