[Qemu-devel] [PATCH v2] block: avoid creating too large iovecs in multiwrite_merge

[Qemu-devel] [PATCH v2] block: avoid creating too large iovecs in multiwrite_merge

[Christoph Hellwig]

If we go over the maximum number of iovecs support by syscall we get
back EINVAL from the kernel which translate to I/O errors for the guest.

Add a MAX_IOV defintion for platforms that don't have it.  For now we use
the same 1024 define that's used on Linux and various other platforms,
but until the windows block backend implements some kind of vectored I/O
it doesn't matter.

Signed-off-by: Christoph Hellwig <hch@lst.de>

Index: qemu/block.c
===================================================================
--- qemu.orig/block.c	2010-01-26 10:59:39.757004445 +0100
+++ qemu/block.c	2010-01-26 11:01:38.056023231 +0100
@@ -1689,6 +1689,10 @@ static int multiwrite_merge(BlockDriverS
             merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
         }
 
+        if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
+            merge = 0;
+        }
+
         if (merge) {
             size_t size;
             QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
Index: qemu/qemu-common.h
===================================================================
--- qemu.orig/qemu-common.h	2010-01-26 14:41:40.894254285 +0100
+++ qemu/qemu-common.h	2010-01-26 14:42:27.267275698 +0100
@@ -54,6 +54,10 @@ struct iovec {
     void *iov_base;
     size_t iov_len;
 };
+/*
+ * Use the same value as Linux for now.
+ */
+#define IOV_MAX		1024
 #else
 #include <sys/uio.h>
 #endif

Re: [Qemu-devel] [PATCH v2] block: avoid creating too large iovecs in multiwrite_merge

[Anthony Liguori]

On 01/26/2010 07:49 AM, Christoph Hellwig wrote:
> If we go over the maximum number of iovecs support by syscall we get
> back EINVAL from the kernel which translate to I/O errors for the guest.
>
> Add a MAX_IOV defintion for platforms that don't have it.  For now we use
> the same 1024 define that's used on Linux and various other platforms,
> but until the windows block backend implements some kind of vectored I/O
> it doesn't matter.
>
> Signed-off-by: Christoph Hellwig<hch@lst.de>
>    

Applied.  Thanks.

Regards,

Anthony Liguori
> Index: qemu/block.c
> ===================================================================
> --- qemu.orig/block.c	2010-01-26 10:59:39.757004445 +0100
> +++ qemu/block.c	2010-01-26 11:01:38.056023231 +0100
> @@ -1689,6 +1689,10 @@ static int multiwrite_merge(BlockDriverS
>               merge = bs->drv->bdrv_merge_requests(bs,&reqs[outidx],&reqs[i]);
>           }
>
> +        if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1>  IOV_MAX) {
> +            merge = 0;
> +        }
> +
>           if (merge) {
>               size_t size;
>               QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
> Index: qemu/qemu-common.h
> ===================================================================
> --- qemu.orig/qemu-common.h	2010-01-26 14:41:40.894254285 +0100
> +++ qemu/qemu-common.h	2010-01-26 14:42:27.267275698 +0100
> @@ -54,6 +54,10 @@ struct iovec {
>       void *iov_base;
>       size_t iov_len;
>   };
> +/*
> + * Use the same value as Linux for now.
> + */
> +#define IOV_MAX		1024
>   #else
>   #include<sys/uio.h>
>   #endif
>
>
>
>    

Re: [Qemu-devel] [PATCH v2] block: avoid creating too large iovecs in multiwrite_merge

[Christoph Hellwig]

On Tue, Jan 26, 2010 at 06:07:45PM -0600, Anthony Liguori wrote:
> On 01/26/2010 07:49 AM, Christoph Hellwig wrote:
> >If we go over the maximum number of iovecs support by syscall we get
> >back EINVAL from the kernel which translate to I/O errors for the guest.
> >
> >Add a MAX_IOV defintion for platforms that don't have it.  For now we use
> >the same 1024 define that's used on Linux and various other platforms,
> >but until the windows block backend implements some kind of vectored I/O
> >it doesn't matter.
> >
> >Signed-off-by: Christoph Hellwig<hch@lst.de>
> >   
> 
> Applied.  Thanks.

Can you also add it to the 0.12-stable series?  There's tons of bug
reports from users seeing this regression in qemu-0.12.