Skip to contentSkip to conversation

mm/khugepaged: fix iteration in collapse_file

PATCH Applied David Stevens 4 participants 9 messages Jun 7, 2023 – Jun 29, 2023

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

PATCH 1/1 mm/khugepaged: fix iteration in collapse_file

From: David Stevens <[email protected]>

Remove an unnecessary call to xas_set(index) when iterating over the target range in collapse_file. The extra call to xas_set reset the xas cursor to the top of the tree, causing the xas_next call on the next iteration to walk the tree to index instead of advancing to index+1. This returned the same page again, which would cause collapse_file to fail because the page is already locked.

This bug was hidden when CONFIG_DEBUG_VM was set. When that config was used, the xas_load in a subsequent VM_BUG_ON assert would walk xas from the top of the tree to index, causing the xas_next call on the next loop iteration to advance the cursor as expected.

Fixes a2e17cc2efc7 ("mm/khugepaged: maintain page cache uptodate flag")
---
 mm/khugepaged.c | 1 -
 1 file changed, 1 deletion(-)
mm/khugepaged.c +0 -1
index 6b9d39d65b73..2d0d58fb4e7f 100644
@@ -2070,7 +2070,6 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
 					TTU_IGNORE_MLOCK | TTU_BATCH_FLUSH);
 
 		xas_lock_irq(&xas);
-		xas_set(&xas, index);
 
 		VM_BUG_ON_PAGE(page != xas_load(&xas), page);
 
Signature · 1 line
2.41.0.rc2.161.g9c6817b8e7-goog
Reviewed-by on v1 1/1
#1 lore raw
Quoted 9 lines · David Stevens

From: David Stevens <[email protected]>

Remove an unnecessary call to xas_set(index) when iterating over the target range in collapse_file. The extra call to xas_set reset the xas cursor to the top of the tree, causing the xas_next call on the next iteration to walk the tree to index instead of advancing to index+1. This returned the same page again, which would cause collapse_file to fail because the page is already locked.

Quoted 7 lines

This bug was hidden when CONFIG_DEBUG_VM was set. When that config was used, the xas_load in a subsequent VM_BUG_ON assert would walk xas from the top of the tree to index, causing the xas_next call on the next loop iteration to advance the cursor as expected.

Fixes a2e17cc2efc7 ("mm/khugepaged: maintain page cache uptodate flag")
Signature · 1 line
Peter Xu
#2 lore raw
In reply to David Stevens
Quoted 8 lines · David Stevens

Remove an unnecessary call to xas_set(index) when iterating over the target range in collapse_file. The extra call to xas_set reset the xas cursor to the top of the tree, causing the xas_next call on the next iteration to walk the tree to index instead of advancing to index+1. This returned the same page again, which would cause collapse_file to fail because the page is already locked.

Quoted 7 lines

This bug was hidden when CONFIG_DEBUG_VM was set. When that config was used, the xas_load in a subsequent VM_BUG_ON assert would walk xas from the top of the tree to index, causing the xas_next call on the next loop iteration to advance the cursor as expected.

Fixes a2e17cc2efc7 ("mm/khugepaged: maintain page cache uptodate flag")

This patch seems to be wrong, but I have not investigated why.

It's certainly an interesting and worrying observation, if a CONFIG_DEBUG_VM=y kernel goes a significantly different way.

I almost always do have CONFIG_DEBUG_VM=y, so you won't be surprised that I never saw the issue. But once I ran an mm-everything with this patch in, I hit that VM_BUG_ON_PAGE(page != xas_load(&xas), page) for the first time (after about 2 hours of huge tmpfs swapping load).

As if you have just transferred the problem from DEBUG_VM=n to DEBUG_VM=y. But I then tried a CONFIG_DEBUG_VM off 6.4-rc1 kernel (including the fixee but not this fixer) under similar load, and saw no problem in 14 hours. So I can't even reproduce the bug that is being fixed here: only hit a bug that it introduces.

Hugh

Quoted 18 lines
---
 mm/khugepaged.c | 1 -
 1 file changed, 1 deletion(-)
mm/khugepaged.c +0 -1
index 6b9d39d65b73..2d0d58fb4e7f 100644
@@ -2070,7 +2070,6 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
 					TTU_IGNORE_MLOCK | TTU_BATCH_FLUSH);
 
 		xas_lock_irq(&xas);
-		xas_set(&xas, index);
 
 		VM_BUG_ON_PAGE(page != xas_load(&xas), page);
 
Signature · 1 line
2.41.0.rc2.161.g9c6817b8e7-goog
#3 lore raw
Quoted 20 lines · Hugh Dickins
Quoted 8 lines · David Stevens

Remove an unnecessary call to xas_set(index) when iterating over the target range in collapse_file. The extra call to xas_set reset the xas cursor to the top of the tree, causing the xas_next call on the next iteration to walk the tree to index instead of advancing to index+1. This returned the same page again, which would cause collapse_file to fail because the page is already locked.

Quoted 7 lines

This bug was hidden when CONFIG_DEBUG_VM was set. When that config was used, the xas_load in a subsequent VM_BUG_ON assert would walk xas from the top of the tree to index, causing the xas_next call on the next loop iteration to advance the cursor as expected.

Fixes a2e17cc2efc7 ("mm/khugepaged: maintain page cache uptodate flag")

This patch seems to be wrong, but I have not investigated why.

Quoted 7 lines

It's certainly an interesting and worrying observation, if a CONFIG_DEBUG_VM=y kernel goes a significantly different way.

I almost always do have CONFIG_DEBUG_VM=y, so you won't be surprised that I never saw the issue. But once I ran an mm-everything with this patch in, I hit that VM_BUG_ON_PAGE(page != xas_load(&xas), page) for the first time (after about 2 hours of huge tmpfs swapping load).

Is the particular workload one you can share? I haven't hit that assert so far with my tests.

Also, I'm a little surprised that this is the assert which is being hit. My patch series didn't make that many changes to the first loop of the function, and the changes it did make were mostly about missing pages, not present pages.

Quoted 5 lines

As if you have just transferred the problem from DEBUG_VM=n to DEBUG_VM=y. But I then tried a CONFIG_DEBUG_VM off 6.4-rc1 kernel (including the fixee but not this fixer) under similar load, and saw no problem in 14 hours. So I can't even reproduce the bug that is being fixed here: only hit a bug that it introduces.

The bug this fixes isn't a crash - it's the fact that khugepage becomes nearly unable to collapse pages. Specifically, it can only collapse if there is exactly one present page, and that page is at index 511. Since MADV_COLLAPSE uses the same code path, it's easy to reproduce the bug I'm trying to fix with this program:

#define _GNU_SOURCE
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <unistd.h>

#define THP_SIZE (2 * 1024 * 1024)

#ifndef MADV_COLLAPSE
#define MADV_COLLAPSE 25
#endif

int main() {
int memfd = memfd_create("memfd", MFD_CLOEXEC);
assert(memfd >= 0);

  int ret = ftruncate(memfd, THP_SIZE);
  assert(ret >= 0);

  char *addr = mmap(NULL, THP_SIZE, PROT_READ|PROT_WRITE, MAP_SHARED, memfd, 0);
  assert(addr != MAP_FAILED);

  addr[0] = 0xff;

ret = madvise(addr, THP_SIZE, MADV_COLLAPSE);
assert(ret == 0);
}

If DEBUG_VM isn't set, then the test will trigger the assert. If DEBUG_VM is set or if this fix is included, then the test will pass.

-David

Quoted 20 lines

Hugh

Quoted 18 lines
---
 mm/khugepaged.c | 1 -
 1 file changed, 1 deletion(-)
mm/khugepaged.c +0 -2
index 6b9d39d65b73..2d0d58fb4e7f 100644
@@ -2070,7 +2070,6 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
                                      TTU_IGNORE_MLOCK | TTU_BATCH_FLUSH);
 
              xas_lock_irq(&xas);
-             xas_set(&xas, index);
 
              VM_BUG_ON_PAGE(page != xas_load(&xas), page);
 
--

2.41.0.rc2.161.g9c6817b8e7-goog

#4 lore raw
Quoted 15 lines · David Stevens
Quoted 14 lines · Hugh Dickins
Quoted 13 lines · David Stevens

Remove an unnecessary call to xas_set(index) when iterating over the target range in collapse_file. The extra call to xas_set reset the xas cursor to the top of the tree, causing the xas_next call on the next iteration to walk the tree to index instead of advancing to index+1. This returned the same page again, which would cause collapse_file to fail because the page is already locked.

This bug was hidden when CONFIG_DEBUG_VM was set. When that config was used, the xas_load in a subsequent VM_BUG_ON assert would walk xas from the top of the tree to index, causing the xas_next call on the next loop iteration to advance the cursor as expected.

Quoted 4 lines
Quoted 4 lines
Quoted 2 lines
Fixes a2e17cc2efc7 ("mm/khugepaged: maintain page cache uptodate flag")

This patch seems to be wrong,

I withdraw the accusation! I haven't thought about the patch itself:
I'd have to learn a bit more about the xarray rules, so cannot quickly ack it myself; but withdraw my implicit nack.

Quoted 11 lines
Quoted 9 lines

but I have not investigated why.

It's certainly an interesting and worrying observation, if a CONFIG_DEBUG_VM=y kernel goes a significantly different way.

I almost always do have CONFIG_DEBUG_VM=y, so you won't be surprised that I never saw the issue. But once I ran an mm-everything with this patch in, I hit that VM_BUG_ON_PAGE(page != xas_load(&xas), page) for the first time (after about 2 hours of huge tmpfs swapping load).

Is the particular workload one you can share?

It's a workload familiar to me, and no secret, but which would take much more time to share than I have to spare; with no great likelihood that you could easily reproduce all the conditions at the end of it.

Quoted 1 line

I haven't hit that assert so far with my tests.

And nor have I since. I've been trying 6.4-rc5 with and without your patch, mm-everything (of a few days ago) with and without your patch, with and without my pte_offset_map changes. Not once have I seen that VM_BUG_ON_PAGE(page != xas_load(&xas), page) again, even with the exact same kernel.

I've no doubt that I did hit it, but I'm now having to assume that it's something very rare (or very rare in my particular testing). And because the first time I saw it coincided with the first time I had in your patch to the last non-blank line above it, I jumped to the conclusion that it was your patch causing it. Not unreasonable at the time, but not justified by later attempts. Perhaps there was something unrelated in that kernel, corrupting an xarray node.

If I ever see it again, I hope I'll be able to spend more time investigating; or if I study that code with more leisure, maybe a hypothesis for a rare race will arise. But for now, we'd better just forget about it.

Quoted 5 lines

Also, I'm a little surprised that this is the assert which is being hit. My patch series didn't make that many changes to the first loop of the function, and the changes it did make were mostly about missing pages, not present pages.

I was suspicious of that patch series, as I said at the time; and might one day want to revert it; but it has not given me any actual problem - that only came with this fix to it.

Quoted 8 lines
Quoted 6 lines

As if you have just transferred the problem from DEBUG_VM=n to DEBUG_VM=y. But I then tried a CONFIG_DEBUG_VM off 6.4-rc1 kernel (including the fixee but not this fixer) under similar load, and saw no problem in 14 hours. So I can't even reproduce the bug that is being fixed here: only hit a bug that it introduces.

The bug this fixes isn't a crash - it's the fact that khugepage

Sorry, yes, I should have re-read your commit message before writing all that.

Thanks,
Hugh

Quoted 59 lines

becomes nearly unable to collapse pages. Specifically, it can only collapse if there is exactly one present page, and that page is at index 511. Since MADV_COLLAPSE uses the same code path, it's easy to reproduce the bug I'm trying to fix with this program:

#define _GNU_SOURCE
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <unistd.h>

#define THP_SIZE (2 * 1024 * 1024)

#ifndef MADV_COLLAPSE
#define MADV_COLLAPSE 25
#endif

int main() {
int memfd = memfd_create("memfd", MFD_CLOEXEC);
assert(memfd >= 0);

  int ret = ftruncate(memfd, THP_SIZE);
  assert(ret >= 0);

  char *addr = mmap(NULL, THP_SIZE, PROT_READ|PROT_WRITE, MAP_SHARED, memfd, 0);
  assert(addr != MAP_FAILED);

  addr[0] = 0xff;

ret = madvise(addr, THP_SIZE, MADV_COLLAPSE);
assert(ret == 0);
}

If DEBUG_VM isn't set, then the test will trigger the assert. If DEBUG_VM is set or if this fix is included, then the test will pass.

-David

Quoted 20 lines

Hugh

Quoted 18 lines
---
 mm/khugepaged.c | 1 -
 1 file changed, 1 deletion(-)
mm/khugepaged.c +0 -2
index 6b9d39d65b73..2d0d58fb4e7f 100644
@@ -2070,7 +2070,6 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
                                      TTU_IGNORE_MLOCK | TTU_BATCH_FLUSH);
 
              xas_lock_irq(&xas);
-             xas_set(&xas, index);
 
              VM_BUG_ON_PAGE(page != xas_load(&xas), page);
 
--

2.41.0.rc2.161.g9c6817b8e7-goog

#5 lore raw

PATCH 1/1 mm/khugepaged: fix regression in collapse_file()

There is no xas_pause(&xas) in collapse_file()'s main loop, at the points where it does xas_unlock_irq(&xas) and then continues.

That would explain why, once two weeks ago and twice yesterday, I have hit the VM_BUG_ON_PAGE(page != xas_load(&xas), page) since "mm/khugepaged:
fix iteration in collapse_file" removed the xas_set(&xas, index) just before it: xas.xa_node could be left pointing to a stale node, if there was concurrent activity on the file which transformed its xarray.

I tried inserting xas_pause()s, but then even bootup crashed on that VM_BUG_ON_PAGE(): there appears to be a subtle "nextness" implicit in xas_pause().

xas_next() and xas_pause() are good for use in simple loops, but not in this one: xas_set() worked well until now, so use xas_set(&xas, index) explicitly at the head of the loop; and change that VM_BUG_ON_PAGE() not to need its own xas_set(), and not to interfere with the xa_state (which would probably stop the crashes from xas_pause(), but I trust that less).

Fixes c8a8f3b4a95a ("mm/khugepaged: fix iteration in collapse_file")

---
Linus, I'm rushing this directly to you, but not really expecting you to put it in at this stage, unless you're very comfortable with it, or perhaps it catches Matthew's eye and gets a quick Ack from him.

The commit being fixed only got in after -rc7, after being held up by my initial report of this crash; but I had to rescind that when I couldn't reproduce it at all. Then yesterday morning it hit again on two machines, and reading XArray Doc reminded me of xas_pause() - seems obvious now. Patch ran for 14 hours overnight on those two machines without a problem.

 mm/khugepaged.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
mm/khugepaged.c +3 -4
index 2d0d58fb4e7f..47b59f2843f6 100644
@@ -1918,9 +1918,9 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
 		}
 	} while (1);
 
-	xas_set(&xas, start);
 	for (index = start; index < end; index++) {
-		page = xas_next(&xas);
+		xas_set(&xas, index);
+		page = xas_load(&xas);
 
 		VM_BUG_ON(index != xas.xa_index);
 		if (is_shmem) {
@@ -1935,7 +1935,6 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
 						result = SCAN_TRUNCATED;
 						goto xa_locked;
 					}
-					xas_set(&xas, index + 1);
 				}
 				if (!shmem_charge(mapping->host, 1)) {
 					result = SCAN_FAIL;
@@ -2071,7 +2070,7 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
 
 		xas_lock_irq(&xas);
 
-		VM_BUG_ON_PAGE(page != xas_load(&xas), page);
+		VM_BUG_ON_PAGE(page != xa_load(xas.xa, index), page);
 
 		/*
 		 * We control three references to the page:
Signature · 1 line
2.35.3
#6 lore raw

Re: [PATCH] mm/khugepaged: fix regression in collapse_file()

Quoted 4 lines · Hugh Dickins

Linus, I'm rushing this directly to you, but not really expecting you to put it in at this stage, unless you're very comfortable with it, or perhaps it catches Matthew's eye and gets a quick Ack from him.

Yeah, the fix and explanation sound fine to me, but the maple tree code confuses me enough and has been subtle enough (as exemplified by this bug and many others) that there's no way I'll apply this just before the final 6.4 without some "Ack, that's obviously correct" from Willy or Liam or somebody.

Willy?

Linus

#7 lore raw

PATCH 1/1 mm/khugepaged: fix regression in collapse_file()

There is no xas_pause(&xas) in collapse_file()'s main loop, at the points where it does xas_unlock_irq(&xas) and then continues.

That would explain why, once two weeks ago and twice yesterday, I have hit the VM_BUG_ON_PAGE(page != xas_load(&xas), page) since "mm/khugepaged:
fix iteration in collapse_file" removed the xas_set(&xas, index) just before it: xas.xa_node could be left pointing to a stale node, if there was concurrent activity on the file which transformed its xarray.

I tried inserting xas_pause()s, but then even bootup crashed on that VM_BUG_ON_PAGE(): there appears to be a subtle "nextness" implicit in xas_pause().

xas_next() and xas_pause() are good for use in simple loops, but not in this one: xas_set() worked well until now, so use xas_set(&xas, index) explicitly at the head of the loop; and change that VM_BUG_ON_PAGE() not to need its own xas_set(), and not to interfere with the xa_state (which would probably stop the crashes from xas_pause(), but I trust that less).

The user-visible effects of this bug (if VM_BUG_ONs are configured out) would be data loss and data leak - potentially - though in practice I expect it is more likely that a subsequent check (e.g. on mapping or on nr_none) would notice an inconsistency, and just abandon the collapse.

Fixes c8a8f3b4a95a ("mm/khugepaged: fix iteration in collapse_file")
---
 mm/khugepaged.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
mm/khugepaged.c +3 -4
index 3beb4ad2ee5e..78c8d5d8b628 100644
@@ -1937,9 +1937,9 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
 		}
 	} while (1);
 
-	xas_set(&xas, start);
 	for (index = start; index < end; index++) {
-		page = xas_next(&xas);
+		xas_set(&xas, index);
+		page = xas_load(&xas);
 
 		VM_BUG_ON(index != xas.xa_index);
 		if (is_shmem) {
@@ -1954,7 +1954,6 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
 						result = SCAN_TRUNCATED;
 						goto xa_locked;
 					}
-					xas_set(&xas, index + 1);
 				}
 				if (!shmem_charge(mapping->host, 1)) {
 					result = SCAN_FAIL;
@@ -2090,7 +2089,7 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
 
 		xas_lock_irq(&xas);
 
-		VM_BUG_ON_PAGE(page != xas_load(&xas), page);
+		VM_BUG_ON_PAGE(page != xa_load(xas.xa, index), page);
 
 		/*
 		 * We control three references to the page:
Signature · 1 line
2.35.3
#8 lore raw

Re: [PATCH hotfix] mm/khugepaged: fix regression in collapse_file()

Quoted 3 lines · Hugh Dickins

There is no xas_pause(&xas) in collapse_file()'s main loop, at the points where it does xas_unlock_irq(&xas) and then continues.

Thanks for reminding me. Now applied,

Linus