> On 05 August 2014 at 23:32 Zach Brown <z...@zabbo.net> wrote: > > > > > > Hello Zach, > > > > > > > > Here's an untested patch which > > > > > > Try testing it. It's easy with virtualization and xfstests. > > > > > > You'll find that sending to a file fails because each individual file > > > write call that makes up a send starts at offset 0 -- at the start of > > > the file. > > > > > > Getting this right means getting the semantics around updating the send > > > descriptors f_pos right. It requires having a bit of a think about send > > > semantics and f_pos update locking. > > > > Thanks for those informations Zach, > > > > I've tried btrfs test scripts related to ioctl in xfstests (tests/btrfs/025, > > 035, 052, 055) > > but was not able to trigger that problem. Do I have to create another > > script, > > use some generic one > > or maybe use big test/scratch devices ? > > No idea, sorry. Maybe your patch is fine and I'm a dummy. Maybe you > didn't test the kernel you thought you were testing. Maybe the test > doesn't test what you changed. You'll have to do some investigating to > find out.
After checking both kernel and btrfs-progs, it seems file offset is only being passed through TLV(see send.c/ send_write/ TLV_PUT_U64(BTRFS_SEND_A_FILE_OFFSET). Tell me if I'm wrong but when we do 'btrfs send backup | btrfs receive backupvolume', btrfs-progs calls btrfs_ioctl_send emitting several BTRFS_SEND_C_WRITE with relevant BTRFS_SEND_A_FILE_OFFSET (growing by 48kb). At the other side of | , 'btrfs receive' executes read_and_process_cmd which does TLV_GET before s->ops->write. All btrfs interactions being done with | maybe we could simply replace vfs_write stuff as you suggested and remove sctx->send_off ? Regards, Fabian > > - z -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
