On Thu, Jul 27, 2023 at 03:56:23PM +0100, Peter Maydell wrote: > On Wed, 26 Jul 2023 at 16:21, Richard Henderson > <richard.hender...@linaro.org> wrote: > > > > On 7/26/23 02:43, Peter Maydell wrote: > > > (Something went wrong with the quoting in your email. I've > > > fixed it up.) > > > > > > On Wed, 26 Jul 2023 at 05:38, <dingli...@cmss.chinamobile.com> wrote: > > >> Peter Maydell wrote: > > >>> The third part here, is that g_malloc() does not ever > > >>> fail -- it will abort() on out of memory. However > > >>> the code here is still handling g_malloc() returning NULL. > > >>> The equivalent for "we expect this might fail" (which we want > > >>> here, because the guest is passing us the length of memory > > >>> to try to allocate) is g_try_malloc(). > > > > > >> g_malloc() is preferred more than g_try_* functions, which return NULL > > >> on error, > > >> when the size of the requested allocation is small. > > >> This is because allocating few bytes should not be a problem in a > > >> healthy system. > > > > > > This is true. But in this particular case we cannot be sure > > > that the size of the allocation is small, because the size > > > is controlled by the guest. So we want g_try_malloc(). > > > > And why do we want to use g_try_malloc instead of just sticking with malloc? > > The only real reason is just consistency
I think it is slightly stronger than that. By using g_try_malloc we make it explicit that this scenario is expecting the allocation to fail and needs to handle that. If we use plain 'malloc' it isn't clear whether we genuinely expect the allocation to fail, or someone just blindly checked malloc return value out of habit, because they didn't realize QEMU wants abort-on-OOM behaviour most of the time. With regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
