[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <i4c6xe6jdei2to6kah4kgjehpjlanaqfulju2jzsu5ny2gmegv@2b2oh44oilnj>
Date: Mon, 22 Apr 2024 13:03:45 +0200
From: "Pankaj Raghav (Samsung)" <kernel@...kajraghav.com>
To: Matthew Wilcox <willy@...radead.org>
Cc: linux-xfs@...r.kernel.org, linux-fsdevel@...r.kernel.org,
gost.dev@...sung.com, chandan.babu@...cle.com, hare@...e.de, mcgrof@...nel.org,
djwong@...nel.org, linux-mm@...ck.org, linux-kernel@...r.kernel.org,
david@...morbit.com, akpm@...ux-foundation.org, Pankaj Raghav <p.raghav@...sung.com>
Subject: Re: [PATCH v3 05/11] readahead: allocate folios with
mapping_min_order in readahead
> > @@ -515,7 +562,7 @@ void page_cache_ra_order(struct readahead_control *ractl,
> > if (index & ((1UL << order) - 1))
> > order = __ffs(index);
> > /* Don't allocate pages past EOF */
> > - while (index + (1UL << order) - 1 > limit)
> > + while (order > min_order && index + (1UL << order) - 1 > limit)
> > order--;
>
> This raises an interesting question that I don't know if we have a test
> for. POSIX says that if we mmap, let's say, the first 16kB of a 10kB
> file, then we can store into offset 0-12287, but stores to offsets
> 12288-16383 get a signal (I forget if it's SEGV or BUS). Thus far,
> we've declined to even create folios in the page cache that would let us
> create PTEs for offset 12288-16383, so I haven't paid too much attention
> to this. Now we're going to have folios that extend into that range, so
> we need to be sure that when we mmap(), we only create PTEs that go as
> far as 12287.
>
> Can you check that we have such an fstest, and that we still pass it
> with your patches applied and a suitably large block size?
>
So the mmap is giving the correct SIGBUS error when we try to do this:
dd if=/dev/zero of=./test bs=10k count=1;
xfs_io -c "mmap -w 0 16384" -c "mwrite 13000 10" test
Logs on bs=64k ps=4k system:
root@...ian:/media/test# dd if=/dev/zero of=./test bs=10k count=1;
root@...ian:/media/test# du -sh test
64K test
root@...ian:/media/test# ls -l --block-size=k test
-rw-r--r-- 1 root root 10K Apr 22 10:42 test
root@...ian:/media/test# xfs_io -c "mmap 0 16384" -c "mwrite 13000 10" test
Bus error
The check in filemap_fault takes care of this:
max_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
if (unlikely(index >= max_idx))
return VM_FAULT_SIGBUS;
The same operation for read should also give a bus error, but it didn't.
Further investigation pointed out that the fault_around() does not take
this condition into account for LBS configuration. When I set fault_around_bytes
to 4096, things worked as expected as we skip fault_around for reads.
I have a patch that return SIGBUS also for the following read operation:
dd if=/dev/zero of=./test bs=10k count=1;
xfs_io -c "mmap -r 0 16384" -c "mread 13000 10" test
This is the patch I have for now that fixes fault_around() logic for LBS
configuration:
diff --git a/mm/filemap.c b/mm/filemap.c
index f0c0cfbbd134..259531dd297b 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -3600,12 +3600,15 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf,
}
do {
unsigned long end;
+ unsigned long i_size;
addr += (xas.xa_index - last_pgoff) << PAGE_SHIFT;
vmf->pte += xas.xa_index - last_pgoff;
last_pgoff = xas.xa_index;
end = folio_next_index(folio) - 1;
- nr_pages = min(end, end_pgoff) - xas.xa_index + 1;
+ i_size = DIV_ROUND_UP(i_size_read(mapping->host),
+ PAGE_SIZE) - 1;
+ nr_pages = min3(end, end_pgoff, i_size) - xas.xa_index + 1;
if (!folio_test_large(folio))
ret |= filemap_map_order0_folio(vmf,
I will send a new version of the series this week after doing some more
testing.
Powered by blists - more mailing lists