Re: [libvirt] [PATCH v2] storage: Sanitize pool target paths
On 05/21/2010 02:33 PM, Eric Blake wrote:
On 05/21/2010 11:05 AM, Cole Robinson wrote:
> Spurious / in a pool target path makes life difficult for apps using the
> GetVolByPath, and doing other path based comparisons with pools. This
> has caused a few issues for virt-manager users:
>
> https://bugzilla.redhat.com/show_bug.cgi?id=494005
> https://bugzilla.redhat.com/show_bug.cgi?id=593565
>
> Add a new util API which removes spurious /, virFileSanitizePath. Sanitize
> target paths when parsing pool XML, and for paths passed to GetVolByPath.
>
> v2: Leading // must be preserved, properly sanitize path=/, sanitize
> away /./ -> /
Nice - you caught one case that my review did not ("/" as the entire path).
> +/* Remove spurious / characters from a path. The result must be freed */
> +char *
> +virFileSanitizePath(const char *path)
> +{
> + const char *cur = path;
> + char *cleanpath;
> + int idx = 0;
> +
> + cleanpath = strdup(path);
> + if (!cleanpath) {
> + virReportOOMError();
> + return NULL;
> + }
> +
> + /* Starting with // is valid posix, but ///foo == /foo */
> + if (cur[0] == '/' && cur[1] == '/' && cur[2] !=
'/') {
> + cleanpath[0] = '/';
> + cleanpath[1] = '/';
Delete these two lines; since cleanpath was created by strdup(), they
are redundant assignments.
Will do
> + idx = 2;
> + cur += 2;
> + }
> +
> + /* Sanitize path in place */
> + while (*cur != '\0') {
> + if (*cur != '/') {
> + cleanpath[idx++] = *cur++;
> + continue;
> + }
> +
> + /* Skip all extra / */
> + while (*cur == '/') {
> + cur++;
> +
> + /* Resolve away /./ to just / */
> + if (cur[0] == '.' && cur[1] == '/')
> + cur++;
> + }
That doesn't simplify "./a" to "a", nor "b/." to
"b"
Argh, good point.
> +
> + /* Don't add a trailing / unless path is only made of / */
> + if (idx != 0 && *cur == '\0')
> + break;
This incorrectly collapses plain "//" to "/".
Actually we never entered the while() loop with a path of just //, so
this worked in practice.
> +
> + cleanpath[idx++] = '/';
> + }
Maybe a better approach is this pseudo-code:
if (2 leading slashes) {
advance 2
if (end) return
} else if (leading slash)
advance 1
do {
advance past all slashes - they are redundant from previous round
if (./ or .-end) advance 1 and continue
copy all non-slashes
copy one slash
} while (!end)
if (trailing slash)
back up 1
I didn't end up going with this configuration, but the resulting code
isn't too bad, and seems to cover all the tests I could think up.
Updated patch sending shortly.
Thanks,
Cole