On Sat, May 28, 2022 at 06:18:27PM +0100, Richard W.M. Jones wrote:
On Fri, May 27, 2022 at 05:25:43PM -0500, Eric Blake wrote:
> Commit c7fbc71d missed an optimization: when we guarantee that a pread
> buffer is pre-zeroed, we don't need to waste time memset()ing it when
> the server sends a hole. But the next commit e0953cb7 lets the user
> skip pre-zeroing, at which point the memset is required to avoid data
> corruption. Thankfully, we are not leaking bogus bytes when a server
> sends a hole, but we CAN speed up the case where we have pre-zeroed
> the buffer to avoid a second memset(). Because the user can change
> the state of pread_initialize on the fly (including while a pread is
> still in flight), we must track the state per-command as it was before
> we send the request to the server.
> ---
Reviewed-by: Richard W.M. Jones <rjones(a)redhat.com>
Thanks; now in as 56d083b885
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization:
qemu.org |
libvirt.org