| Message ID | 20220223011416.323085-1-lsahlber@redhat.com |
|---|---|
| State | New |
| Headers | show |
| Series | cifs: truncate the inode and mapping when we simulate fcollapse | expand |
Tentatively merged into cifs-2.6.git for-next pending review and testing On Tue, Feb 22, 2022 at 7:14 PM Ronnie Sahlberg <lsahlber@redhat.com> wrote: > > RHBZ:1997367 > > When we collapse a range in smb3_collapse_range() we must make sure > we update the inode size and pagecache accordingly. > > If not, both inode size and pagecahce may be stale until it is refreshed. > > This can be demonstrated for the inode size by running : > > xfs_io -i -f -c "truncate 320k" -c "fcollapse 64k 128k" -c "fiemap -v" \ > /mnt/testfile > > where we can see the result of stale data in the fiemap output. > The third line of the output is wrong, all this data should be truncated. > > EXT: FILE-OFFSET BLOCK-RANGE TOTAL FLAGS > 0: [0..127]: hole 128 > 1: [128..383]: 128..383 256 0x1 > 2: [384..639]: hole 256 > > And the correct output, when the inode size has been updated correctly should > look like this: > > EXT: FILE-OFFSET BLOCK-RANGE TOTAL FLAGS > 0: [0..127]: hole 128 > 1: [128..383]: 128..383 256 0x1 > > Reported-by: Xiaoli Feng <xifeng@redhat.com> > Reported-by: kernel test robot <lkp@intel.com> > Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> > --- > fs/cifs/smb2ops.c | 18 ++++++++++++++---- > 1 file changed, 14 insertions(+), 4 deletions(-) > > diff --git a/fs/cifs/smb2ops.c b/fs/cifs/smb2ops.c > index af5d0830bc8a..891b11576e55 100644 > --- a/fs/cifs/smb2ops.c > +++ b/fs/cifs/smb2ops.c > @@ -25,6 +25,7 @@ > #include "smb2glob.h" > #include "cifs_ioctl.h" > #include "smbdirect.h" > +#include "fscache.h" > #include "fs_context.h" > > /* Change credits for different ops and return the total number of credits */ > @@ -3887,29 +3888,38 @@ static long smb3_collapse_range(struct file *file, struct cifs_tcon *tcon, > { > int rc; > unsigned int xid; > + struct inode *inode; > struct cifsFileInfo *cfile = file->private_data; > + struct cifsInodeInfo *cifsi; > __le64 eof; > > xid = get_xid(); > > - if (off >= i_size_read(file->f_inode) || > - off + len >= i_size_read(file->f_inode)) { > + inode = d_inode(cfile->dentry); > + cifsi = CIFS_I(inode); > + > + if (off >= i_size_read(inode) || > + off + len >= i_size_read(inode)) { > rc = -EINVAL; > goto out; > } > > rc = smb2_copychunk_range(xid, cfile, cfile, off + len, > - i_size_read(file->f_inode) - off - len, off); > + i_size_read(inode) - off - len, off); > if (rc < 0) > goto out; > > - eof = cpu_to_le64(i_size_read(file->f_inode) - len); > + eof = cpu_to_le64(i_size_read(inode) - len); > rc = SMB2_set_eof(xid, tcon, cfile->fid.persistent_fid, > cfile->fid.volatile_fid, cfile->pid, &eof); > if (rc < 0) > goto out; > > rc = 0; > + > + cifsi->server_eof = i_size_read(inode) - len; > + truncate_setsize(inode, cifsi->server_eof); > + fscache_resize_cookie(cifs_inode_cookie(inode), cifsi->server_eof); > out: > free_xid(xid); > return rc; > -- > 2.30.2 >
diff --git a/fs/cifs/smb2ops.c b/fs/cifs/smb2ops.c index af5d0830bc8a..891b11576e55 100644 --- a/fs/cifs/smb2ops.c +++ b/fs/cifs/smb2ops.c @@ -25,6 +25,7 @@ #include "smb2glob.h" #include "cifs_ioctl.h" #include "smbdirect.h" +#include "fscache.h" #include "fs_context.h" /* Change credits for different ops and return the total number of credits */ @@ -3887,29 +3888,38 @@ static long smb3_collapse_range(struct file *file, struct cifs_tcon *tcon, { int rc; unsigned int xid; + struct inode *inode; struct cifsFileInfo *cfile = file->private_data; + struct cifsInodeInfo *cifsi; __le64 eof; xid = get_xid(); - if (off >= i_size_read(file->f_inode) || - off + len >= i_size_read(file->f_inode)) { + inode = d_inode(cfile->dentry); + cifsi = CIFS_I(inode); + + if (off >= i_size_read(inode) || + off + len >= i_size_read(inode)) { rc = -EINVAL; goto out; } rc = smb2_copychunk_range(xid, cfile, cfile, off + len, - i_size_read(file->f_inode) - off - len, off); + i_size_read(inode) - off - len, off); if (rc < 0) goto out; - eof = cpu_to_le64(i_size_read(file->f_inode) - len); + eof = cpu_to_le64(i_size_read(inode) - len); rc = SMB2_set_eof(xid, tcon, cfile->fid.persistent_fid, cfile->fid.volatile_fid, cfile->pid, &eof); if (rc < 0) goto out; rc = 0; + + cifsi->server_eof = i_size_read(inode) - len; + truncate_setsize(inode, cifsi->server_eof); + fscache_resize_cookie(cifs_inode_cookie(inode), cifsi->server_eof); out: free_xid(xid); return rc;
RHBZ:1997367 When we collapse a range in smb3_collapse_range() we must make sure we update the inode size and pagecache accordingly. If not, both inode size and pagecahce may be stale until it is refreshed. This can be demonstrated for the inode size by running : xfs_io -i -f -c "truncate 320k" -c "fcollapse 64k 128k" -c "fiemap -v" \ /mnt/testfile where we can see the result of stale data in the fiemap output. The third line of the output is wrong, all this data should be truncated. EXT: FILE-OFFSET BLOCK-RANGE TOTAL FLAGS 0: [0..127]: hole 128 1: [128..383]: 128..383 256 0x1 2: [384..639]: hole 256 And the correct output, when the inode size has been updated correctly should look like this: EXT: FILE-OFFSET BLOCK-RANGE TOTAL FLAGS 0: [0..127]: hole 128 1: [128..383]: 128..383 256 0x1 Reported-by: Xiaoli Feng <xifeng@redhat.com> Reported-by: kernel test robot <lkp@intel.com> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> --- fs/cifs/smb2ops.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-)