On Tue, May 10, 2022 at 02:35:12PM +0200, Karel Zak wrote:
> On Mon, May 09, 2022 at 02:48:15PM +0200, Christian Brauner wrote:
> > One comment about this. We really need to have this interface support
> > giving us mount options like "relatime" back in numeric form (I assume
> > this will be possible.). It is royally annoying having to maintain a
> > mapping table in userspace just to do:
> >
> > relatime -> MS_RELATIME/MOUNT_ATTR_RELATIME
> > ro -> MS_RDONLY/MOUNT_ATTR_RDONLY
> >
> > A library shouldn't be required to use this interface. Conservative
> > low-level software that keeps its shared library dependencies minimal
> > will need to be able to use that interface without having to go to an
> > external library that transforms text-based output to binary form (Which
> > I'm very sure will need to happen if we go with a text-based
> > interface.).
>
> Sounds like David's fsinfo() :-)
>
> We need an interface where the kernel returns a consistent mount table
> entry (more syscalls to get more key=value could be a way how to get
> inconsistent data).
>
> IMHO all the attempts to make a trivial interface will be unsuccessful
> because the mount table is complex (tree) and mixes strings, paths,
> and flags. We will always end with a complex interface or complex
> strings (like the last xatts attempt). There is no 3rd path to go ...
>
> The best would be simplified fsinfo() where userspace defines
> a request (wanted "keys"), and the kernel fills a buffer with data
> separated by some header metadata struct. In this case, the kernel can
> return strings and structs with binary data.
>
>
> I'd love something like:
>
> ssize_t sz;
> fsinfo_query query[] = {
> { .request = FSINFO_MOUNT_PATH },
> { .request = FSINFO_PROPAGATION },
> { .request = FSINFO_CHILDREN_IDS },
> };
>
> sz = fsinfo(dfd, "", AT_EMPTY_PATH,
> &query, ARRAY_SIZE(query),
> buf, sizeof(buf));
>
> for (p = buf; p < buf + sz; ) {
> {
> fsinfo_entry *e = (struct fsinfo_entry) p;
> char *data = p + sizeof(struct fsinfo_entry);
>
> switch(e->request) {
> case FSINFO_MOUNT_PATH:
> printf("mountpoint %s\n", data);
> break;
> case FSINFO_PROPAGATION:
> printf("propagation %x\n", (uintptr_t) data);
> break;
> case FSINFO_CHILDREN_IDS:
> fsinfo_child *x = (fsinfo_child *) data;
> for (i = 0; i < e->count; i++) {
> printf("child: %d\n", x[i].mnt_id);
> }
> break;
> ...
> }
>
> p += sizeof(struct fsinfo_entry) + e->len;
> }
That's pretty much what a multi-xattr get operation looks like.
It's a bit more more intricate in the setup of the request/return
buffer, but otherwise the structure of the code is the same.
I just don't see why we need special purpose interfaces like this
for key/value information when small tweaks to the existing
generic key/value interfaces can provide exactly the same
functionality....
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
