Skip to contentSkip to conversation

Define a UNIQUE value for AS_UNEVICTABLE flag

PATCH In review Lee Schermerhorn 4 participants 5 messages Apr 2, 2009 Page cache

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Define a UNIQUE value for AS_UNEVICTABLE flag

[PATCH] Define UNIQUE value of AS_UNEVICTABLE

Needed in 2.6.28, 2.6.29, ...

A new "address_space flag"--AS_MM_ALL_LOCKS--was defined to use the next available AS flag while the Unevictable LRU was under development. The Unevictable LRU was using the same flag and "no one" noticed. Current mainline, since 2.6.28, has same value for two symbolic flag names.

So, define a unique flag value for AS_UNEVICTABLE--up close to the other flags, [at the cost of an additional #ifdef] so we'll notice next time. Note that #ifdef is not actually required, if we don't mind having the unused flag value defined.

Replace #defines with an enum.

 include/linux/pagemap.h |   12 ++++++++----
 1 files changed, 8 insertions(+), 4 deletions(-)
include/linux/pagemap.h +8 -4
Index: linux-2.6.29/include/linux/pagemap.h
===================================================================
@@ -18,9 +18,14 @@
  * Bits in mapping->flags.  The lower __GFP_BITS_SHIFT bits are the page
  * allocation mode flags.
  */
-#define	AS_EIO		(__GFP_BITS_SHIFT + 0)	/* IO error on async write */
-#define AS_ENOSPC	(__GFP_BITS_SHIFT + 1)	/* ENOSPC on async write */
-#define AS_MM_ALL_LOCKS	(__GFP_BITS_SHIFT + 2)	/* under mm_take_all_locks() */
+enum mapping_flags {
+	AS_EIO		= __GFP_BITS_SHIFT + 0,	/* IO error on async write */
+	AS_ENOSPC	= __GFP_BITS_SHIFT + 1,	/* ENOSPC on async write */
+	AS_MM_ALL_LOCKS	= __GFP_BITS_SHIFT + 2,	/* under mm_take_all_locks() */
+#ifdef CONFIG_UNEVICTABLE_LRU
+	AS_UNEVICTABLE	= __GFP_BITS_SHIFT + 3,	/* e.g., ramdisk, SHM_LOCK */
+#endif
+};
 
 static inline void mapping_set_error(struct address_space *mapping, int error)
 {
@@ -33,7 +38,6 @@ static inline void mapping_set_error(str
 }
 
 #ifdef CONFIG_UNEVICTABLE_LRU
-#define AS_UNEVICTABLE	(__GFP_BITS_SHIFT + 2)	/* e.g., ramdisk, SHM_LOCK */
 
 static inline void mapping_set_unevictable(struct address_space *mapping)
 {
#1 lore raw
Quoted 8 lines · On Thu, 02 Apr 2009 12:47:15 -0400 Lee Schermerhorn <[email protected]>…

[PATCH] Define UNIQUE value of AS_UNEVICTABLE

Needed in 2.6.28, 2.6.29, ...

A new "address_space flag"--AS_MM_ALL_LOCKS--was defined to use the next available AS flag while the Unevictable LRU was under development. The Unevictable LRU was using the same flag and "no one" noticed. Current mainline, since 2.6.28, has same value for two symbolic flag names.

argh.

What are the user-observable effects of the bug, and why didn't anyone notice it until now?

#2 lore raw

(cc to Avi)

Quoted 9 lines
Quoted 8 lines

[PATCH] Define UNIQUE value of AS_UNEVICTABLE

Needed in 2.6.28, 2.6.29, ...

A new "address_space flag"--AS_MM_ALL_LOCKS--was defined to use the next available AS flag while the Unevictable LRU was under development.  The Unevictable LRU was using the same flag and "no one" noticed.  Current mainline, since 2.6.28, has same value for two symbolic flag names.

Quoted 4 lines

argh.

What are the user-observable effects of the bug, and why didn't anyone notice it until now?

AS_MM_ALL_LOCKS is used by mmu_notifier. it mean it is used by only kvm. In the other hand, AS_UNEVICTABLE mean unevictable shmem or ramfs.

Then, if shmem opend process use ioctl(KVM_CREATE_VM), unevictable flag on the shmem accidentally turn off.
but, fortunatelly, In modern desktop environment, only KVM control program use above ioctl. then, we can assume this doesn't use shmem and ramfs.

Am I missing anything?

#4 lore raw
In reply to Andrew Morton
Quoted 11 lines · Andrew Morton
Quoted 8 lines · On Thu, 02 Apr 2009 12:47:15 -0400 Lee Schermerhorn <[email protected]>…

[PATCH] Define UNIQUE value of AS_UNEVICTABLE

Needed in 2.6.28, 2.6.29, ...

A new "address_space flag"--AS_MM_ALL_LOCKS--was defined to use the next available AS flag while the Unevictable LRU was under development. The Unevictable LRU was using the same flag and "no one" noticed. Current mainline, since 2.6.28, has same value for two symbolic flag names.

Quoted 4 lines

argh.

What are the user-observable effects of the bug, and why didn't anyone notice it until now?

Well, AS_MM_ALL_LOCKS seems to be used for mmu notifiers. So, I expect I've never enabled it. However, I think that if it got set, all of the pages in all of the vmas that had it set will appear to be unevictable. This would only matter if/when one tried to reclaim/evict them. The pages would probably get stranded on the unevictable lru [until freed] in that case as there would be no scan to rescue them when AS_MM_ALL_LOCKS is cleared.

Not sure about the other way around: flag set as AS_UNEVICTABLE and code examining AS_MM_ALL_LOCKS sees it...

#3 lore raw
In reply to KOSAKI Motohiro
Quoted 21 lines · KOSAKI Motohiro

(cc to Avi)

Quoted 9 lines
Quoted 9 lines

[PATCH] Define UNIQUE value of AS_UNEVICTABLE

Needed in 2.6.28, 2.6.29, ...

A new "address_space flag"--AS_MM_ALL_LOCKS--was defined to use the next available AS flag while the Unevictable LRU was under development. The Unevictable LRU was using the same flag and "no one" noticed. Current mainline, since 2.6.28, has same value for two symbolic flag names.

Quoted 5 lines

argh.

What are the user-observable effects of the bug, and why didn't anyone notice it until now?

AS_MM_ALL_LOCKS is used by mmu_notifier. it mean it is used by only kvm. In the other hand, AS_UNEVICTABLE mean unevictable shmem or ramfs.

Quoted 8 lines

Then, if shmem opend process use ioctl(KVM_CREATE_VM), unevictable flag on the shmem accidentally turn off.
but, fortunatelly, In modern desktop environment, only KVM control program use above ioctl. then, we can assume this doesn't use shmem and ramfs.

Am I missing anything?

Qemu, the main kvm user, indeed doesn't use shm or ramfs. However it is not inconceivable that one day it will, as a way to share memory among guests.

It isn't surprising that no bug was observed, but let's fix it.

(If hugetlbfs is considered unevictable, then qemu can trigger this bug today)

Signature · 2 lines
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.