Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.kernel > #1529630
| From | Amir Goldstein <amir73il@gmail.com> |
|---|---|
| Newsgroups | linux.kernel |
| Subject | Re: [PATCH 6/7] ovl: intercept mmap |
| Date | 2016-11-24 20:10 +0100 |
| Message-ID | <sH7zc-6SI-25@gated-at.bofh.it> (permalink) |
| References | <sGZUZ-1zn-13@gated-at.bofh.it> <sH04G-1RN-37@gated-at.bofh.it> <sH2ga-3f2-23@gated-at.bofh.it> <sH6D7-6ey-19@gated-at.bofh.it> |
| Organization | linux.* mail to news gateway |
On Thu, Nov 24, 2016 at 8:03 PM, Amir Goldstein <amir73il@gmail.com> wrote:
> On Thu, Nov 24, 2016 at 3:25 PM, Amir Goldstein <amir73il@gmail.com> wrote:
>> On Thu, Nov 24, 2016 at 12:55 PM, Miklos Szeredi <mszeredi@redhat.com> wrote:
>>> ... in order to handle the corner case when the file is copied up after
>>> being opened read-only and mapped shared.
>>>
>>> Can be verified with the following script:
>>>
>>> - 8< - - - - - 8< - - - - - 8< - - - - - 8< - - - -
>>> cd /
>>> rm -rf /tmp/ovl-rorw-test
>>> mkdir /tmp/ovl-rorw-test
>>> cd /tmp/ovl-rorw-test
>>> cat << EOF > rorw-map.c
>>> #include <fcntl.h>
>>> #include <unistd.h>
>>> #include <string.h>
>>> #include <err.h>
>>> #include <sys/mman.h>
>>>
>>> int main(int argc, char *argv[])
>>> {
>>> int rofd, rwfd;
>>> int ret;
>>> char buf[4];
>>> char *addr;
>>>
>>> rofd = open(argv[1], O_RDONLY);
>>> if (rofd == -1)
>>> err(1, "ro open");
>>>
>>> addr = mmap(NULL, 4, PROT_READ, MAP_SHARED, rofd, 0);
>>> if (addr == MAP_FAILED)
>>> err(1, "ro mmap");
>>>
>>> if (memcmp(addr, "bubu", 4) == 0)
>>> errx(1, "identical startup data");
>>>
>>> rwfd = open(argv[1], O_WRONLY);
>>> if (rwfd == -1)
>>> err(1, "rw open");
>>>
>>> ret = write(rwfd, "bubu", 4);
>>> if (ret == -1)
>>> err(1, "write");
>>> if (ret < 4)
>>> errx(1, "short write");
>>>
>>> if (memcmp(addr, "bubu", 4) != 0)
>>> errx(1, "bad mmap data");
>>>
>>> ret = read(rofd, buf, 4);
>>> if (ret == -1)
>>> err(1, "read");
>>> if (ret < 4)
>>> errx(1, "short read");
>>> if (memcmp(buf, "bubu", 4) != 0)
>>> errx(1, "bad read data");
>>>
>>> return 0;
>>> }
>>> EOF
>>
>>
>> Good timing :-)
>> I just started working on an xfstest this morning to bring this corner
>> case into attention.
>> Once I get the xfs_io commands in order, you could use them for a
>> shorter commit message.
>> Unless you like it that way...
>>
>>> gcc -o rorw-map rorw-map.c
>>> mkdir -p mnt lower upper work
>>> echo baba > lower/foo
>>> mount -t overlay overlay -olowerdir=lower,upperdir=upper,workdir=work mnt
>>> ./rorw-map mnt/foo
>
> Well, there is a bug/feature is xfs_io that puts it in a spin when an
> open command is
> used from command line, so I will need to fix that before sumbitting
> the xfstest, but for
> documentation purpose, you can use the following command in place of rorw-map:
>
> $ xfs_io -r -c "mmap -r 0 4" -c "open foo" -c "pwrite -S 0x61 0 4" -c
> "mread -v 0 4" foo |tail -n 1
> 00000000: 62 61 62 61 baba
>
> Naturally, 'baba' is bad. 'aaaa' is good.
>
Or maybe it was meant to be used this way, which does not blow up:
$ xfs_io << EOF
open -r foo
mmap -r 0 4
open foo
pwrite -S 0x61 0 4
mread -v 0 4
EOF
>>> umount mnt
>>> - 8< - - - - - 8< - - - - - 8< - - - - - 8< - - - -
>>>
>>> No output means success, "rorw-map: bad mmap data" means failure.
>>>
>>> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
>>> ---
>>> fs/overlayfs/inode.c | 23 ++++++++++++++++++++++-
>>> 1 file changed, 22 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/fs/overlayfs/inode.c b/fs/overlayfs/inode.c
>>> index 09c6f99bd5db..25b31c6ebe9e 100644
>>> --- a/fs/overlayfs/inode.c
>>> +++ b/fs/overlayfs/inode.c
>>> @@ -11,6 +11,7 @@
>>> #include <linux/slab.h>
>>> #include <linux/xattr.h>
>>> #include <linux/posix_acl.h>
>>> +#include <linux/mm.h>
>>> #include <linux/module.h>
>>> #include <linux/file.h>
>>> #include <linux/hashtable.h>
>>> @@ -392,6 +393,25 @@ static ssize_t ovl_read_iter(struct kiocb *iocb, struct iov_iter *to)
>>> return ret;
>>> }
>>>
>>> +static int ovl_mmap(struct file *file, struct vm_area_struct *vma)
>>> +{
>>> + if (likely(ovl_file_is_lower(file)))
>>> + return OVL_CALL_REAL_FOP(file, mmap(file, vma));
>>> +
>>> + file = filp_clone_open(file);
>>> + if (IS_ERR(file))
>>> + return PTR_ERR(file);
>>> +
>>> + fput(vma->vm_file);
>>> + /* transfer ref: */
>>> + vma->vm_file = file;
>>> +
>>> + if (!file->f_op->mmap)
>>> + return -ENODEV;
>>> +
>>> + return file->f_op->mmap(file, vma);
>>> +}
>>> +
>>> static struct ovl_fops *ovl_fops_find(const struct file_operations *orig)
>>> {
>>> struct ovl_fops *ofop;
>>> @@ -431,9 +451,10 @@ static struct ovl_fops *ovl_fops_get(struct file *file)
>>> /* Intercept these: */
>>> if (orig->read_iter)
>>> ofop->fops.read_iter = ovl_read_iter;
>>> + if (orig->mmap)
>>> + ofop->fops.mmap = ovl_mmap;
>>>
>>> /* These will need to be intercepted: */
>>> - ofop->fops.mmap = orig->mmap;
>>> ofop->fops.fsync = orig->fsync;
>>>
>>> /*
>>> --
>>> 2.5.5
>>>
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in
>>> the body of a message to majordomo@vger.kernel.org
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
Back to linux.kernel | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread
[PATCH 0/7] overlayfs: fix ro/rw fd data inconsistecies Miklos Szeredi <mszeredi@redhat.com> - 2016-11-24 12:00 +0100
[PATCH 0/7] overlayfs: fix ro/rw fd data inconsistecies Miklos Szeredi <mszeredi@redhat.com> - 2016-11-24 12:00 +0100
Re: [PATCH 0/7] overlayfs: fix ro/rw fd data inconsistecies Miklos Szeredi <miklos@szeredi.hu> - 2016-11-24 15:20 +0100
Re: [PATCH 0/7] overlayfs: fix ro/rw fd data inconsistecies Amir Goldstein <amir73il@gmail.com> - 2016-11-24 22:00 +0100
[PATCH 3/7] mm: ovl: copy-up on MAP_SHARED Miklos Szeredi <mszeredi@redhat.com> - 2016-11-24 12:00 +0100
[PATCH 4/7] ovl: add infrastructure for intercepting file ops Miklos Szeredi <mszeredi@redhat.com> - 2016-11-24 12:00 +0100
Re: [PATCH 4/7] ovl: add infrastructure for intercepting file ops Amir Goldstein <amir73il@gmail.com> - 2016-11-24 13:00 +0100
Re: [PATCH 4/7] ovl: add infrastructure for intercepting file ops Miklos Szeredi <miklos@szeredi.hu> - 2016-11-24 13:10 +0100
Re: [PATCH 4/7] ovl: add infrastructure for intercepting file ops Amir Goldstein <amir73il@gmail.com> - 2016-11-24 14:20 +0100
Re: [PATCH 4/7] ovl: add infrastructure for intercepting file ops Miklos Szeredi <miklos@szeredi.hu> - 2016-11-24 15:00 +0100
Re: [PATCH 4/7] ovl: add infrastructure for intercepting file ops Amir Goldstein <amir73il@gmail.com> - 2016-11-24 15:10 +0100
Re: [PATCH 4/7] ovl: add infrastructure for intercepting file ops Amir Goldstein <amir73il@gmail.com> - 2016-11-25 06:30 +0100
[PATCH 5/7] ovl: intercept read_iter Miklos Szeredi <mszeredi@redhat.com> - 2016-11-24 12:00 +0100
[PATCH 1/7] vfs: allow overlayfs to intercept file ops Miklos Szeredi <mszeredi@redhat.com> - 2016-11-24 12:00 +0100
[PATCH 6/7] ovl: intercept mmap Miklos Szeredi <mszeredi@redhat.com> - 2016-11-24 12:10 +0100
Re: [PATCH 6/7] ovl: intercept mmap Amir Goldstein <amir73il@gmail.com> - 2016-11-24 14:30 +0100
Re: [PATCH 6/7] ovl: intercept mmap Amir Goldstein <amir73il@gmail.com> - 2016-11-24 19:10 +0100
Re: [PATCH 6/7] ovl: intercept mmap Amir Goldstein <amir73il@gmail.com> - 2016-11-24 20:10 +0100
[PATCH 2/7] vfs: export filp_clone_open() Miklos Szeredi <mszeredi@redhat.com> - 2016-11-24 12:10 +0100
[PATCH 7/7] ovl: intercept fsync Miklos Szeredi <mszeredi@redhat.com> - 2016-11-24 12:10 +0100
csiph-web