[Qemu-devel] [PATCH v3] block: Switch to host monotonic clock for IO throttling

[Qemu-devel] [PATCH v3] block: Switch to host monotonic clock for IO throttling

[Fam Zheng]

Currently, throttle timers won't make any progress when VCPU is not
running, which would stall the request queue in utils, qtest, vm
suspending, and live migration without special handling.

Block jobs are confusingly inconsistent between with and without
throttling: if user sets a bps limit, starts a block job, then stops vm,
the block job will not make any progress; in contrary, if user unsets
the bps limit, the block job will run normally.

After this patch, with the host clock, even if the VCPUs are stopped,
the throttle queues will be processed.

This patch also enables potential to add throttle to bdrv_drain_all.
Currently all requests are drained immediately. In other words whenever
it is called, IO throttling goes ineffective (examples: system reset,
migration and many block job operations.). This is a loophole that guest
could exploit. If we use the host clock, we can later just trust the
nested poll. This could be done on top.

Note that for qemu-iotests case 093, which uses qtest, we still keep vm
clock so the script can control the clock stepping in order to be
deterministic.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

---
v3: More justification in commit message. [Stefan]
    Add Paolo's and Alberto's rev-bys.
v2: Don't break qemu-iotests 093.
---
 block.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/block.c b/block.c
index 0fe97de..89a1d5b 100644
--- a/block.c
+++ b/block.c
@@ -30,6 +30,7 @@
 #include "qapi/qmp/qjson.h"
 #include "sysemu/block-backend.h"
 #include "sysemu/sysemu.h"
+#include "sysemu/qtest.h"
 #include "qemu/notify.h"
 #include "block/coroutine.h"
 #include "block/qapi.h"
@@ -181,10 +182,16 @@ static void bdrv_throttle_write_timer_cb(void *opaque)
 /* should be called before bdrv_set_io_limits if a limit is set */
 void bdrv_io_limits_enable(BlockDriverState *bs)
 {
+    int clock_type = QEMU_CLOCK_REALTIME;
+
+    if (qtest_enabled()) {
+        /* For testing block IO throttling only */
+        clock_type = QEMU_CLOCK_VIRTUAL;
+    }
     assert(!bs->io_limits_enabled);
     throttle_init(&bs->throttle_state,
                   bdrv_get_aio_context(bs),
-                  QEMU_CLOCK_VIRTUAL,
+                  clock_type,
                   bdrv_throttle_read_timer_cb,
                   bdrv_throttle_write_timer_cb,
                   bs);
-- 
1.9.3

Re: [Qemu-devel] [PATCH v3] block: Switch to host monotonic clock for IO throttling

[Alberto Garcia]

On Wed, Mar 25, 2015 at 10:42:32AM +0800, Fam Zheng wrote:

> Block jobs are confusingly inconsistent between with and without
> throttling: if user sets a bps limit, starts a block job, then stops
> vm, the block job will not make any progress; in contrary, if user
> unsets the bps limit, the block job will run normally.

Wait, that's not what happens:

a) If you start a job then stop the vm then the job will be
   completed immediately.
b) If you stop the vm then start the job then it won't make any
   progress.

This patch fixes b), but a) still happens.

Berto

Re: [Qemu-devel] [PATCH v3] block: Switch to host monotonic clock for IO throttling

[Fam Zheng]

On Wed, 03/25 08:16, Alberto Garcia wrote:
> On Wed, Mar 25, 2015 at 10:42:32AM +0800, Fam Zheng wrote:
> 
> > Block jobs are confusingly inconsistent between with and without
> > throttling: if user sets a bps limit, starts a block job, then stops
> > vm, the block job will not make any progress; in contrary, if user
> > unsets the bps limit, the block job will run normally.
> 
> Wait, that's not what happens:
> 
> a) If you start a job then stop the vm then the job will be
>    completed immediately.
> b) If you stop the vm then start the job then it won't make any
>    progress.
> 
> This patch fixes b), but a) still happens.
> 

Ah you're right, I put the order wrong. I'v just started looking at issue a).

I'll fix the text.

Fam