Skip to content
GitLab
  1. Mar 26, 2015
  3. Mar 14, 2015
    • Christoph Hellwig's avatar
      fs: split generic and aio kiocb · 04b2fa9f
      Christoph Hellwig authored 
      

Most callers in the kernel want to perform synchronous file I/O, but
still have to bloat the stack with a full struct kiocb.  Split out
the parts needed in filesystem code from those in the aio code, and
only allocate those needed to pass down argument on the stack.  The
aio code embedds the generic iocb in the one it allocates and can
easily get back to it by using container_of.

Also add a ->ki_complete method to struct kiocb, this is used to call
into the aio code and thus removes the dependency on aio for filesystems
impementing asynchronous operations.  It will also allow other callers
to substitute their own completion callback.

We also add a new ->ki_flags field to work around the nasty layering
violation recently introduced in commit 5e33f6 ("usb: gadget: ffs: add
eventfd notification about ffs events").

Signed-off-by: default avatarChristoph Hellwig <hch@lst.de>
Signed-off-by: default avatarAl Viro <viro@zeniv.linux.org.uk>
      04b2fa9f
    • Christoph Hellwig's avatar
      fs: don't allow to complete sync iocbs through aio_complete · 599bd19b
      Christoph Hellwig authored 
      

The AIO interface is fairly complex because it tries to allow
filesystems to always work async and then wakeup a synchronous
caller through aio_complete.  It turns out that basically no one
was doing this to avoid the complexity and context switches,
and we've already fixed up the remaining users and can now
get rid of this case.

Signed-off-by: default avatarChristoph Hellwig <hch@lst.de>
Signed-off-by: default avatarAl Viro <viro@zeniv.linux.org.uk>
      599bd19b
    • Christoph Hellwig's avatar
      fuse: handle synchronous iocbs internally · 9d5722b7
      Christoph Hellwig authored 
      

Based on a patch from Maxim Patlasov <MPatlasov@parallels.com>.

Signed-off-by: default avatarChristoph Hellwig <hch@lst.de>
Signed-off-by: default avatarAl Viro <viro@zeniv.linux.org.uk>
      9d5722b7
  5. Mar 13, 2015
  7. Mar 09, 2015
    • Alan Stern's avatar
      gadgetfs: really get rid of switching ->f_op · 96b62a57
      Alan Stern authored 
      

... for ep0 as well

Signed-off-by: default avatarAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: default avatarAl Viro <viro@zeniv.linux.org.uk>
      96b62a57
    • Al Viro's avatar
      gadgetfs: get rid of flipping ->f_op in ep_config() · d4461a60
      Al Viro authored 
      

Final methods start with get_ready_ep(), which will fail unless we have
->state == STATE_EP_ENABLED.  So they'd be failing just fine until that
first write() anyway.  Let's do the following:
	* get_ready_ep() gets a new argument - true when called from
ep_write_iter(), false otherwise.
	* make it quiet when it finds STATE_EP_READY (no printk, that is;
the case won't be impossible after that change).
	* when that new argument is true, treat STATE_EP_READY the same
way as STATE_EP_ENABLED (i.e. return zero and do not unlock).
	* in ep_write_iter(), after success of get_ready_ep() turn
	if (!usb_endpoint_dir_in(&epdata->desc)) {
into
	if (epdata->state == STATE_EP_ENABLED &&
	    !usb_endpoint_dir_in(&epdata->desc)) {
- that logics only applies after config.
	* have ep_config() take kernel-side buffer (i.e. use memcpy()
instead of copy_from_user() in there) and in the "let's call ep_io or
ep_aio" (again, in ep_write_iter()) add "... or ep_config() in case it's
not configured yet"

Signed-off-by: default avatarAl Viro <viro@zeniv.linux.org.uk>
      d4461a60
  9. Feb 18, 2015