Re: [PATCH v3 12/17] block/block-backend: convert blk io path to use int64_t parameters

[Eric Blake <eblake@redhat.com>]

On 4/30/20 6:10 AM, Vladimir Sementsov-Ogievskiy wrote:
> We are generally moving to int64_t for both offset and bytes parameters
> on all io paths.
> 
> Main motivation is realization of 64-bit write_zeroes operation for
> fast zeroing large disk chunks, up to the whole disk.
> 
> We chose signed type, to be consistent with off_t (which is signed) and
> with possibility for signed return type (where negative value means
> error).
> 
> Now bdrv layer is converted, convert blk layer too.

In fact, I just discovered thanks to 
https://bugs.launchpad.net/qemu/+bug/1884831 that NBD is a case of a 
client that can currently pass values larger than 2G into 
blk_co_pdiscard() which in turn appears as a negative value and instant 
EIO failure.  So this is a bug fix visible to NBD clients.

$ gdb --args ./qemu-nbd --trace=nbd_\* -f raw f --port 10810
...
(gdb) b blk_co_pdiscard
(gdb) r
...
$ nbdsh -u nbd://localhost:10810 -c 'h.trim(3*1024*1024*1024,0)'
...
Thread 1 "qemu-nbd" hit Breakpoint 3, blk_co_pdiscard (blk=0x555555832dc0,
     offset=0, bytes=-1073741824)

Looks like I now have even more reason to accelerate my review of the 
remainder of this series, and to take some (if not all) of it through 
the NBD tree.


> +++ b/include/sysemu/block-backend.h
> @@ -119,14 +119,14 @@ BlockBackend *blk_by_dev(void *dev);
>   BlockBackend *blk_by_qdev_id(const char *id, Error **errp);
>   void blk_set_dev_ops(BlockBackend *blk, const BlockDevOps *ops, void *opaque);
>   int coroutine_fn blk_co_preadv(BlockBackend *blk, int64_t offset,
> -                               unsigned int bytes, QEMUIOVector *qiov,
> +                               int64_t bytes, QEMUIOVector *qiov,
>                                  BdrvRequestFlags flags);
>   int coroutine_fn blk_co_pwritev_part(BlockBackend *blk, int64_t offset,
> -                                     unsigned int bytes,
> +                                     int64_t bytes,
>                                        QEMUIOVector *qiov, size_t qiov_offset,
>                                        BdrvRequestFlags flags);
>   int coroutine_fn blk_co_pwritev(BlockBackend *blk, int64_t offset,
> -                               unsigned int bytes, QEMUIOVector *qiov,
> +                               int64_t bytes, QEMUIOVector *qiov,
>                                  BdrvRequestFlags flags);

pread and pwrite weren't necessarily problems for NBD (since our NBD 
implementation caps things to 32M per packet).

>   
>   static inline int coroutine_fn blk_co_pread(BlockBackend *blk, int64_t offset,
> @@ -148,13 +148,13 @@ static inline int coroutine_fn blk_co_pwrite(BlockBackend *blk, int64_t offset,
>   }
>   
>   int blk_pwrite_zeroes(BlockBackend *blk, int64_t offset,
> -                      int bytes, BdrvRequestFlags flags);
> +                      int64_t bytes, BdrvRequestFlags flags);
>   BlockAIOCB *blk_aio_pwrite_zeroes(BlockBackend *blk, int64_t offset,
> -                                  int bytes, BdrvRequestFlags flags,
> +                                  int64_t bytes, BdrvRequestFlags flags,
>                                     BlockCompletionFunc *cb, void *opaque);

But this change to writing zeroes,

>   int blk_make_zero(BlockBackend *blk, BdrvRequestFlags flags);
> -int blk_pread(BlockBackend *blk, int64_t offset, void *buf, int bytes);
> -int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int bytes,
> +int blk_pread(BlockBackend *blk, int64_t offset, void *buf, int64_t bytes);
> +int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int64_t bytes,
>                  BdrvRequestFlags flags);
>   int64_t blk_getlength(BlockBackend *blk);
>   void blk_get_geometry(BlockBackend *blk, uint64_t *nb_sectors_ptr);
> @@ -167,14 +167,14 @@ BlockAIOCB *blk_aio_pwritev(BlockBackend *blk, int64_t offset,
>                               BlockCompletionFunc *cb, void *opaque);
>   BlockAIOCB *blk_aio_flush(BlockBackend *blk,
>                             BlockCompletionFunc *cb, void *opaque);
> -BlockAIOCB *blk_aio_pdiscard(BlockBackend *blk, int64_t offset, int bytes,
> +BlockAIOCB *blk_aio_pdiscard(BlockBackend *blk, int64_t offset, int64_t bytes,
>                                BlockCompletionFunc *cb, void *opaque);

and this change to discard are definitely both bug fixes for NBD 
clients, especially now that we have a real-world case of a client 
(namely the blkdiscard app triggering ioctl(BLKDISCARD) handling through 
nbd.ko as client) that actually triggers a >2G trim request.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org