Message ID | 20171023144646.50107-5-bfoster@redhat.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c index e282fd8..5966cab 100644 --- a/fs/xfs/xfs_log.c +++ b/fs/xfs/xfs_log.c @@ -1453,6 +1453,14 @@ xlog_alloc_log( xlog_get_iclog_buffer_size(mp, log); + error = -EINVAL; + if (log->l_iclog_bufs * log->l_iclog_size > (log->l_logsize >> 1)) { + xfs_warn(mp, + "total iclog buffer size (logbufs * logbsize) cannot exceed %d bytes", + (log->l_logsize >> 1)); + goto out_free_log; + } + /* * Use a NULL block for the extra log buffer used during splits so that * it will trigger errors if we ever try to do IO on it without first
Since mkfs has historically had an issue creating a very small filesystems with a log smaller than the minimum log size, add an extra layer of runtime protection around the log buffer count and size parameters. Restrict the total log buffer size to 1/2 of the physical log size, otherwise fail the mount. The default log buffer count and size is 8 and 32k, respectively. This total size of 256k results in a minimum log size requirement of 512k, which is well outside even the smallest logs created by broken formats. Therefore, this change should only affect users who explicitly attempt to use larger log buffer counts/sizes with such filesystems. Signed-off-by: Brian Foster <bfoster@redhat.com> --- fs/xfs/xfs_log.c | 8 ++++++++ 1 file changed, 8 insertions(+)