| Message ID | 20210723012124.3405007-1-lsahlber@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | cifs: fix fallocate when trying to allocate a hole. | expand |
With the minor fix to change the rc from EFBIG to 0 to address the xfstest 071 failure, tentatively merged into cifs-2.6.git for-next pending more testing On Thu, Jul 22, 2021 at 8:21 PM Ronnie Sahlberg <lsahlber@redhat.com> wrote: > > Remove the conditional checking for out_data_len and skipping the fallocate > if it is 0. This is wrong will actually change any legitimate the fallocate > where the entire region is unallocated into a no-op. > > Additionally, before allocating the range, if FALLOC_FL_KEEP_SIZE is set then > we need to clamp the length of the fallocate region as to not extend the size of the file. > > Fixes: 966a3cb7c7db ("cifs: improve fallocate emulation") > Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> > --- > fs/cifs/smb2ops.c | 23 ++++++++++++++++++----- > 1 file changed, 18 insertions(+), 5 deletions(-) > > diff --git a/fs/cifs/smb2ops.c b/fs/cifs/smb2ops.c > index 5cefb5972396..238717654e46 100644 > --- a/fs/cifs/smb2ops.c > +++ b/fs/cifs/smb2ops.c > @@ -3667,11 +3667,6 @@ static int smb3_simple_fallocate_range(unsigned int xid, > (char **)&out_data, &out_data_len); > if (rc) > goto out; > - /* > - * It is already all allocated > - */ > - if (out_data_len == 0) > - goto out; > > buf = kzalloc(1024 * 1024, GFP_KERNEL); > if (buf == NULL) { > @@ -3794,6 +3789,24 @@ static long smb3_simple_falloc(struct file *file, struct cifs_tcon *tcon, > goto out; > } > > + if (keep_size == true) { > + /* > + * We can not preallocate pages beyond the end of the file > + * in SMB2. > + */ > + if (off >= i_size_read(inode)) { > + rc = -EFBIG; > + goto out; > + } > + /* > + * For fallocates that are partially beyond the end of file, > + * clamp len so we only fallocate up to the end of file. > + */ > + if (off + len > i_size_read(inode)) { > + len = i_size_read(inode) - off; > + } > + } > + > if ((keep_size == true) || (i_size_read(inode) >= off + len)) { > /* > * At this point, we are trying to fallocate an internal > -- > 2.30.2 >
diff --git a/fs/cifs/smb2ops.c b/fs/cifs/smb2ops.c index 5cefb5972396..238717654e46 100644 --- a/fs/cifs/smb2ops.c +++ b/fs/cifs/smb2ops.c @@ -3667,11 +3667,6 @@ static int smb3_simple_fallocate_range(unsigned int xid, (char **)&out_data, &out_data_len); if (rc) goto out; - /* - * It is already all allocated - */ - if (out_data_len == 0) - goto out; buf = kzalloc(1024 * 1024, GFP_KERNEL); if (buf == NULL) { @@ -3794,6 +3789,24 @@ static long smb3_simple_falloc(struct file *file, struct cifs_tcon *tcon, goto out; } + if (keep_size == true) { + /* + * We can not preallocate pages beyond the end of the file + * in SMB2. + */ + if (off >= i_size_read(inode)) { + rc = -EFBIG; + goto out; + } + /* + * For fallocates that are partially beyond the end of file, + * clamp len so we only fallocate up to the end of file. + */ + if (off + len > i_size_read(inode)) { + len = i_size_read(inode) - off; + } + } + if ((keep_size == true) || (i_size_read(inode) >= off + len)) { /* * At this point, we are trying to fallocate an internal
Remove the conditional checking for out_data_len and skipping the fallocate if it is 0. This is wrong will actually change any legitimate the fallocate where the entire region is unallocated into a no-op. Additionally, before allocating the range, if FALLOC_FL_KEEP_SIZE is set then we need to clamp the length of the fallocate region as to not extend the size of the file. Fixes: 966a3cb7c7db ("cifs: improve fallocate emulation") Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> --- fs/cifs/smb2ops.c | 23 ++++++++++++++++++----- 1 file changed, 18 insertions(+), 5 deletions(-)