After commit ea2f0f77538c, a 416-CPU VM running on Hyper-V hangs during
boot because scsi_add_host_with_dma() sets shost->cmd_per_lun to a
negative number:
'max_outstanding_req_per_channel' is 352,
'max_sub_channels' is (416 - 1) / 4 = 103, so in storvsc_probe(),
scsi_driver.can_queue = 352 * (103 + 1) * (100 - 10) / 100 = 32947, which
is bigger than SHRT_MAX (i.e. 32767).
Fix the hang issue by capping scsi_driver.can_queue.
Add the below Fixed tag though ea2f0f77538c itself is good.
Fixes: ea2f0f77538c ("scsi: core: Cap scsi_host cmd_per_lun at can_queue")
Cc: [email protected]
Signed-off-by: Dexuan Cui <[email protected]>
---
drivers/scsi/storvsc_drv.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c
index ebbbc1299c62..ba374908aec2 100644
--- a/drivers/scsi/storvsc_drv.c
+++ b/drivers/scsi/storvsc_drv.c
@@ -1976,6 +1976,16 @@ static int storvsc_probe(struct hv_device *device,
(max_sub_channels + 1) *
(100 - ring_avail_percent_lowater) / 100;
+ /*
+ * v5.14 (see commit ea2f0f77538c) implicitly requires that
+ * scsi_driver.can_queue should not exceed SHRT_MAX, otherwise
+ * scsi_add_host_with_dma() sets shost->cmd_per_lun to a negative
+ * number (note: the type of the "cmd_per_lun" field is "short"), and
+ * the system may hang during early boot.
+ */
+ if (scsi_driver.can_queue > SHRT_MAX)
+ scsi_driver.can_queue = SHRT_MAX;
+
host = scsi_host_alloc(&scsi_driver,
sizeof(struct hv_host_device));
if (!host)
--
2.17.1
On Wed, Oct 06, 2021 at 12:03:45AM -0700, Dexuan Cui wrote:
> After commit ea2f0f77538c, a 416-CPU VM running on Hyper-V hangs during
> boot because scsi_add_host_with_dma() sets shost->cmd_per_lun to a
> negative number:
> 'max_outstanding_req_per_channel' is 352,
> 'max_sub_channels' is (416 - 1) / 4 = 103, so in storvsc_probe(),
> scsi_driver.can_queue = 352 * (103 + 1) * (100 - 10) / 100 = 32947, which
> is bigger than SHRT_MAX (i.e. 32767).
>
> Fix the hang issue by capping scsi_driver.can_queue.
>
> Add the below Fixed tag though ea2f0f77538c itself is good.
>
> Fixes: ea2f0f77538c ("scsi: core: Cap scsi_host cmd_per_lun at can_queue")
> Cc: [email protected]
> Signed-off-by: Dexuan Cui <[email protected]>
> ---
> drivers/scsi/storvsc_drv.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c
> index ebbbc1299c62..ba374908aec2 100644
> --- a/drivers/scsi/storvsc_drv.c
> +++ b/drivers/scsi/storvsc_drv.c
> @@ -1976,6 +1976,16 @@ static int storvsc_probe(struct hv_device *device,
> (max_sub_channels + 1) *
> (100 - ring_avail_percent_lowater) / 100;
>
> + /*
> + * v5.14 (see commit ea2f0f77538c) implicitly requires that
No need to put a version number in a comment, they do not track well
when people backport changes all over the place in other kernel trees.
If you want to refer to a commit, please do so in the documented way.
For this case, that would be:
ea2f0f77538c ("scsi: core: Cap scsi_host cmd_per_lun at can_queue")
thanks,
greg k-h
On 06/10/2021 08:03, Dexuan Cui wrote:
> After commit ea2f0f77538c, a 416-CPU VM running on Hyper-V hangs during
> boot because scsi_add_host_with_dma() sets shost->cmd_per_lun to a
> negative number:
> 'max_outstanding_req_per_channel' is 352,
> 'max_sub_channels' is (416 - 1) / 4 = 103, so in storvsc_probe(),
> scsi_driver.can_queue = 352 * (103 + 1) * (100 - 10) / 100 = 32947, which
> is bigger than SHRT_MAX (i.e. 32767).
Out of curiosity, are these values realistic? You're capping can_queue
just because of a data size issue, so, if these values are realistic,
seems a weak reason.
>
> Fix the hang issue by capping scsi_driver.can_queue.
>
> Add the below Fixed tag though ea2f0f77538c itself is good.
>
> Fixes: ea2f0f77538c ("scsi: core: Cap scsi_host cmd_per_lun at can_queue")
> Cc: [email protected]
> Signed-off-by: Dexuan Cui <[email protected]>
> ---
> drivers/scsi/storvsc_drv.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c
> index ebbbc1299c62..ba374908aec2 100644
> --- a/drivers/scsi/storvsc_drv.c
> +++ b/drivers/scsi/storvsc_drv.c
> @@ -1976,6 +1976,16 @@ static int storvsc_probe(struct hv_device *device,
> (max_sub_channels + 1) *
> (100 - ring_avail_percent_lowater) / 100;
>
> + /*
> + * v5.14 (see commit ea2f0f77538c) implicitly requires that
> + * scsi_driver.can_queue should not exceed SHRT_MAX, otherwise
> + * scsi_add_host_with_dma() sets shost->cmd_per_lun to a negative
> + * number (note: the type of the "cmd_per_lun" field is "short"), and
> + * the system may hang during early boot.
> + */
The different data sizes for cmd_per_lun and can_queue are problematic here.
I'd be more inclined to set cmd_per_lun to the same data size as
can_queue. We did discuss this when ea2f0f77538c was upstreamed
(actually it was the other way around - setting can_queue to 16b).
Thanks,
John
> + if (scsi_driver.can_queue > SHRT_MAX)
> + scsi_driver.can_queue = SHRT_MAX;
> +
> host = scsi_host_alloc(&scsi_driver,
> sizeof(struct hv_host_device));
> if (!host)
>
From: John Garry <[email protected]> Sent: Wednesday, October 6, 2021 1:17 AM
>
> On 06/10/2021 08:03, Dexuan Cui wrote:
> > After commit ea2f0f77538c, a 416-CPU VM running on Hyper-V hangs during
> > boot because scsi_add_host_with_dma() sets shost->cmd_per_lun to a
> > negative number:
> > 'max_outstanding_req_per_channel' is 352,
> > 'max_sub_channels' is (416 - 1) / 4 = 103, so in storvsc_probe(),
> > scsi_driver.can_queue = 352 * (103 + 1) * (100 - 10) / 100 = 32947, which
> > is bigger than SHRT_MAX (i.e. 32767).
>
> Out of curiosity, are these values realistic? You're capping can_queue
> just because of a data size issue, so, if these values are realistic,
> seems a weak reason.
>
The calculated value of can_queue is not realistic. The blk-mq layer
caps the number of tags at 10240, so the excessively large value
calculated here didn't definitively break anything, though it can be
poor from a performance tuning standpoint. The algorithm used here
is fairly broken, particularly in VMs with large CPU counts. I have an
effort underway to fix it, but its part of a bigger set of changes to also
do a better job on the perf tuning aspects.
> >
> > Fix the hang issue by capping scsi_driver.can_queue.
> >
> > Add the below Fixed tag though ea2f0f77538c itself is good.
> >
> > Fixes: ea2f0f77538c ("scsi: core: Cap scsi_host cmd_per_lun at can_queue")
> > Cc: [email protected]
> > Signed-off-by: Dexuan Cui <[email protected]>
> > ---
> > drivers/scsi/storvsc_drv.c | 10 ++++++++++
> > 1 file changed, 10 insertions(+)
> >
> > diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c
> > index ebbbc1299c62..ba374908aec2 100644
> > --- a/drivers/scsi/storvsc_drv.c
> > +++ b/drivers/scsi/storvsc_drv.c
> > @@ -1976,6 +1976,16 @@ static int storvsc_probe(struct hv_device *device,
> > (max_sub_channels + 1) *
> > (100 - ring_avail_percent_lowater) / 100;
> >
> > + /*
> > + * v5.14 (see commit ea2f0f77538c) implicitly requires that
> > + * scsi_driver.can_queue should not exceed SHRT_MAX, otherwise
> > + * scsi_add_host_with_dma() sets shost->cmd_per_lun to a negative
> > + * number (note: the type of the "cmd_per_lun" field is "short"), and
> > + * the system may hang during early boot.
> > + */
>
> The different data sizes for cmd_per_lun and can_queue are problematic here.
>
> I'd be more inclined to set cmd_per_lun to the same data size as
> can_queue. We did discuss this when ea2f0f77538c was upstreamed
> (actually it was the other way around - setting can_queue to 16b).
I can see that making can_queue be 16 bits would make sense.
And it also seems that both cmd_per_lun and can_queue should be
unsigned, though I don't the implications of making such a change.
But in today's world where cmd_per_lun is "short" and can_queue
is "int", ea2f0f77538c seems incorrect to me. The comparison should
be done as "int", not "short", in order to prevent the truncation
problem with can_queue that Dexuan's patch is trying to address.
The result will always fit in back into the "short" cmd_per_lun since
it is calculating a "min" function.
>
> Thanks,
> John
>
>
> > + if (scsi_driver.can_queue > SHRT_MAX)
> > + scsi_driver.can_queue = SHRT_MAX;
> > +
This fix works, but is a more of a temporary hack until I can finish
a larger overhaul of the algorithm. But for now, I think the better
fix is for ea2f0f77538c to do the comparison as "int" instead of "short".
Michael
> > host = scsi_host_alloc(&scsi_driver,
> > sizeof(struct hv_host_device));
> > if (!host)
> >
On 06/10/2021 16:01, Michael Kelley wrote:
> From: John Garry <[email protected]> Sent: Wednesday, October 6, 2021 1:17 AM
>>
>> On 06/10/2021 08:03, Dexuan Cui wrote:
>>> After commit ea2f0f77538c, a 416-CPU VM running on Hyper-V hangs during
>>> boot because scsi_add_host_with_dma() sets shost->cmd_per_lun to a
>>> negative number:
>>> 'max_outstanding_req_per_channel' is 352,
>>> 'max_sub_channels' is (416 - 1) / 4 = 103, so in storvsc_probe(),
>>> scsi_driver.can_queue = 352 * (103 + 1) * (100 - 10) / 100 = 32947, which
>>> is bigger than SHRT_MAX (i.e. 32767).
>>
>> Out of curiosity, are these values realistic? You're capping can_queue
>> just because of a data size issue, so, if these values are realistic,
>> seems a weak reason.
>>
>
> The calculated value of can_queue is not realistic. The blk-mq layer
> caps the number of tags at 10240,
nit: 1024, I think
> so the excessively large value
> calculated here didn't definitively break anything, though it can be
> poor from a performance tuning standpoint. The algorithm used here
> is fairly broken, particularly in VMs with large CPU counts. I have an
> effort underway to fix it, but its part of a bigger set of changes to also
> do a better job on the perf tuning aspects.
>
>>>
>>> Fix the hang issue by capping scsi_driver.can_queue.
>>>
>>> Add the below Fixed tag though ea2f0f77538c itself is good.
>>>
>>> Fixes: ea2f0f77538c ("scsi: core: Cap scsi_host cmd_per_lun at can_queue")
>>> Cc: [email protected]
>>> Signed-off-by: Dexuan Cui <[email protected]>
>>> ---
>>> drivers/scsi/storvsc_drv.c | 10 ++++++++++
>>> 1 file changed, 10 insertions(+)
>>>
>>> diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c
>>> index ebbbc1299c62..ba374908aec2 100644
>>> --- a/drivers/scsi/storvsc_drv.c
>>> +++ b/drivers/scsi/storvsc_drv.c
>>> @@ -1976,6 +1976,16 @@ static int storvsc_probe(struct hv_device *device,
>>> (max_sub_channels + 1) *
>>> (100 - ring_avail_percent_lowater) / 100;
>>>
>>> + /*
>>> + * v5.14 (see commit ea2f0f77538c) implicitly requires that
>>> + * scsi_driver.can_queue should not exceed SHRT_MAX, otherwise
>>> + * scsi_add_host_with_dma() sets shost->cmd_per_lun to a negative
>>> + * number (note: the type of the "cmd_per_lun" field is "short"), and
>>> + * the system may hang during early boot.
>>> + */
>>
>> The different data sizes for cmd_per_lun and can_queue are problematic here.
>>
>> I'd be more inclined to set cmd_per_lun to the same data size as
>> can_queue. We did discuss this when ea2f0f77538c was upstreamed
>> (actually it was the other way around - setting can_queue to 16b).
>
> I can see that making can_queue be 16 bits would make sense.
> And it also seems that both cmd_per_lun and can_queue should be
> unsigned, though I don't the implications of making such a change.
>
> But in today's world where cmd_per_lun is "short" and can_queue
> is "int", ea2f0f77538c seems incorrect to me. The comparison should
> be done as "int", not "short", in order to prevent the truncation
> problem with can_queue that Dexuan's patch is trying to address.
Yeah, right. I think can_queue values > short_max was considered outside
the realms of what is realistic then, hence my sloppy programming.
> The result will always fit in back into the "short" cmd_per_lun since
> it is calculating a "min" function.
>
>>
>> Thanks,
>> John
>>
>>
>>> + if (scsi_driver.can_queue > SHRT_MAX)
>>> + scsi_driver.can_queue = SHRT_MAX;
>>> +
>
> This fix works, but is a more of a temporary hack until I can finish
> a larger overhaul of the algorithm.
> But for now, I think the better
> fix is for ea2f0f77538c to do the comparison as "int" instead of "short".
>
That seems better to me. But Let's wait for other possible opinion.
Thanks,
John
From: John Garry <[email protected]> Sent: Wednesday, October 6, 2021 9:03 AM
>
> On 06/10/2021 16:01, Michael Kelley wrote:
> > From: John Garry <[email protected]> Sent: Wednesday, October 6, 2021 1:17 AM
> >>
> >> On 06/10/2021 08:03, Dexuan Cui wrote:
> >>> After commit ea2f0f77538c, a 416-CPU VM running on Hyper-V hangs during
> >>> boot because scsi_add_host_with_dma() sets shost->cmd_per_lun to a
> >>> negative number:
> >>> 'max_outstanding_req_per_channel' is 352,
> >>> 'max_sub_channels' is (416 - 1) / 4 = 103, so in storvsc_probe(),
> >>> scsi_driver.can_queue = 352 * (103 + 1) * (100 - 10) / 100 = 32947, which
> >>> is bigger than SHRT_MAX (i.e. 32767).
> >>
> >> Out of curiosity, are these values realistic? You're capping can_queue
> >> just because of a data size issue, so, if these values are realistic,
> >> seems a weak reason.
> >>
> >
> > The calculated value of can_queue is not realistic. The blk-mq layer
> > caps the number of tags at 10240,
>
> nit: 1024, I think
I was thinking about BLK_MQ_MAX_DEPTH (#define'd as 10240), which
is used to limit the tag set size in blk_mq_alloc_tag_set(). When running
on large VMs on Hyper-V, we can see the "blk-mq: reduced tag depth
to 10240" message. :-(
Michael
On 06/10/2021 17:08, Michael Kelley wrote:
>>> The calculated value of can_queue is not realistic. The blk-mq layer
>>> caps the number of tags at 10240,
>> nit: 1024, I think
> I was thinking about BLK_MQ_MAX_DEPTH (#define'd as 10240), which
> is used to limit the tag set size in blk_mq_alloc_tag_set(). When running
> on large VMs on Hyper-V, we can see the "blk-mq: reduced tag depth
> to 10240" message.:-(
Ah, right. The other related capping is the sdev queue depth, which is
now capped at max(1024, can_queue), see scsi_device_max_queue_depth().
Thanks,
John
> From: John Garry <[email protected]>
> Sent: Wednesday, October 6, 2021 9:03 AM
> >> ...
> >>> + if (scsi_driver.can_queue > SHRT_MAX)
> >>> + scsi_driver.can_queue = SHRT_MAX;
> >>> +
> >
> > This fix works, but is a more of a temporary hack until I can finish
> > a larger overhaul of the algorithm.
>
> > But for now, I think the better
> > fix is for ea2f0f77538c to do the comparison as "int" instead of "short".
> >
>
> That seems better to me. But Let's wait for other possible opinion.
>
> Thanks,
> John
It looks like shost->cmd_per_lun has been "short" since day 1.
I don't know whether it should be changed to unsigned int.
Thanks for the thoughts! I'll post a v2 like the below in 24 hours.
diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c
index 3f6f14f0cafb..24b72ee4246f 100644
--- a/drivers/scsi/hosts.c
+++ b/drivers/scsi/hosts.c
@@ -220,7 +220,8 @@ int scsi_add_host_with_dma(struct Scsi_Host *shost, struct device *dev,
goto fail;
}
- shost->cmd_per_lun = min_t(short, shost->cmd_per_lun,
+ /* Use min_t(int, ...) in case shost->can_queue exceeds SHRT_MAX */
+ shost->cmd_per_lun = min_t(int, shost->cmd_per_lun,
shost->can_queue);
error = scsi_init_sense_cache(shost)