Re: [PULL v2 39/58] Transmit vhost-user memory regions individually

2020-06-22 Thread Raphael Norwitz
On Fri, Jun 19, 2020 at 6:03 AM Peter Maydell wrote: > > > As noted in my other email, I think the best fix for this is to > have vhost_user_fill_msg_region() take an extra mmap_offset > argument to fill in the mmap_offset itself. In this callsite in > send_add_regions() we would pass in 'offset'

Re: [PULL v2 39/58] Transmit vhost-user memory regions individually

2020-06-19 Thread Peter Maydell
On Fri, 12 Jun 2020 at 15:52, Michael S. Tsirkin wrote: > > From: Raphael Norwitz > > With this change, when the VHOST_USER_PROTOCOL_F_CONFIGURE_MEM_SLOTS > protocol feature has been negotiated, Qemu no longer sends the backend > all the memory regions in a single message. Rather, when the

[PULL v2 39/58] Transmit vhost-user memory regions individually

2020-06-12 Thread Michael S. Tsirkin
From: Raphael Norwitz With this change, when the VHOST_USER_PROTOCOL_F_CONFIGURE_MEM_SLOTS protocol feature has been negotiated, Qemu no longer sends the backend all the memory regions in a single message. Rather, when the memory tables are set or updated, a series of VHOST_USER_ADD_MEM_REG and