mm: shmem: do not call PageHWPoison on a ERR-page
Subsystems (2)
PATCH 1/1 mm: shmem: do not call PageHWPoison on a ERR-page
commit b9d02f1bdd98
("mm: shmem: don't truncate page if memory failure happens")
introduced a PageHWPoison(page) call in "shmem_read_mapping_page_gfp" in shmem.c.
Now, if "shmem_getpage_gfp" returns an error, page is set to ERR-page. Therafter, calling PageHWPoison() on this ERR-page, causes KASAN to OOP the kernel :
#############################
BUG: unable to handle page fault for address: fffffffffffffff4 PF: supervisor read access in kernel mode PF: error_code(0x0000) - not-present page PGD 18e019067 P4D 18e019067 PUD 18e01b067 PMD 0 Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC KASAN PTI CPU: 0 PID: 4836 Comm: MATLAB Not tainted 5.15.0+ #18 Hardware name: Dell Inc. Latitude E6320/0GJF11, BIOS A19 11/14/2013 RIP: 0010:shmem_read_mapping_page_gfp+0xd3/0x140 Code: 4c 89 ff e8 6f eb ff ff 5a 59 85 c0 74 64 48 63 d8 48 89 5d 98 be 08 00 00 00 48 89 df e8 e5 67 0c 00 48 89 df e8 6d 5c 0c 00 <48> 8b 13 48 c7 c0 fb ff ff ff f7 c2 00 00 80 00 74 30 48 ba 00 00 RSP: 0018:ffff88806b33f998 EFLAGS: 00010246 RAX: 0000000000000000 RBX: fffffffffffffff4 RCX: ffffffffb7a37ba3 RDX: 0000000000000003 RSI: dffffc0000000000 RDI: fffffffffffffff4 RBP: ffff88806b33fa20 R08: 1ffffffffffffffe R09: fffffffffffffffb R10: fffffbffffffffff R11: 0000000000000001 R12: 1ffff1100d667f33 R13: 00000000001120d2 R14: 00000000000005db R15: ffff88814e64e2d8 FS: 00007f379a384640(0000) GS:ffff888161a00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: fffffffffffffff4 CR3: 00000000269dc004 CR4: 00000000000606f0 Call Trace: <TASK> ? shmem_fault+0x480/0x480 ? __cond_resched+0x1c/0x30 ? __kasan_check_read+0x11/0x20 shmem_get_pages+0x3a4/0xa70 [i915] ? shmem_writeback+0x3b0/0x3b0 [i915] ? i915_gem_object_wait_reservation+0x330/0x560 [i915]
...
...
################################
So, we proceed with PageHWPoison() call, only if the page is not a ERR-page.
P.S. : Alternative (optimised) solution :
===========================================
We could save some CPU cycles, if we directly replace
if (error)
page = ERR_PTR(error);
else
unlock_page(page);with
if (error)
return ERR_PTR(error);--- mm/shmem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
mm/shmem.c +1 -1
@@ -4222,7 +4222,7 @@ struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,else unlock_page (page); -if (PageHWPoison (page)) +if (!IS_ERR (page) &&PageHWPoison (page)) page =ERR_PTR (-EIO );return page;
Signature · 1 line
2.30.2
Quoted 72 lines · Ajay Garg
commit
b9d02f1bdd98
("mm: shmem: don't truncate page if memory failure happens")introduced a PageHWPoison(page) call in "
shmem_read_mapping_page_gfp" inshmem.c.Now, if "
shmem_getpage_gfp" returns an error, page is set to ERR-page. Therafter, callingPageHWPoison()on this ERR-page, causes KASAN to OOP the kernel :#############################
BUG: unable to handle page fault for address: fffffffffffffff4 PF: supervisor read access in kernel mode PF: error_code(0x0000) - not-present page PGD 18e019067 P4D 18e019067 PUD 18e01b067 PMD 0 Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC KASAN PTI CPU: 0 PID: 4836 Comm: MATLAB Not tainted 5.15.0+ #18 Hardware name: Dell Inc. Latitude E6320/0GJF11, BIOS A19 11/14/2013 RIP: 0010:shmem_read_mapping_page_gfp+0xd3/0x140 Code: 4c 89 ff e8 6f eb ff ff 5a 59 85 c0 74 64 48 63 d8 48 89 5d 98 be 08 00 00 00 48 89 df e8 e5 67 0c 00 48 89 df e8 6d 5c 0c 00 <48> 8b 13 48 c7 c0 fb ff ff ff f7 c2 00 00 80 00 74 30 48 ba 00 00 RSP: 0018:ffff88806b33f998 EFLAGS: 00010246 RAX: 0000000000000000 RBX: fffffffffffffff4 RCX: ffffffffb7a37ba3 RDX: 0000000000000003 RSI: dffffc0000000000 RDI: fffffffffffffff4 RBP: ffff88806b33fa20 R08: 1ffffffffffffffe R09: fffffffffffffffb R10: fffffbffffffffff R11: 0000000000000001 R12: 1ffff1100d667f33 R13: 00000000001120d2 R14: 00000000000005db R15: ffff88814e64e2d8 FS: 00007f379a384640(0000) GS:ffff888161a00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: fffffffffffffff4 CR3: 00000000269dc004 CR4: 00000000000606f0 Call Trace: <TASK> ? shmem_fault+0x480/0x480 ? __cond_resched+0x1c/0x30 ? __kasan_check_read+0x11/0x20 shmem_get_pages+0x3a4/0xa70 [i915] ? shmem_writeback+0x3b0/0x3b0 [i915] ? i915_gem_object_wait_reservation+0x330/0x560 [i915]...
...
################################So, we proceed with
PageHWPoison()call, only if the page is not a ERR-page.
P.S. : Alternative (optimised) solution :
===========================================We could save some CPU cycles, if we directly replace
if (error) page = ERR_PTR(error); else unlock_page(page);with
if (error) return ERR_PTR(error);Fixes b9d02f1bdd98 ("mm: shmem: don't truncate page if memory failure happens")Signed-off-by Ajay Garg <[email protected]>--- mm/shmem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)mm/shmem.c +0
-0
mm/shmem.c struct page *shmem_read_mapping_page_gfp(struct address_space *mapping, · line 4226
Quoted 6 lines
mm/shmem.c +1
-1@@ -4222,7 +4222,7 @@ struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,else unlock_page (page); -if (PageHWPoison (page)) +if (!IS_ERR (page) &&PageHWPoison (page)) page =ERR_PTR (-EIO );
How about the following changes since the above if block already do the judgment?
mm/shmem.c +5 -5
@@ -4195,13 +4195,13 @@ struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,BUG_ON (!shmem_mapping (mapping)); error =shmem_getpage_gfp (inode, index, &page,SGP_CACHE , gfp,NULL ,NULL ,NULL ); -if (error) +if (error) { page =ERR_PTR (error); -else + }else {unlock_page (page); - -if (PageHWPoison (page)) - page =ERR_PTR (-EIO ); +if (PageHWPoison (page)) + page =ERR_PTR (-EIO ); + }return page;#else
Quoted 3 lines
mm/shmem.c +0
-0return page;
Quoted 17 lines
How about the following changes since the above if block already do the judgment?
mm/shmem.c +1
-2@@ -4195,13 +4195,13 @@ struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,BUG_ON (!shmem_mapping (mapping)); error =shmem_getpage_gfp (inode, index, &page,SGP_CACHE , gfp,NULL ,NULL ,NULL ); -if (error) +if (error) { page =ERR_PTR (error); -else
mm/shmem.c struct page *shmem_read_mapping_page_gfp(struct · line 4207
Quoted 10 lines
mm/shmem.c +4
-3+ }else {unlock_page (page); - -if (PageHWPoison (page)) - page =ERR_PTR (-EIO ); +if (PageHWPoison (page)) + page =ERR_PTR (-EIO ); + }return page;
You have
* simply put braces (not required for 1-liner if/else blocks)
* contributed nothing to the issue the patch addresses.
I hope this is not a deliberate joke/spam.
Quoted 32 lines · Ajay Garg
Quoted 18 lines
How about the following changes since the above if block already do the judgment?
mm/shmem.c +1
-2@@ -4195,13 +4195,13 @@ struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,BUG_ON (!shmem_mapping (mapping)); error =shmem_getpage_gfp (inode, index, &page,SGP_CACHE , gfp,NULL ,NULL ,NULL ); -if (error) +if (error) { page =ERR_PTR (error); -else mm/shmem.c struct page *shmem_read_mapping_page_gfp(struct · line 4207
Quoted 10 lines
mm/shmem.c +4
-3+ }else {unlock_page (page); - -if (PageHWPoison (page)) - page =ERR_PTR (-EIO ); +if (PageHWPoison (page)) + page =ERR_PTR (-EIO ); + }return page;You have
Quoted 4 lines
* simply put braces (not required for 1-liner if/else blocks)
* contributed nothing to the issue the patch addresses.I hope this is not a deliberate joke/spam.
What? I put "if (PageHWPoison(page))" into the else branch, it should be added braces since it's not 1-line blocks. Why do you think it does not address the issue? What am I missing here?
Quoted 3 lines
What? I put "if (PageHWPoison(page))" into the else branch, it should be added braces since it's not 1-line blocks.
Oh, ok.
Quoted 3 lines
Why do
you think it does not address the issue? What am I missing here?
That might work, however I haven't tested it.
Upto the reviewers to see what they see as the best way going forward.
Thanks and Regards,
Ajay
Quoted 7 lines · Ajay Garg
Quoted 4 lines
What? I put "if (PageHWPoison(page))" into the else branch, it should be added braces since it's not 1-line blocks.
Oh, ok.
Quoted 7 lines
Quoted 3 lines
Why do
you think it does not address the issue? What am I missing here?That might work, however I haven't tested it.
Upto the reviewers to see what they see as the best way going forward.
I just suggest, it's up to you.
Quoted 5 lines
Thanks and Regards,
Ajay
Quoted 97 lines · Muchun Song
Quoted 72 lines · Ajay Garg
commit
b9d02f1bdd98
("mm: shmem: don't truncate page if memory failure happens")introduced a PageHWPoison(page) call in "
shmem_read_mapping_page_gfp" inshmem.c.Now, if "
shmem_getpage_gfp" returns an error, page is set to ERR-page. Therafter, callingPageHWPoison()on this ERR-page, causes KASAN to OOP the kernel :#############################
BUG: unable to handle page fault for address: fffffffffffffff4 PF: supervisor read access in kernel mode PF: error_code(0x0000) - not-present page PGD 18e019067 P4D 18e019067 PUD 18e01b067 PMD 0 Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC KASAN PTI CPU: 0 PID: 4836 Comm: MATLAB Not tainted 5.15.0+ #18 Hardware name: Dell Inc. Latitude E6320/0GJF11, BIOS A19 11/14/2013 RIP: 0010:shmem_read_mapping_page_gfp+0xd3/0x140 Code: 4c 89 ff e8 6f eb ff ff 5a 59 85 c0 74 64 48 63 d8 48 89 5d 98 be 08 00 00 00 48 89 df e8 e5 67 0c 00 48 89 df e8 6d 5c 0c 00 <48> 8b 13 48 c7 c0 fb ff ff ff f7 c2 00 00 80 00 74 30 48 ba 00 00 RSP: 0018:ffff88806b33f998 EFLAGS: 00010246 RAX: 0000000000000000 RBX: fffffffffffffff4 RCX: ffffffffb7a37ba3 RDX: 0000000000000003 RSI: dffffc0000000000 RDI: fffffffffffffff4 RBP: ffff88806b33fa20 R08: 1ffffffffffffffe R09: fffffffffffffffb R10: fffffbffffffffff R11: 0000000000000001 R12: 1ffff1100d667f33 R13: 00000000001120d2 R14: 00000000000005db R15: ffff88814e64e2d8 FS: 00007f379a384640(0000) GS:ffff888161a00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: fffffffffffffff4 CR3: 00000000269dc004 CR4: 00000000000606f0 Call Trace: <TASK> ? shmem_fault+0x480/0x480 ? __cond_resched+0x1c/0x30 ? __kasan_check_read+0x11/0x20 shmem_get_pages+0x3a4/0xa70 [i915] ? shmem_writeback+0x3b0/0x3b0 [i915] ? i915_gem_object_wait_reservation+0x330/0x560 [i915]...
...
################################So, we proceed with
PageHWPoison()call, only if the page is not a ERR-page.
P.S. : Alternative (optimised) solution :
===========================================We could save some CPU cycles, if we directly replace
if (error) page = ERR_PTR(error); else unlock_page(page);with
if (error) return ERR_PTR(error);Fixes b9d02f1bdd98 ("mm: shmem: don't truncate page if memory failure happens")Signed-off-by Ajay Garg <[email protected]>--- mm/shmem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)mm/shmem.c +0
-0mm/shmem.c struct page *shmem_read_mapping_page_gfp(struct address_space *mapping, · line 4226
Quoted 6 lines
mm/shmem.c +1
-1@@ -4222,7 +4222,7 @@ struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,else unlock_page (page); -if (PageHWPoison (page)) +if (!IS_ERR (page) &&PageHWPoison (page)) page =ERR_PTR (-EIO );How about the following changes since the above if block already do the judgment?
mm/shmem.c +1
-1@@ -4195,13 +4195,13 @@ struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,BUG_ON (!shmem_mapping (mapping)); error =shmem_getpage_gfp (inode, index, &page,SGP_CACHE , gfp,NULL ,NULL ,NULL ); -if (error) +if (error) {
mm/shmem.c struct page *shmem_read_mapping_page_gfp(struct · line 4205
Quoted 10 lines
mm/shmem.c +4
-4page =ERR_PTR (error); - else + }else {unlock_page (page); - -if (PageHWPoison (page)) - page =ERR_PTR (-EIO ); +if (PageHWPoison (page)) + page =ERR_PTR (-EIO ); + }
Thanks guys. My bad. It is an apparent bug. IMHO, Muchun's version seems better to me.
Anyway, whatever Andrew picks up could have Reviewed-by: Yang Shi <[email protected]>
Quoted 8 lines
mm/shmem.c +0
-0return page;#else Quoted 4 lines
mm/shmem.c +0
-0return page;
Another report of the issue (different call-flow, but the same error at "shmem_read_mapping_page_gfp") at :
https://lore.kernel.org/lkml/[email protected]/T/#m52d98b6bdb05764524a118b15cec048b34e5ca76
with a tentative approval for the patch :
https://lore.kernel.org/lkml/[email protected]/T/#m24c2888a879d428cde5b34c43838301de544eb7e
Thanks and Regards,
Ajay
Quoted 84 lines · Ajay Garg
commit
b9d02f1bdd98
("mm: shmem: don't truncate page if memory failure happens")introduced a PageHWPoison(page) call in "
shmem_read_mapping_page_gfp" inshmem.c.Now, if "
shmem_getpage_gfp" returns an error, page is set to ERR-page. Therafter, callingPageHWPoison()on this ERR-page, causes KASAN to OOP the kernel :#############################
BUG: unable to handle page fault for address: fffffffffffffff4 PF: supervisor read access in kernel mode PF: error_code(0x0000) - not-present page PGD 18e019067 P4D 18e019067 PUD 18e01b067 PMD 0 Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC KASAN PTI CPU: 0 PID: 4836 Comm: MATLAB Not tainted 5.15.0+ #18 Hardware name: Dell Inc. Latitude E6320/0GJF11, BIOS A19 11/14/2013 RIP: 0010:shmem_read_mapping_page_gfp+0xd3/0x140 Code: 4c 89 ff e8 6f eb ff ff 5a 59 85 c0 74 64 48 63 d8 48 89 5d 98 be 08 00 00 00 48 89 df e8 e5 67 0c 00 48 89 df e8 6d 5c 0c 00 <48> 8b 13 48 c7 c0 fb ff ff ff f7 c2 00 00 80 00 74 30 48 ba 00 00 RSP: 0018:ffff88806b33f998 EFLAGS: 00010246 RAX: 0000000000000000 RBX: fffffffffffffff4 RCX: ffffffffb7a37ba3 RDX: 0000000000000003 RSI: dffffc0000000000 RDI: fffffffffffffff4 RBP: ffff88806b33fa20 R08: 1ffffffffffffffe R09: fffffffffffffffb R10: fffffbffffffffff R11: 0000000000000001 R12: 1ffff1100d667f33 R13: 00000000001120d2 R14: 00000000000005db R15: ffff88814e64e2d8 FS: 00007f379a384640(0000) GS:ffff888161a00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: fffffffffffffff4 CR3: 00000000269dc004 CR4: 00000000000606f0 Call Trace: <TASK> ? shmem_fault+0x480/0x480 ? __cond_resched+0x1c/0x30 ? __kasan_check_read+0x11/0x20 shmem_get_pages+0x3a4/0xa70 [i915] ? shmem_writeback+0x3b0/0x3b0 [i915] ? i915_gem_object_wait_reservation+0x330/0x560 [i915]...
...
################################So, we proceed with
PageHWPoison()call, only if the page is not a ERR-page.
P.S. : Alternative (optimised) solution :
===========================================We could save some CPU cycles, if we directly replace
if (error) page = ERR_PTR(error); else unlock_page(page);with
if (error) return ERR_PTR(error);Fixes b9d02f1bdd98 ("mm: shmem: don't truncate page if memory failure happens")Signed-off-by Ajay Garg <[email protected]>--- mm/shmem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)mm/shmem.c +1
-2@@ -4222,7 +4222,7 @@ struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,else unlock_page (page); -if (PageHWPoison (page)) +if (!IS_ERR (page) &&PageHWPoison (page)) page =ERR_PTR (-EIO );return page; --2.30.2
Quoted 6 lines · Ajay Garg
Another report of the issue (different call-flow, but the same error at "
shmem_read_mapping_page_gfp") at :
https://lore.kernel.org/lkml/[email protected]/T/#m52d98b6bdb05764524a118b15cec048b34e5ca76with a tentative approval for the patch :
https://lore.kernel.org/lkml/[email protected]/T/#m24c2888a879d428cde5b34c43838301de544eb7e
Andrew, I've hit this multiple times on resume on my laptop, and it's a very apparent bug in that b9d02f1bdd98 commit that it doesn't factor in error pointers. My oops clearly shows it being one too, with ...fff4 being the dereference address.
Can we get this pushed upstream asap if you're fine with this patch?
Signature · 1 line
Jens Axboe
Quoted 8 lines · Jens Axboe
Quoted 6 lines · Ajay Garg
Another report of the issue (different call-flow, but the same error at "
shmem_read_mapping_page_gfp") at :
https://lore.kernel.org/lkml/[email protected]/T/#m52d98b6bdb05764524a118b15cec048b34e5ca76with a tentative approval for the patch :
https://lore.kernel.org/lkml/[email protected]/T/#m24c2888a879d428cde5b34c43838301de544eb7e
Quoted 6 lines
Andrew, I've hit this multiple times on resume on my laptop, and it's a very apparent bug in that
b9d02f1bdd98commit that it doesn't factor in error pointers. My oops clearly shows it being one too, with ...fff4 being the dereference address.Can we get this pushed upstream asap if you're fine with this patch?
Maybe Andrew is out - Linus, if you follow this thread, there are two proposed fixes for this. It'd be nice to have one of them in -rc1.
Backstory is that commit:
commit b9d02f1bdd98f38e6e5ecacc9786a8f58f3f8b2c
Author: Yang Shi <[email protected]> Date: Fri Nov 5 13:41:10 2021 -0700
mm: shmem: don't truncate page if memory failure happens
adds a pretty obvious error, where PageHWPoison() is called on an error pointer in shmem_read_mapping_page_gfp(). I can trivially hit this oops on resuming the laptop.
Signature · 1 line
Jens Axboe
Quoted 3 lines · Jens Axboe
Maybe Andrew is out - Linus, if you follow this thread, there are two proposed fixes for this. It'd be nice to have one of them in -rc1.
Neither of the fixes were sent to me, and honestly, I think the real issue is that the original commit is just too broken for words.
The error handling in that commit is just entirely lacking.
I've reverted the commit rather than try to fix up the breakage.
Because at some point it's just better to say "that was broken" rather than try to fix it up. And that original commit was too broken to be worth fixing up. It already had one fix for uninitialized variables before it even hit my tree, and then caused problems and had obvious broken error handling.
Linus
Quoted 10 lines · Linus Torvalds
Quoted 3 lines · Jens Axboe
Maybe Andrew is out - Linus, if you follow this thread, there are two proposed fixes for this. It'd be nice to have one of them in -rc1.
Neither of the fixes were sent to me, and honestly, I think the real issue is that the original commit is just too broken for words.
The error handling in that commit is just entirely lacking.
Quoted 7 lines
I've reverted the commit rather than try to fix up the breakage.
Because at some point it's just better to say "that was broken" rather than try to fix it up. And that original commit was too broken to be worth fixing up. It already had one fix for uninitialized variables before it even hit my tree, and then caused problems and had obvious broken error handling.
Thanks, that obviously fixes the problem as far as I'm concerned as well.
Signature · 1 line
Jens Axboe
Quoted 3 lines · Linus Torvalds
Neither of the fixes were sent to me, and honestly, I think the real issue is that the original commit is just too broken for words.
Side note: the one you pointed to (by Ajay), had the comment that it could be done differently as an optimization.
And I very much agree with that, although I think it would be a lot more than an optimization: just doing
if (error)
return ERR_PTR(error);earlier in the function would have avoided the issue entirely, and would have made the code a lot easier to read too.
But what made me decide to just revert it entirely was that the original commit that caused this all also had stuff like this:
- return shmem_getpage(inode, index, pagep, SGP_WRITE);
+ ret = shmem_getpage(inode, index, pagep, SGP_WRITE);
+
+ if (*pagep && PageHWPoison(*pagep)) {
+ unlock_page(*pagep); + put_page(*pagep); + ret = -EIO;
+ }
+
+ return ret;
which is another example of exactly the same issue: ignoring errors, and then acting on other information and creating new errors.
Again, that code should have checked and handled errors first, and then - if there wasn't an error - added that new HWpoison handling.
So that just made me go "this is not worth fixing up, this just needs re-doing", and thus I just went for the revert instead.
Linus
Quoted 33 lines · Linus Torvalds
Quoted 3 lines · Linus Torvalds
Neither of the fixes were sent to me, and honestly, I think the real issue is that the original commit is just too broken for words.
Side note: the one you pointed to (by Ajay), had the comment that it could be done differently as an optimization.
And I very much agree with that, although I think it would be a lot more than an optimization: just doing
if (error) return ERR_PTR(error);earlier in the function would have avoided the issue entirely, and would have made the code a lot easier to read too.
But what made me decide to just revert it entirely was that the original commit that caused this all also had stuff like this:
(quoted diff) +9
-1- return shmem_getpage(inode, index, pagep, SGP_WRITE); + ret = shmem_getpage(inode, index, pagep, SGP_WRITE); + + if (*pagep && PageHWPoison(*pagep)) { + unlock_page(*pagep); + put_page(*pagep); + ret = -EIO; + } + + return ret;
Quoted 8 lines
which is another example of exactly the same issue: ignoring errors, and then acting on other information and creating new errors.
Again, that code should have checked and handled errors first, and then - if there wasn't an error - added that new HWpoison handling.
So that just made me go "this is not worth fixing up, this just needs re-doing", and thus I just went for the revert instead.
I'm so sorry for the inconvenience.
The above snippet is actually ok since if *pagep returned via shmem_getpage()'s parameter is not NULL, then ret is 0. When shmem_getpage() returns error code, *pagep is NULL IIUC. So it actually doesn't ignore errors then create and return new error.
But I do agree it seems tricky for someone who is not familiar with shmem code. And if shmem code is changed in the future it may be error prone. I could rewrite it to:
if (ret < 0)
goto out;
if (*pagep && PageHWPoison(*pagep)) {
unlock_page(*pagep);
put_page(*pagep);
ret = -EIO;
}out:
return ret;
And fold in Ajay's fix (will take Muchun's version which returns error earlier). Hopefully it still can make -rc1. Of course rc2 is fine either.
Quoted 2 lines
Linus
Quoted 3 lines · Yang Shi
The above snippet is actually ok since if *pagep returned via shmem_getpage()'s parameter is not NULL, then ret is 0.
That's a random implementation detail, and is not ok to rely on.
It may or may not be true, and is not part of the rules of error handling.
If a function returns an error, you shouldn't be looking at the other stuff it returned.
Here's a very recent example of the same kind of problem:
https://lore.kernel.org/lkml/163663333331.414.639840290224641315.tip-bot2@tip-bot2/
where people didn't actually look properly at the return value of the function, and instead looked at the page pointers that the function filled in.
See? EXACT same logic. And completely buggy.
Quoted 1 line
When
shmem_getpage()returns error code, *pagep is NULL IIUC.
No.
When a function returns an error code, you check for the error code, and don't rely on weather the function then filled in other data (or left it alone, or whatever).
So the code should
(a) check and handle error returns properly
(b) be legible
That (b) basically means that if it's not entirely trivial (and none of this was entirely trivial), then when you get an error, you just deal with it right away. You return early, and undo anything you need to undo.
You don't do "oh, let's keep that error, and then do something else that maybe also generates an error".
That "don't handle the error directly" was why shmem_read_mapping_page_gfp() was buggy and would cause an oops.
And while the shmem_write_begin() code migth not cause an oops, it had the same fundamental bad pattern.
Error handling is where 99% of all problems occur. But that also means that you should do the obvious thing wrt error handling, and not have some crazy "if function X returned an error, it will have left the return array untouched" which may or may not be true.
When a function returns an error code, you do error handling based on that code. Not on some random other state.
Linus
Quoted 15 lines · Linus Torvalds
Quoted 3 lines · Yang Shi
The above snippet is actually ok since if *pagep returned via shmem_getpage()'s parameter is not NULL, then ret is 0.
That's a random implementation detail, and is not ok to rely on.
It may or may not be true, and is not part of the rules of error handling.
If a function returns an error, you shouldn't be looking at the other stuff it returned.
Here's a very recent example of the same kind of problem:
Quoted 7 lines
https://lore.kernel.org/lkml/163663333331.414.639840290224641315.tip-bot2@tip-bot2/where people didn't actually look properly at the return value of the function, and instead looked at the page pointers that the function filled in.
See? EXACT same logic. And completely buggy.
Yes, I agree it is too fragile to rely on.
Quoted 29 lines
Quoted 2 lines
When
shmem_getpage()returns error code, *pagep is NULL IIUC.No.
When a function returns an error code, you check for the error code, and don't rely on weather the function then filled in other data (or left it alone, or whatever).
So the code should
(a) check and handle error returns properly
(b) be legible
That (b) basically means that if it's not entirely trivial (and none of this was entirely trivial), then when you get an error, you just deal with it right away. You return early, and undo anything you need to undo.
You don't do "oh, let's keep that error, and then do something else that maybe also generates an error".
That "don't handle the error directly" was why
shmem_read_mapping_page_gfp()was buggy and would cause an oops.And while the
shmem_write_begin()code migth not cause an oops, it had the same fundamental bad pattern.
Quoted 7 lines
Error handling is where 99% of all problems occur. But that also means that you should do the obvious thing wrt error handling, and not have some crazy "if function X returned an error, it will have left the return array untouched" which may or may not be true.
When a function returns an error code, you do error handling based on that code. Not on some random other state.
Thanks a lot for the thorough explanation. Preparing a new patch.
Quoted 2 lines
Linus