On Mon, Apr 07, 2025 at 01:38:24PM +0200, Jan Beulich wrote:
> On 27.03.2025 14:32, Jan Beulich wrote:
> > From their introduction all xc_hypercall_bounce_pre() uses, when they
> > failed, would properly cause exit from the function including cleanup,
> > yet without informing the caller of the failure. Purge the unlock_1
> > label for being both pointless and mis-named.
> >
> > An earlier attempt to switch to the usual split between return value and
> > errno wasn't quite complete.
> >
> > HWP work made the cleanup of the "available governors" array
> > conditional, neglecting the fact that the condition used may not be the
> > condition that was used to allocate the buffer (as the structure field
> > is updated upon getting back EAGAIN). Throughout the function, use the
> > local variable being introduced to address that.
> >
> > Fixes: 4513025a8790 ("libxc: convert sysctl interfaces over to hypercall
> > buffers")
> > Amends: 73367cf3b4b4 ("libxc: Fix xc_pm API calls to return negative error
> > and stash error in errno")
> > Fixes: 31e264c672bc ("pmstat&xenpm: Re-arrage for cpufreq union")
> > Signed-off-by: Jan Beulich <jbeul...@suse.com>
>
> May I ask for an ack or comments towards what needs changing?
Calling xc_get_cpufreq_para with:
user_para = {
.cpu_num = 0,
.freq_num = 0,
.gov_num = 9,
};
seems broken. It's looks like the `scaling_available_governors` bounce
buffer is going to be used without been allocated properly handled, with
this patch.
Cheers,
--
Anthony Perard | Vates XCP-ng Developer
XCP-ng & Xen Orchestra - Vates solutions
web: https://vates.tech
