Pali Rohár <pali@xxxxxxxxxx> writes:
> On Monday 30 September 2024 12:25:27 Paulo Alcantara wrote:
>> Pali Rohár <pali@xxxxxxxxxx> writes:
>>
>> > If CREATE was successful but SMB2_OP_SET_REPARSE failed then remove the
>> > intermediate object created by CREATE. Otherwise empty object stay on the
>> > server when reparse call failed.
>> >
>> > This ensures that if the creating of special files is unsupported by the
>> > server then no empty file stay on the server as a result of unsupported
>> > operation.
>> >
>> > Fixes: 102466f303ff ("smb: client: allow creating special files via reparse points")
>> > Signed-off-by: Pali Rohár <pali@xxxxxxxxxx>
>> > ---
>> > fs/smb/client/smb2inode.c | 21 +++++++++++++++++++--
>> > 1 file changed, 19 insertions(+), 2 deletions(-)
>> >
>> > diff --git a/fs/smb/client/smb2inode.c b/fs/smb/client/smb2inode.c
>> > index 11a1c53c64e0..af42f44bdcf4 100644
>> > --- a/fs/smb/client/smb2inode.c
>> > +++ b/fs/smb/client/smb2inode.c
>> > @@ -1205,6 +1205,8 @@ struct inode *smb2_get_reparse_inode(struct cifs_open_info_data *data,
>> > struct cifs_sb_info *cifs_sb = CIFS_SB(sb);
>> > struct cifsFileInfo *cfile;
>> > struct inode *new = NULL;
>> > + int out_buftype[2] = {};
>> > + struct kvec out_iov[2];
>> > struct kvec in_iov[2];
>> > int cmds[2];
>> > int rc;
>> > @@ -1228,7 +1230,7 @@ struct inode *smb2_get_reparse_inode(struct cifs_open_info_data *data,
>> > cmds[1] = SMB2_OP_POSIX_QUERY_INFO;
>> > cifs_get_writable_path(tcon, full_path, FIND_WR_ANY, &cfile);
>> > rc = smb2_compound_op(xid, tcon, cifs_sb, full_path, &oparms,
>> > - in_iov, cmds, 2, cfile, NULL, NULL, NULL);
>> > + in_iov, cmds, 2, cfile, out_iov, out_buftype, NULL);
>> > if (!rc) {
>> > rc = smb311_posix_get_inode_info(&new, full_path,
>> > data, sb, xid);
>> > @@ -1237,12 +1239,27 @@ struct inode *smb2_get_reparse_inode(struct cifs_open_info_data *data,
>> > cmds[1] = SMB2_OP_QUERY_INFO;
>> > cifs_get_writable_path(tcon, full_path, FIND_WR_ANY, &cfile);
>> > rc = smb2_compound_op(xid, tcon, cifs_sb, full_path, &oparms,
>> > - in_iov, cmds, 2, cfile, NULL, NULL, NULL);
>> > + in_iov, cmds, 2, cfile, out_iov, out_buftype, NULL);
>> > if (!rc) {
>> > rc = cifs_get_inode_info(&new, full_path,
>> > data, sb, xid, NULL);
>> > }
>> > }
>> > +
>> > + if (rc) {
>> > + /*
>> > + * If CREATE was successful but SMB2_OP_SET_REPARSE failed then
>> > + * remove the intermediate object created by CREATE. Otherwise
>> > + * empty object stay on the server when reparse call failed.
>> > + */
>> > + if (((struct smb2_hdr *)out_iov[0].iov_base)->Status == STATUS_SUCCESS &&
>> > + ((struct smb2_hdr *)out_iov[1].iov_base)->Status != STATUS_SUCCESS)
>> > + smb2_unlink(xid, tcon, full_path, cifs_sb, NULL);
>> > + }
>>
>> You should handle the case where ->iov_base is NULL or out_buftype ==
>> CIFS_NO_BUFFER, otherwise you'll end up with a NULL ptr deref.
>
> Ok, thanks for info! I will send v3 with those checks.
>
> Anyway, what does it mean if iov_base stay NULL or out_buftype is
> CIFS_NO_BUFFER? Does it mean that the server has not returned reply for
> that command?
Not necessarily. Just consider a simple case where smb2_compound_op()
would fail to allocate memory for @vars, then it would return -ENOMEM
and you would end up dereferencing ->iov_base which is still NULL. That
is, compound_send_recv() wasn't called and then no response buffers were
set yet.
