Il 19/03/2013 14:12, Michael R. Hines ha scritto: > On 03/19/2013 05:18 AM, Paolo Bonzini wrote: >> Il 18/03/2013 21:33, Michael R. Hines ha scritto: >>>> +int qemu_drain(QEMUFile *f) >>>> +{ >>>> + return f->ops->drain ? f->ops->drain(f->opaque) : 0; >>>> +} >>>> Hmm, this is very similar to qemu_fflush, but not quite. :/ >>>> >>>> Why exactly is this needed? >>> Good idea - I'll replace drain with flush once I added >>> the "qemu_file_ops_are(const QEMUFile *, const QEMUFileOps *) " >>> that you recommended...... >> If I understand correctly, the problem is that save_rdma_page is >> asynchronous and you have to wait for pending operations to do the >> put_buffer protocol correctly. >> >> Would it work to just do the "drain" in the put_buffer operation, if and >> only if it was preceded by a save_rdma_page operation? > > Yes, the drain needs to happen in a few places already: > > 1. During save_rdma_page (if the current "chunk" is full of pages)
Ok, this is internal to RDMA so no problem. > 2. During the end of each iteration (now using qemu_fflush in my current > patch) Why? > 3. And also during qemu_savem_state_complete(), also using qemu_fflush. This would be caught by put_buffer, but (2) would not. >>>>> /** Flushes QEMUFile buffer >>>>> * >>>>> */ >>>>> @@ -723,6 +867,8 @@ int qemu_get_byte(QEMUFile *f) >>>>> int64_t qemu_ftell(QEMUFile *f) >>>>> { >>>>> qemu_fflush(f); >>>>> + if(migrate_use_rdma(f)) >>>>> + return delta_norm_mig_bytes_transferred(); >>>> Not needed, and another undesirable dependency (savevm.c -> >>>> arch_init.c). Just update f->pos in save_rdma_page. >>> f->pos isn't good enough because save_rdma_page does not >>> go through QEMUFile directly - only non-live state goes >>> through QEMUFile ....... pc.ram uses direct RDMA writes. >>> >>> As a result, the position pointer does not get updated >>> and the accounting is missed........ >> Yes, I am suggesting to modify f->pos in save_rdma_page instead. >> >> Paolo >> > > Would that not confuse the other QEMUFile users? > If I change that pointer (without actually putting bytes > in into QEMUFile), won't the f->pos pointer be > incorrectly updated? f->pos is never used directly by QEMUFile, it is almost an opaque value. It is accumulated on every qemu_fflush (so that it can be passed to the ->put_buffer function), and returned by qemu_ftell; nothing else. If you make somehow save_rdma_page a new op, returning a value from that op and adding it to f->pos would be a good way to achieve this. Paolo