On Tue, Jan 14, 2014 at 05:20:33PM +0000, Al Viro wrote:
> And that's less than half of fs/*... I'm not saying that the current
> situation on the write side is good; hell, just the mess with write/aio_write
> alone is ugly enough - we have
> * a bunch of file_operations without ->aio_write(); simple enough.
> * a bunch with ->write == do_sync_write. Also simple.
> * several with NULL ->write and non-NULL ->aio_write(); same as
> do_sync_write() for ->write (socket, android/logger, kmsg, macvtap)
> * several with ->aio_write being an optimized equivalent of
> do_sync_write() (blackhole for /dev/null and /dev/zero, error for bad_inode)
> * 9p cached with its "oh, but if we have O_DIRECT we want ->write()
> to be different" (why not use a separate file_operations, then? It's not
> as if ->open() couldn't switch to it if it sees O_DIRECT...)
> * two infinibad things (ipath and qib), with completely unrelated
> semantics on write(2) and writev(2) (the latter shared with aio). As in
> "writev() of a single-element iovec array does things that do not even
> resemble what write() of the same data would've done". Yes, really - check
> for yourself.
> * snd_pcm - hell knows; it might be that it tries to collect the
> data from iovec and push it in one go, as if it was a single write, but
> then it might be something as bogus as what ipath is doing...
> * gadgetfs - hell knows; ep_write() seems to be doing something
> beyond what ep_aio_write() does, but I haven't traced them down the call
> chain... That one, BTW, won't be fun for splice - looks like it cares
> about datagram boundaries a lot, so it's not obvious what the semantics
> should be.
> * lustre. I _think_ do_sync_write() would work there, but I'm might
> be easily missing something in all those layers of obfusca^Wgood software
> development practices.
BTW, ->read/->aio_read situation is only slighlty better - of file_operations
instances that have ->aio_read, most have do_sync_read() for ->read() (as
they ought to). Exceptions:
* 9p O_DIRECT (again)
* NULL ->read where do_sync_read ought to be (socket, macvtap)
* optimized ->read (/dev/zero, /dev/null, bad_inode)
* snd_pcm - magic. It (and its ->aio_write counterpart) wants exactly
one iovec per channel. IOW, it's not a general-purpose ->aio_{read,write}
at all - it's a magic API shoehorned into readv(2)/writev(2) (and aio
IOCB_CMD_P{READ,WRITE}V as well).
* lustre - probably could live with do_sync_read(), but there might
be stack footprint considerations or some really weird magic going on
(the difference is that instead of iocb on stack they appear to be using
per-thread one allocated on heap and hashed by pid, of all things).
It's really weird - they end up doing repeated hash lookups for that
per-thread wastebasket of a structure on different levels of call chain.
Looks like they have swept a lot of local variables of a lot of functions
into that thing; worse, it appears to be one of several dynamically allocated
bits of that thing, hidden behind a bunch of wrappers... Overall feel is
Lovecraftian, complete with lurking horrors of the deep... BTW, its ->aio_read
would better never return -EIOCBQUEUED - its ->read does *not* wait for
completion of iocb it has submitted.
* gadgetfs - it appears to be seriously datagram-oriented; basically,
they want to reduce readv/writev to read/write, not the other way round.
> BTW, speaking of ->aio_write() - why the devil do we pass the pos
> argument (long long, at that) separately, when all call sites provably
> have it equal to iocb->ki_pos? If nothing else, on a bunch of architectures
> it makes the difference between passing arguments in registers and spilling
> them on stack; moreover, if we do something and only then call
> generic_file_aio_write(), we get to propagate it all way down. And
> generic_file_aio_write() has had explicit BUG_ON(iocb->ki_pos != pos)
> since 2.5.55, for crying out loud...
The same goes for ->aio_read() (except for s/2.5.55/2.5.39/)...
|