Re: [PATCH 02/16] vfs: check kiocb->ki_flags instead filp->fl_flags

[Dmitry Monakhov]

Al Viro <viro@xxxxxxxxxxxxxxxxxx> writes:

> On Sat, Apr 04, 2015 at 11:13:11PM +0400, Dmitry Monakhov wrote:
>> generic_write_checks now accept kiocb as an argument
>> Unfortunetly it is impossible to get rid of old interface because some crappy
>> do not support write_iter interface so leave __generic_write_checks as backward
>> compatibility helper.
>
> Check the current vfs.git#for-next (there's even some generic_write_checks()
> work in it).  The same goes for the rest of the series.  Please, rebase it.
Ok. Yes you right. I've prepared patches against vfs.git#for-next(48a0c62) but
it is too old. Will rebase. Also it is reasonable to fold my fs-xxx
conversion to one combo patch similar to (d04cfe7840)
>
> What's more, generic_write_checks() should take iov_iter *, not the address
> of something its ->count had been copied into.  Note that all callers of that
> thing end up doing iov_iter_truncate() pretty soon afterwards.  I hadn't
> pushed that one out yet (there is some weirdness in ocfs2 which might be
> a bug; I want to sort that out first), but that's where it's going.
I'm not sure I have get your point about ocfs2 because it does
iov_iter_truncate() right after generic_write_checks()
But nfs definitely has a bug because iter was not truncated after generic_write_checks
->nfs_file_direct_write
  ->generic_write_checks(file, &pos, &count);
  dreq->bytes_left = count;
  task_io_account_write(count);
  ->nfs_direct_write_schedule_iovec(dreq, iter, pos);
    ###Ignore dreq->bytes_left(ala count) and submit non truncated iter
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Attachment: signature.asc
Description: PGP signature