Re: [Libguestfs] [PATCH v2] Use less stack.
On Mon, Mar 07, 2016 at 05:12:33PM +0100, Pino Toscano wrote:
On Monday 07 March 2016 11:18:32 Richard W.M. Jones wrote:
> +#if defined(__GNUC__) && GUESTFS_GCC_VERSION >= 40800 /* gcc >= 4.8.0
*/
> +#pragma GCC diagnostic push
> +#pragma GCC diagnostic ignored "-Wstack-usage=10000"
> +#endif
> +
> /* Takes optional arguments, consult optargs_bitmask. */
> int
> do_btrfs_filesystem_resize (const char *filesystem, int64_t size)
Is this because of the argv[MAX_ARGS] here ^ and in do_mkfs_btrfs?
(Similar notes for all the other diagnostic markers added by this
patch.)
Yes.
As slightly related change (although that can be done after this
patch),
something to reduce a bit the stacks in the daemon would be reduce the
MAX_ARGS to values closer (but not too much) than the actual needs.
I see most of the MAX_ARGS=64 cases have no more than 10 arguments
added, even in the worst case...
OK, but I wouldn't want to go too small. There's not any particular
need to reduce stack frames below, say, 4K. The real issues with
large stack frames are:
- JVM & other environments with small per-thread stacks
- arbitrarily large stack frames controlled from untrusted sources
- avoiding using BUFSIZ, PATH_MAX and similar where we don't directly
control those constants, so we might encounter a system where they
are much larger [mostly we've got rid of PATH_MAX for other reasons]
- to a lesser extent the previous point applies even to things we do
control like GUESTFS_MAX_CHUNK_SIZE where we might conceivably wish
to increase them in future.
I have fixed all the other things you noted, and am just testing
everything again.
Thanks,
Rich.
--
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-p2v converts physical machines to virtual machines. Boot with a
live CD or over the network (PXE) and turn machines into KVM guests.
http://libguestfs.org/virt-v2v