Re: [PATCH 2/3] virtiofs: split requests that exceed virtqueue size

Re: [PATCH 2/3] virtiofs: split requests that exceed virtqueue size

[Vivek Goyal]

On Thu, Mar 18, 2021 at 08:52:22AM -0500, Connor Kuehl wrote:
> If an incoming FUSE request can't fit on the virtqueue, the request is
> placed onto a workqueue so a worker can try to resubmit it later where
> there will (hopefully) be space for it next time.
> 
> This is fine for requests that aren't larger than a virtqueue's maximum
> capacity. However, if a request's size exceeds the maximum capacity of
> the virtqueue (even if the virtqueue is empty), it will be doomed to a
> life of being placed on the workqueue, removed, discovered it won't fit,
> and placed on the workqueue yet again.
> 
> Furthermore, from section 2.6.5.3.1 (Driver Requirements: Indirect
> Descriptors) of the virtio spec:
> 
>   "A driver MUST NOT create a descriptor chain longer than the Queue
>   Size of the device."
> 
> To fix this, limit the number of pages FUSE will use for an overall
> request. This way, each request can realistically fit on the virtqueue
> when it is decomposed into a scattergather list and avoid violating
> section 2.6.5.3.1 of the virtio spec.

Hi Connor,

So as of now if a request is bigger than what virtqueue can support,
it never gets dispatched and caller waits infinitely? So this patch
will fix it by forcing fuse to split the request. That sounds good.


[..]
> diff --git a/fs/fuse/virtio_fs.c b/fs/fuse/virtio_fs.c
> index 8868ac31a3c0..a6ffba85d59a 100644
> --- a/fs/fuse/virtio_fs.c
> +++ b/fs/fuse/virtio_fs.c
> @@ -18,6 +18,12 @@
>  #include <linux/uio.h>
>  #include "fuse_i.h"
>  
> +/* Used to help calculate the FUSE connection's max_pages limit for a request's
> + * size. Parts of the struct fuse_req are sliced into scattergather lists in
> + * addition to the pages used, so this can help account for that overhead.
> + */
> +#define FUSE_HEADER_OVERHEAD    4

How did yo arrive at this overhead. Is it following.

- One sg element for fuse_in_header.
- One sg element for input arguments.
- One sg element for fuse_out_header.
- One sg element for output args.

Thanks
Vivek

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 2/3] virtiofs: split requests that exceed virtqueue size

[Connor Kuehl]

On 3/19/21 8:49 AM, Vivek Goyal wrote:
> On Thu, Mar 18, 2021 at 08:52:22AM -0500, Connor Kuehl wrote:
>> If an incoming FUSE request can't fit on the virtqueue, the request is
>> placed onto a workqueue so a worker can try to resubmit it later where
>> there will (hopefully) be space for it next time.
>>
>> This is fine for requests that aren't larger than a virtqueue's maximum
>> capacity. However, if a request's size exceeds the maximum capacity of
>> the virtqueue (even if the virtqueue is empty), it will be doomed to a
>> life of being placed on the workqueue, removed, discovered it won't fit,
>> and placed on the workqueue yet again.
>>
>> Furthermore, from section 2.6.5.3.1 (Driver Requirements: Indirect
>> Descriptors) of the virtio spec:
>>
>>    "A driver MUST NOT create a descriptor chain longer than the Queue
>>    Size of the device."
>>
>> To fix this, limit the number of pages FUSE will use for an overall
>> request. This way, each request can realistically fit on the virtqueue
>> when it is decomposed into a scattergather list and avoid violating
>> section 2.6.5.3.1 of the virtio spec.
> 
> Hi Connor,
> 
> So as of now if a request is bigger than what virtqueue can support,
> it never gets dispatched and caller waits infinitely? So this patch
> will fix it by forcing fuse to split the request. That sounds good.

Right, in theory. Certain configurations make it easier to avoid this 
from happening, such as using indirect descriptors; however, in that 
case, the virtio spec says even if indirect descriptors are used, the 
descriptor chain length shouldn't exceed the length of the queue's size 
anyways. So having FUSE split the request also helps to uphold that 
property.

This is my reading of the potential looping problem:

virtio_fs_wake_pending_and_unlock
   calls
     virtio_fs_enqueue_req
       calls
         virtqueue_add_sgs

virtqueue_add_sgs can return -ENOSPC if there aren't enough descriptors 
available.

This error gets propagated back down to 
virtio_fs_wake_pending_and_unlock which checks for this exact issue and 
places the request on a workqueue to retry submission later.

Resubmission occurs in virtio_fs_request_dispatch_work, which does a 
similar dance, where if the request fails with -ENOSPC it just puts it 
back in the queue. However, for a sufficiently large request that would 
exceed the capacity of the virtqueue (even when empty), no amount of 
retrying will ever make it fit.

> 
> 
> [..]
>> diff --git a/fs/fuse/virtio_fs.c b/fs/fuse/virtio_fs.c
>> index 8868ac31a3c0..a6ffba85d59a 100644
>> --- a/fs/fuse/virtio_fs.c
>> +++ b/fs/fuse/virtio_fs.c
>> @@ -18,6 +18,12 @@
>>   #include <linux/uio.h>
>>   #include "fuse_i.h"
>>   
>> +/* Used to help calculate the FUSE connection's max_pages limit for a request's
>> + * size. Parts of the struct fuse_req are sliced into scattergather lists in
>> + * addition to the pages used, so this can help account for that overhead.
>> + */
>> +#define FUSE_HEADER_OVERHEAD    4
> 
> How did yo arrive at this overhead. Is it following.
> 
> - One sg element for fuse_in_header.
> - One sg element for input arguments.
> - One sg element for fuse_out_header.
> - One sg element for output args.

Yes, that's exactly how I got to that number.

Connor


_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 2/3] virtiofs: split requests that exceed virtqueue size

[Michael S. Tsirkin]

On Thu, Mar 18, 2021 at 10:52:14AM -0500, Connor Kuehl wrote:
> On 3/18/21 10:17 AM, Miklos Szeredi wrote:
> > I removed the conditional compilation and renamed the limit.  Also made
> > virtio_fs_get_tree() bail out if it hit the WARN_ON().  Updated patch below.
> 
> Thanks, Miklos. I think it looks better with those changes.
> 
> > The virtio_ring patch in this series should probably go through the respective
> > subsystem tree.
> 
> Makes sense. I've CC'd everyone that ./scripts/get_maintainers.pl suggested
> for that patch on this entire series as well. Should I resend patch #1
> through just that subsystem to avoid confusion or wait to see if it gets
> picked out of this series?

Yes pls post separately. Thanks!

> Thanks again,
> 
> Connor

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 1/3] virtio_ring: always warn when descriptor chain exceeds queue size

[Jason Wang]

在 2021/3/18 下午9:52, Connor Kuehl 写道:
>  From section 2.6.5.3.1 (Driver Requirements: Indirect Descriptors)
> of the virtio spec:
>
>    "A driver MUST NOT create a descriptor chain longer than the Queue
>    Size of the device."
>
> This text suggests that the warning should trigger even if
> indirect descriptors are in use.


So I think at least the commit log needs some tweak.

For split virtqueue. We had:

2.6.5.2 Driver Requirements: The Virtqueue Descriptor Table

Drivers MUST NOT add a descriptor chain longer than 2^32 bytes in total; 
this implies that loops in the descriptor chain are forbidden!

2.6.5.3.1 Driver Requirements: Indirect Descriptors

A driver MUST NOT create a descriptor chain longer than the Queue Size 
of the device.

If I understand the spec correctly, the check is only needed for a 
single indirect descriptor table?

For packed virtqueue. We had:

2.7.17 Driver Requirements: Scatter-Gather Support

A driver MUST NOT create a descriptor list longer than allowed by the 
device.

A driver MUST NOT create a descriptor list longer than the Queue Size.

2.7.19 Driver Requirements: Indirect Descriptors

A driver MUST NOT create a descriptor chain longer than allowed by the 
device.

So it looks to me the packed part is fine.

Note that if I understand the spec correctly 2.7.17 implies 2.7.19.

Thanks


>
> Reported-by: Stefan Hajnoczi <stefanha@redhat.com>
> Signed-off-by: Connor Kuehl <ckuehl@redhat.com>
> ---
>   drivers/virtio/virtio_ring.c | 7 ++++---
>   1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 71e16b53e9c1..1bc290f9ba13 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -444,11 +444,12 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
>   
>   	head = vq->free_head;
>   
> +	WARN_ON_ONCE(total_sg > vq->split.vring.num);
> +
>   	if (virtqueue_use_indirect(_vq, total_sg))
>   		desc = alloc_indirect_split(_vq, total_sg, gfp);
>   	else {
>   		desc = NULL;
> -		WARN_ON_ONCE(total_sg > vq->split.vring.num && !vq->indirect);
>   	}
>   
>   	if (desc) {
> @@ -1118,6 +1119,8 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>   
>   	BUG_ON(total_sg == 0);
>   
> +	WARN_ON_ONCE(total_sg > vq->packed.vring.num);
> +
>   	if (virtqueue_use_indirect(_vq, total_sg))
>   		return virtqueue_add_indirect_packed(vq, sgs, total_sg,
>   				out_sgs, in_sgs, data, gfp);
> @@ -1125,8 +1128,6 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>   	head = vq->packed.next_avail_idx;
>   	avail_used_flags = vq->packed.avail_used_flags;
>   
> -	WARN_ON_ONCE(total_sg > vq->packed.vring.num && !vq->indirect);
> -
>   	desc = vq->packed.vring.desc;
>   	i = head;
>   	descs_used = total_sg;

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 1/3] virtio_ring: always warn when descriptor chain exceeds queue size

[Jason Wang]

在 2021/3/22 上午11:22, Jason Wang 写道:
>
> 在 2021/3/18 下午9:52, Connor Kuehl 写道:
>>  From section 2.6.5.3.1 (Driver Requirements: Indirect Descriptors)
>> of the virtio spec:
>>
>>    "A driver MUST NOT create a descriptor chain longer than the Queue
>>    Size of the device."
>>
>> This text suggests that the warning should trigger even if
>> indirect descriptors are in use.
>
>
> So I think at least the commit log needs some tweak.
>
> For split virtqueue. We had:
>
> 2.6.5.2 Driver Requirements: The Virtqueue Descriptor Table
>
> Drivers MUST NOT add a descriptor chain longer than 2^32 bytes in 
> total; this implies that loops in the descriptor chain are forbidden!
>
> 2.6.5.3.1 Driver Requirements: Indirect Descriptors
>
> A driver MUST NOT create a descriptor chain longer than the Queue Size 
> of the device.
>
> If I understand the spec correctly, the check is only needed for a 
> single indirect descriptor table?
>
> For packed virtqueue. We had:
>
> 2.7.17 Driver Requirements: Scatter-Gather Support
>
> A driver MUST NOT create a descriptor list longer than allowed by the 
> device.
>
> A driver MUST NOT create a descriptor list longer than the Queue Size.
>
> 2.7.19 Driver Requirements: Indirect Descriptors
>
> A driver MUST NOT create a descriptor chain longer than allowed by the 
> device.
>
> So it looks to me the packed part is fine.
>
> Note that if I understand the spec correctly 2.7.17 implies 2.7.19.


Actually not. So in 2.7.7, spec said:

"

Some devices benefit by concurrently dispatching a large number of large 
requests. The VIRTIO_F_INDIRECT_DESC feature allows this. To increase 
ring capacity the driver can store a (read-only by the device) table of 
indirect descriptors anywhere in memory, and insert a descriptor in the 
main virtqueue (with Flags bit VIRTQ_DESC_F_INDIRECT on) that refers to 
a buffer element containing this indirect descriptor table; addr and len 
refer to the indirect table address and length in bytes, respectively.

"

And in 2.7.5, spec said

"

While unusual (most implementations either create all lists solely using 
non-indirect descriptors, or always use a single indirect element), if 
both features have been negotiated, mixing indirect and non-indirect 
descriptors in a ring is valid, as long as each list only contains 
descriptors of a given type.
"

So my understanding is that the indirect descriptor is used to sumbit 
the request whose #buffers is greater than the virtqueue size. And the 
spec allows the driver to create a list of indirect descriptors just 
need to make sure the number of indirect descriptors in this list must 
not exceed the size of the virtqueue (2.7.17). And for each indirector 
descriptor, the number of chained descriptor must not exceed the 
virtqueue size. So actually this aligns with split virtqueue.

So if I understand the spec correctly, what we need to do is to make 
sure the descriptor chained in the indirect descriptor table does not 
exceed the virtqueue size. That means we probably need to chain indirect 
descriptors instead of a warn here.

Thanks


>
> Thanks
>
>
>>
>> Reported-by: Stefan Hajnoczi <stefanha@redhat.com>
>> Signed-off-by: Connor Kuehl <ckuehl@redhat.com>
>> ---
>>   drivers/virtio/virtio_ring.c | 7 ++++---
>>   1 file changed, 4 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
>> index 71e16b53e9c1..1bc290f9ba13 100644
>> --- a/drivers/virtio/virtio_ring.c
>> +++ b/drivers/virtio/virtio_ring.c
>> @@ -444,11 +444,12 @@ static inline int virtqueue_add_split(struct 
>> virtqueue *_vq,
>>         head = vq->free_head;
>>   +    WARN_ON_ONCE(total_sg > vq->split.vring.num);
>> +
>>       if (virtqueue_use_indirect(_vq, total_sg))
>>           desc = alloc_indirect_split(_vq, total_sg, gfp);
>>       else {
>>           desc = NULL;
>> -        WARN_ON_ONCE(total_sg > vq->split.vring.num && !vq->indirect);
>>       }
>>         if (desc) {
>> @@ -1118,6 +1119,8 @@ static inline int virtqueue_add_packed(struct 
>> virtqueue *_vq,
>>         BUG_ON(total_sg == 0);
>>   +    WARN_ON_ONCE(total_sg > vq->packed.vring.num);
>> +
>>       if (virtqueue_use_indirect(_vq, total_sg))
>>           return virtqueue_add_indirect_packed(vq, sgs, total_sg,
>>                   out_sgs, in_sgs, data, gfp);
>> @@ -1125,8 +1128,6 @@ static inline int virtqueue_add_packed(struct 
>> virtqueue *_vq,
>>       head = vq->packed.next_avail_idx;
>>       avail_used_flags = vq->packed.avail_used_flags;
>>   -    WARN_ON_ONCE(total_sg > vq->packed.vring.num && !vq->indirect);
>> -
>>       desc = vq->packed.vring.desc;
>>       i = head;
>>       descs_used = total_sg;
>
> _______________________________________________
> Virtualization mailing list
> Virtualization@lists.linux-foundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/virtualization

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 3/3] fuse: fix typo for fuse_conn.max_pages comment

[Jason Wang]

在 2021/3/18 下午9:52, Connor Kuehl 写道:
> 'Maxmum' -> 'Maximum'


Need a better log here.

With the commit log fixed.

Acked-by: Jason Wang <jasowang@redhat.com>


>
> Signed-off-by: Connor Kuehl <ckuehl@redhat.com>
> ---
>   fs/fuse/fuse_i.h | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/fuse/fuse_i.h b/fs/fuse/fuse_i.h
> index f0e4ee906464..8bdee79ba593 100644
> --- a/fs/fuse/fuse_i.h
> +++ b/fs/fuse/fuse_i.h
> @@ -552,7 +552,7 @@ struct fuse_conn {
>   	/** Maximum write size */
>   	unsigned max_write;
>   
> -	/** Maxmum number of pages that can be used in a single request */
> +	/** Maximum number of pages that can be used in a single request */
>   	unsigned int max_pages;
>   
>   #if IS_ENABLED(CONFIG_VIRTIO_FS)

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 1/3] virtio_ring: always warn when descriptor chain exceeds queue size

[Michael S. Tsirkin]

On Mon, Mar 22, 2021 at 11:22:15AM +0800, Jason Wang wrote:
> 
> 在 2021/3/18 下午9:52, Connor Kuehl 写道:
> >  From section 2.6.5.3.1 (Driver Requirements: Indirect Descriptors)
> > of the virtio spec:
> > 
> >    "A driver MUST NOT create a descriptor chain longer than the Queue
> >    Size of the device."
> > 
> > This text suggests that the warning should trigger even if
> > indirect descriptors are in use.
> 
> 
> So I think at least the commit log needs some tweak.
> 
> For split virtqueue. We had:
> 
> 2.6.5.2 Driver Requirements: The Virtqueue Descriptor Table
> 
> Drivers MUST NOT add a descriptor chain longer than 2^32 bytes in total;
> this implies that loops in the descriptor chain are forbidden!
> 
> 2.6.5.3.1 Driver Requirements: Indirect Descriptors
> 
> A driver MUST NOT create a descriptor chain longer than the Queue Size of
> the device.
> 
> If I understand the spec correctly, the check is only needed for a single
> indirect descriptor table?
> 
> For packed virtqueue. We had:
> 
> 2.7.17 Driver Requirements: Scatter-Gather Support
> 
> A driver MUST NOT create a descriptor list longer than allowed by the
> device.
> 
> A driver MUST NOT create a descriptor list longer than the Queue Size.
> 
> 2.7.19 Driver Requirements: Indirect Descriptors
> 
> A driver MUST NOT create a descriptor chain longer than allowed by the
> device.
> 
> So it looks to me the packed part is fine.
> 
> Note that if I understand the spec correctly 2.7.17 implies 2.7.19.
> 
> Thanks

It would be quite strange for packed and split to differ here:
so for packed would you say there's no limit on # of descriptors at all?

I am guessing I just forgot to move this part from
the format specific to the common part of the spec.

This needs discussion in the TC mailing list - want to start a thread
there?



> 
> > 
> > Reported-by: Stefan Hajnoczi <stefanha@redhat.com>
> > Signed-off-by: Connor Kuehl <ckuehl@redhat.com>
> > ---
> >   drivers/virtio/virtio_ring.c | 7 ++++---
> >   1 file changed, 4 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > index 71e16b53e9c1..1bc290f9ba13 100644
> > --- a/drivers/virtio/virtio_ring.c
> > +++ b/drivers/virtio/virtio_ring.c
> > @@ -444,11 +444,12 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
> >   	head = vq->free_head;
> > +	WARN_ON_ONCE(total_sg > vq->split.vring.num);
> > +
> >   	if (virtqueue_use_indirect(_vq, total_sg))
> >   		desc = alloc_indirect_split(_vq, total_sg, gfp);
> >   	else {
> >   		desc = NULL;
> > -		WARN_ON_ONCE(total_sg > vq->split.vring.num && !vq->indirect);
> >   	}
> >   	if (desc) {
> > @@ -1118,6 +1119,8 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
> >   	BUG_ON(total_sg == 0);
> > +	WARN_ON_ONCE(total_sg > vq->packed.vring.num);
> > +
> >   	if (virtqueue_use_indirect(_vq, total_sg))
> >   		return virtqueue_add_indirect_packed(vq, sgs, total_sg,
> >   				out_sgs, in_sgs, data, gfp);
> > @@ -1125,8 +1128,6 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
> >   	head = vq->packed.next_avail_idx;
> >   	avail_used_flags = vq->packed.avail_used_flags;
> > -	WARN_ON_ONCE(total_sg > vq->packed.vring.num && !vq->indirect);
> > -
> >   	desc = vq->packed.vring.desc;
> >   	i = head;
> >   	descs_used = total_sg;

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 2/3] virtiofs: split requests that exceed virtqueue size

[Stefan Hajnoczi]

[-- Attachment #1.1: Type: text/plain, Size: 1397 bytes --]

On Thu, Mar 18, 2021 at 08:52:22AM -0500, Connor Kuehl wrote:
> If an incoming FUSE request can't fit on the virtqueue, the request is
> placed onto a workqueue so a worker can try to resubmit it later where
> there will (hopefully) be space for it next time.
> 
> This is fine for requests that aren't larger than a virtqueue's maximum
> capacity. However, if a request's size exceeds the maximum capacity of
> the virtqueue (even if the virtqueue is empty), it will be doomed to a
> life of being placed on the workqueue, removed, discovered it won't fit,
> and placed on the workqueue yet again.
> 
> Furthermore, from section 2.6.5.3.1 (Driver Requirements: Indirect
> Descriptors) of the virtio spec:
> 
>   "A driver MUST NOT create a descriptor chain longer than the Queue
>   Size of the device."
> 
> To fix this, limit the number of pages FUSE will use for an overall
> request. This way, each request can realistically fit on the virtqueue
> when it is decomposed into a scattergather list and avoid violating
> section 2.6.5.3.1 of the virtio spec.
> 
> Signed-off-by: Connor Kuehl <ckuehl@redhat.com>
> ---
>  fs/fuse/fuse_i.h    |  5 +++++
>  fs/fuse/inode.c     |  7 +++++++
>  fs/fuse/virtio_fs.c | 14 ++++++++++++++
>  3 files changed, 26 insertions(+)

Nice that FUSE already has max_pages :-).

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[-- Attachment #2: Type: text/plain, Size: 183 bytes --]

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 2/3] virtiofs: split requests that exceed virtqueue size

[Vivek Goyal]

On Thu, Mar 18, 2021 at 04:17:51PM +0100, Miklos Szeredi wrote:
> On Thu, Mar 18, 2021 at 08:52:22AM -0500, Connor Kuehl wrote:
> > If an incoming FUSE request can't fit on the virtqueue, the request is
> > placed onto a workqueue so a worker can try to resubmit it later where
> > there will (hopefully) be space for it next time.
> > 
> > This is fine for requests that aren't larger than a virtqueue's maximum
> > capacity. However, if a request's size exceeds the maximum capacity of
> > the virtqueue (even if the virtqueue is empty), it will be doomed to a
> > life of being placed on the workqueue, removed, discovered it won't fit,
> > and placed on the workqueue yet again.
> > 
> > Furthermore, from section 2.6.5.3.1 (Driver Requirements: Indirect
> > Descriptors) of the virtio spec:
> > 
> >   "A driver MUST NOT create a descriptor chain longer than the Queue
> >   Size of the device."
> > 
> > To fix this, limit the number of pages FUSE will use for an overall
> > request. This way, each request can realistically fit on the virtqueue
> > when it is decomposed into a scattergather list and avoid violating
> > section 2.6.5.3.1 of the virtio spec.
> 
> I removed the conditional compilation and renamed the limit.  Also made
> virtio_fs_get_tree() bail out if it hit the WARN_ON().  Updated patch below.
> 
> The virtio_ring patch in this series should probably go through the respective
> subsystem tree.
> 
> 
> Thanks,
> Miklos
> 
> ---
> From: Connor Kuehl <ckuehl@redhat.com>
> Subject: virtiofs: split requests that exceed virtqueue size
> Date: Thu, 18 Mar 2021 08:52:22 -0500
> 
> If an incoming FUSE request can't fit on the virtqueue, the request is
> placed onto a workqueue so a worker can try to resubmit it later where
> there will (hopefully) be space for it next time.
> 
> This is fine for requests that aren't larger than a virtqueue's maximum
> capacity.  However, if a request's size exceeds the maximum capacity of the
> virtqueue (even if the virtqueue is empty), it will be doomed to a life of
> being placed on the workqueue, removed, discovered it won't fit, and placed
> on the workqueue yet again.
> 
> Furthermore, from section 2.6.5.3.1 (Driver Requirements: Indirect
> Descriptors) of the virtio spec:
> 
>   "A driver MUST NOT create a descriptor chain longer than the Queue
>   Size of the device."
> 
> To fix this, limit the number of pages FUSE will use for an overall
> request.  This way, each request can realistically fit on the virtqueue
> when it is decomposed into a scattergather list and avoid violating section
> 2.6.5.3.1 of the virtio spec.
> 
> Signed-off-by: Connor Kuehl <ckuehl@redhat.com>
> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
> ---

Looks good to me.

Reviewed-by: Vivek Goyal <vgoyal@redhat.com>

Vivek

>  fs/fuse/fuse_i.h    |    3 +++
>  fs/fuse/inode.c     |    3 ++-
>  fs/fuse/virtio_fs.c |   19 +++++++++++++++++--
>  3 files changed, 22 insertions(+), 3 deletions(-)
> 
> --- a/fs/fuse/fuse_i.h
> +++ b/fs/fuse/fuse_i.h
> @@ -555,6 +555,9 @@ struct fuse_conn {
>  	/** Maxmum number of pages that can be used in a single request */
>  	unsigned int max_pages;
>  
> +	/** Constrain ->max_pages to this value during feature negotiation */
> +	unsigned int max_pages_limit;
> +
>  	/** Input queue */
>  	struct fuse_iqueue iq;
>  
> --- a/fs/fuse/inode.c
> +++ b/fs/fuse/inode.c
> @@ -712,6 +712,7 @@ void fuse_conn_init(struct fuse_conn *fc
>  	fc->pid_ns = get_pid_ns(task_active_pid_ns(current));
>  	fc->user_ns = get_user_ns(user_ns);
>  	fc->max_pages = FUSE_DEFAULT_MAX_PAGES_PER_REQ;
> +	fc->max_pages_limit = FUSE_MAX_MAX_PAGES;
>  
>  	INIT_LIST_HEAD(&fc->mounts);
>  	list_add(&fm->fc_entry, &fc->mounts);
> @@ -1040,7 +1041,7 @@ static void process_init_reply(struct fu
>  				fc->abort_err = 1;
>  			if (arg->flags & FUSE_MAX_PAGES) {
>  				fc->max_pages =
> -					min_t(unsigned int, FUSE_MAX_MAX_PAGES,
> +					min_t(unsigned int, fc->max_pages_limit,
>  					max_t(unsigned int, arg->max_pages, 1));
>  			}
>  			if (IS_ENABLED(CONFIG_FUSE_DAX) &&
> --- a/fs/fuse/virtio_fs.c
> +++ b/fs/fuse/virtio_fs.c
> @@ -18,6 +18,12 @@
>  #include <linux/uio.h>
>  #include "fuse_i.h"
>  
> +/* Used to help calculate the FUSE connection's max_pages limit for a request's
> + * size. Parts of the struct fuse_req are sliced into scattergather lists in
> + * addition to the pages used, so this can help account for that overhead.
> + */
> +#define FUSE_HEADER_OVERHEAD    4
> +
>  /* List of virtio-fs device instances and a lock for the list. Also provides
>   * mutual exclusion in device removal and mounting path
>   */
> @@ -1413,9 +1419,10 @@ static int virtio_fs_get_tree(struct fs_
>  {
>  	struct virtio_fs *fs;
>  	struct super_block *sb;
> -	struct fuse_conn *fc;
> +	struct fuse_conn *fc = NULL;
>  	struct fuse_mount *fm;
> -	int err;
> +	unsigned int virtqueue_size;
> +	int err = -EIO;
>  
>  	/* This gets a reference on virtio_fs object. This ptr gets installed
>  	 * in fc->iq->priv. Once fuse_conn is going away, it calls ->put()
> @@ -1427,6 +1434,10 @@ static int virtio_fs_get_tree(struct fs_
>  		return -EINVAL;
>  	}
>  
> +	virtqueue_size = virtqueue_get_vring_size(fs->vqs[VQ_REQUEST].vq);
> +	if (WARN_ON(virtqueue_size <= FUSE_HEADER_OVERHEAD))
> +		goto out_err;
> +
>  	err = -ENOMEM;
>  	fc = kzalloc(sizeof(struct fuse_conn), GFP_KERNEL);
>  	if (!fc)
> @@ -1442,6 +1453,10 @@ static int virtio_fs_get_tree(struct fs_
>  	fc->delete_stale = true;
>  	fc->auto_submounts = true;
>  
> +	/* Tell FUSE to split requests that exceed the virtqueue's size */
> +	fc->max_pages_limit = min_t(unsigned int, fc->max_pages_limit,
> +				    virtqueue_size - FUSE_HEADER_OVERHEAD);
> +
>  	fsc->s_fs_info = fm;
>  	sb = sget_fc(fsc, virtio_fs_test_super, set_anon_super_fc);
>  	if (fsc->s_fs_info) {
> 

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 1/3] virtio_ring: always warn when descriptor chain exceeds queue size

[Jason Wang]

在 2021/3/22 下午4:17, Michael S. Tsirkin 写道:
> On Mon, Mar 22, 2021 at 11:22:15AM +0800, Jason Wang wrote:
>> 在 2021/3/18 下午9:52, Connor Kuehl 写道:
>>>   From section 2.6.5.3.1 (Driver Requirements: Indirect Descriptors)
>>> of the virtio spec:
>>>
>>>     "A driver MUST NOT create a descriptor chain longer than the Queue
>>>     Size of the device."
>>>
>>> This text suggests that the warning should trigger even if
>>> indirect descriptors are in use.
>>
>> So I think at least the commit log needs some tweak.
>>
>> For split virtqueue. We had:
>>
>> 2.6.5.2 Driver Requirements: The Virtqueue Descriptor Table
>>
>> Drivers MUST NOT add a descriptor chain longer than 2^32 bytes in total;
>> this implies that loops in the descriptor chain are forbidden!
>>
>> 2.6.5.3.1 Driver Requirements: Indirect Descriptors
>>
>> A driver MUST NOT create a descriptor chain longer than the Queue Size of
>> the device.
>>
>> If I understand the spec correctly, the check is only needed for a single
>> indirect descriptor table?
>>
>> For packed virtqueue. We had:
>>
>> 2.7.17 Driver Requirements: Scatter-Gather Support
>>
>> A driver MUST NOT create a descriptor list longer than allowed by the
>> device.
>>
>> A driver MUST NOT create a descriptor list longer than the Queue Size.
>>
>> 2.7.19 Driver Requirements: Indirect Descriptors
>>
>> A driver MUST NOT create a descriptor chain longer than allowed by the
>> device.
>>
>> So it looks to me the packed part is fine.
>>
>> Note that if I understand the spec correctly 2.7.17 implies 2.7.19.
>>
>> Thanks
> It would be quite strange for packed and split to differ here:
> so for packed would you say there's no limit on # of descriptors at all?
>
> I am guessing I just forgot to move this part from
> the format specific to the common part of the spec.
>
> This needs discussion in the TC mailing list - want to start a thread
> there?


Will do.

Thanks


>
>
>
>>> Reported-by: Stefan Hajnoczi <stefanha@redhat.com>
>>> Signed-off-by: Connor Kuehl <ckuehl@redhat.com>
>>> ---
>>>    drivers/virtio/virtio_ring.c | 7 ++++---
>>>    1 file changed, 4 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
>>> index 71e16b53e9c1..1bc290f9ba13 100644
>>> --- a/drivers/virtio/virtio_ring.c
>>> +++ b/drivers/virtio/virtio_ring.c
>>> @@ -444,11 +444,12 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
>>>    	head = vq->free_head;
>>> +	WARN_ON_ONCE(total_sg > vq->split.vring.num);
>>> +
>>>    	if (virtqueue_use_indirect(_vq, total_sg))
>>>    		desc = alloc_indirect_split(_vq, total_sg, gfp);
>>>    	else {
>>>    		desc = NULL;
>>> -		WARN_ON_ONCE(total_sg > vq->split.vring.num && !vq->indirect);
>>>    	}
>>>    	if (desc) {
>>> @@ -1118,6 +1119,8 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>>>    	BUG_ON(total_sg == 0);
>>> +	WARN_ON_ONCE(total_sg > vq->packed.vring.num);
>>> +
>>>    	if (virtqueue_use_indirect(_vq, total_sg))
>>>    		return virtqueue_add_indirect_packed(vq, sgs, total_sg,
>>>    				out_sgs, in_sgs, data, gfp);
>>> @@ -1125,8 +1128,6 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>>>    	head = vq->packed.next_avail_idx;
>>>    	avail_used_flags = vq->packed.avail_used_flags;
>>> -	WARN_ON_ONCE(total_sg > vq->packed.vring.num && !vq->indirect);
>>> -
>>>    	desc = vq->packed.vring.desc;
>>>    	i = head;
>>>    	descs_used = total_sg;

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 2/3] virtiofs: split requests that exceed virtqueue size

[Connor Kuehl]

On 3/18/21 10:17 AM, Miklos Szeredi wrote:
> I removed the conditional compilation and renamed the limit.  Also made
> virtio_fs_get_tree() bail out if it hit the WARN_ON().  Updated patch below.

Hi Miklos,

Has this patch been queued?

Connor

> ---
> From: Connor Kuehl <ckuehl@redhat.com>
> Subject: virtiofs: split requests that exceed virtqueue size
> Date: Thu, 18 Mar 2021 08:52:22 -0500
> 
> If an incoming FUSE request can't fit on the virtqueue, the request is
> placed onto a workqueue so a worker can try to resubmit it later where
> there will (hopefully) be space for it next time.
> 
> This is fine for requests that aren't larger than a virtqueue's maximum
> capacity.  However, if a request's size exceeds the maximum capacity of the
> virtqueue (even if the virtqueue is empty), it will be doomed to a life of
> being placed on the workqueue, removed, discovered it won't fit, and placed
> on the workqueue yet again.
> 
> Furthermore, from section 2.6.5.3.1 (Driver Requirements: Indirect
> Descriptors) of the virtio spec:
> 
>    "A driver MUST NOT create a descriptor chain longer than the Queue
>    Size of the device."
> 
> To fix this, limit the number of pages FUSE will use for an overall
> request.  This way, each request can realistically fit on the virtqueue
> when it is decomposed into a scattergather list and avoid violating section
> 2.6.5.3.1 of the virtio spec.
> 
> Signed-off-by: Connor Kuehl <ckuehl@redhat.com>
> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
> ---
>   fs/fuse/fuse_i.h    |    3 +++
>   fs/fuse/inode.c     |    3 ++-
>   fs/fuse/virtio_fs.c |   19 +++++++++++++++++--
>   3 files changed, 22 insertions(+), 3 deletions(-)
> 
> --- a/fs/fuse/fuse_i.h
> +++ b/fs/fuse/fuse_i.h
> @@ -555,6 +555,9 @@ struct fuse_conn {
>   	/** Maxmum number of pages that can be used in a single request */
>   	unsigned int max_pages;
>   
> +	/** Constrain ->max_pages to this value during feature negotiation */
> +	unsigned int max_pages_limit;
> +
>   	/** Input queue */
>   	struct fuse_iqueue iq;
>   
> --- a/fs/fuse/inode.c
> +++ b/fs/fuse/inode.c
> @@ -712,6 +712,7 @@ void fuse_conn_init(struct fuse_conn *fc
>   	fc->pid_ns = get_pid_ns(task_active_pid_ns(current));
>   	fc->user_ns = get_user_ns(user_ns);
>   	fc->max_pages = FUSE_DEFAULT_MAX_PAGES_PER_REQ;
> +	fc->max_pages_limit = FUSE_MAX_MAX_PAGES;
>   
>   	INIT_LIST_HEAD(&fc->mounts);
>   	list_add(&fm->fc_entry, &fc->mounts);
> @@ -1040,7 +1041,7 @@ static void process_init_reply(struct fu
>   				fc->abort_err = 1;
>   			if (arg->flags & FUSE_MAX_PAGES) {
>   				fc->max_pages =
> -					min_t(unsigned int, FUSE_MAX_MAX_PAGES,
> +					min_t(unsigned int, fc->max_pages_limit,
>   					max_t(unsigned int, arg->max_pages, 1));
>   			}
>   			if (IS_ENABLED(CONFIG_FUSE_DAX) &&
> --- a/fs/fuse/virtio_fs.c
> +++ b/fs/fuse/virtio_fs.c
> @@ -18,6 +18,12 @@
>   #include <linux/uio.h>
>   #include "fuse_i.h"
>   
> +/* Used to help calculate the FUSE connection's max_pages limit for a request's
> + * size. Parts of the struct fuse_req are sliced into scattergather lists in
> + * addition to the pages used, so this can help account for that overhead.
> + */
> +#define FUSE_HEADER_OVERHEAD    4
> +
>   /* List of virtio-fs device instances and a lock for the list. Also provides
>    * mutual exclusion in device removal and mounting path
>    */
> @@ -1413,9 +1419,10 @@ static int virtio_fs_get_tree(struct fs_
>   {
>   	struct virtio_fs *fs;
>   	struct super_block *sb;
> -	struct fuse_conn *fc;
> +	struct fuse_conn *fc = NULL;
>   	struct fuse_mount *fm;
> -	int err;
> +	unsigned int virtqueue_size;
> +	int err = -EIO;
>   
>   	/* This gets a reference on virtio_fs object. This ptr gets installed
>   	 * in fc->iq->priv. Once fuse_conn is going away, it calls ->put()
> @@ -1427,6 +1434,10 @@ static int virtio_fs_get_tree(struct fs_
>   		return -EINVAL;
>   	}
>   
> +	virtqueue_size = virtqueue_get_vring_size(fs->vqs[VQ_REQUEST].vq);
> +	if (WARN_ON(virtqueue_size <= FUSE_HEADER_OVERHEAD))
> +		goto out_err;
> +
>   	err = -ENOMEM;
>   	fc = kzalloc(sizeof(struct fuse_conn), GFP_KERNEL);
>   	if (!fc)
> @@ -1442,6 +1453,10 @@ static int virtio_fs_get_tree(struct fs_
>   	fc->delete_stale = true;
>   	fc->auto_submounts = true;
>   
> +	/* Tell FUSE to split requests that exceed the virtqueue's size */
> +	fc->max_pages_limit = min_t(unsigned int, fc->max_pages_limit,
> +				    virtqueue_size - FUSE_HEADER_OVERHEAD);
> +
>   	fsc->s_fs_info = fm;
>   	sb = sget_fc(fsc, virtio_fs_test_super, set_anon_super_fc);
>   	if (fsc->s_fs_info) {
> 

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH 2/3] virtiofs: split requests that exceed virtqueue size

[Connor Kuehl]

On 3/24/21 10:30 AM, Miklos Szeredi wrote:
> On Wed, Mar 24, 2021 at 4:09 PM Connor Kuehl <ckuehl@redhat.com> wrote:
>>
>> On 3/18/21 10:17 AM, Miklos Szeredi wrote:
>>> I removed the conditional compilation and renamed the limit.  Also made
>>> virtio_fs_get_tree() bail out if it hit the WARN_ON().  Updated patch below.
>>
>> Hi Miklos,
>>
>> Has this patch been queued?
> 
> It's in my internal patch queue at the moment.   Will push to
> fuse.git#for-next in a couple of days.

Cool! Thank you :-)

Connor

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization