On Wed, Apr 22, 2015 at 10:34 AM, James Bottomley
<james.bottom...@hansenpartnership.com> wrote:
> On Wed, 2015-04-22 at 09:50 -0700, Andy Lutomirski wrote:
>> On Apr 21, 2015 9:51 PM, "James Bottomley"
>> <james.bottom...@hansenpartnership.com> wrote:
>> >
>> > On Tue, 2015-04-21 at 20:24 -0700, Andy Lutomirski wrote:
>> > > On Tue, Apr 21, 2015 at 7:20 PM, James Bottomley
>> > > <james.bottom...@hansenpartnership.com> wrote:
>> > > > On Tue, 2015-04-21 at 18:58 -0700, Andy Lutomirski wrote:
>> > > >> On Tue, Apr 21, 2015 at 6:21 PM, James Bottomley
>> > > >> <james.bottom...@hansenpartnership.com> wrote:
>> > > >> > Andy, just on the misc device idea, what about triggering the
>> > > >> > capsule
>> > > >> > update from close()? In theory close returns an error code (not
>> > > >> > sure if
>> > > >> > most tools actually check this, though). That means we can do the
>> > > >> > write
>> > > >> > in chunks but pass it in atomically on the close and cat will work
>> > > >> > (provided it checks the return code of close).
>> > > >>
>> > > >> I thought about this but IIRC cat doesn't check the return value from
>> > > >> close.
>> > > >
>> > > > It does in my copy (coreutils-8.23) :
>> > > >
>> > > > if (!STREQ (infile, "-") && close (input_desc) < 0)
>> > > > {
>> > > > error (0, errno, "%s", infile);
>> > > > ok = false;
>> > > > }
>> > > > [...]
>> > > > if (have_read_stdin && close (STDIN_FILENO) < 0)
>> > > > error (EXIT_FAILURE, errno, _("closing standard input"));
>> > > >
>> > >
>> > > True, but it's stdout that we care about, not stdin :(
>> >
>> > Gosh you're determined to force me to wade through this source code,
>> > aren't you? That's handled in lib/closeout.c:
>> >
>> > /* Close standard output. On error, issue a diagnostic and _exit
>> > with status 'exit_failure'.
>> >
>> > ...
>> >
>> >
>> > The point is that, admittedly much to my surprise, it all looks to be
>> > handled by cat ... so we could proceed to have the transaction completed
>> > in close in a misc device (or a sysfs file).
>> >
>> > Unless there are any other rabbits you'd like to pull out of the hat?
>>
>> No, maybe it's okay, unless there's an issue where the error would
>> only be returned on the close of the last reference of the struct
>> file. After all, 'cat foo >/sys/bar' doesn't fully close /sys/bar
>> until after cat exits.
>
> No, cat handles that too. It has an atexit() handler for closing
> stdout.
Indeed. Strace tells me that:
$ cat foo >bar
opens bar and then execs cat, so cat has the only reference to bar.
So no more rabbits from me :)
--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
