| Message ID | 1426069565-7933-1-git-send-email-wsa@the-dreams.de (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On Wed, Mar 11, 2015 at 11:26:02AM +0100, Wolfram Sang wrote: > From: Wolfram Sang <wsa+renesas@sang-engineering.com> > > Grabbing the parent is not happening anymore since 2010 (e72ceb8ccac5f7 > "sysfs: Remove sysfs_get/put_active_two"). Remove this confusing > comment. > > Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Acked-by: Tejun Heo <tj@kernel.org> > Found while trying to find out why my sysfs-bin file always returns POLLPRI > after opening, reading, polling. Any pointers while I am here? :) Hmm... poll isn't hooked up for bin files and the behavior isn't well defined either. The open file's event counter is never synchronized to the backing node's. Before the first kernfs_notify(), it won't trigger and after always. We'll prolly need to define that better. What's the use case that you're interested in? Thanks.
> Hmm... poll isn't hooked up for bin files and the behavior isn't well Ah, thanks! I assumed something like that but could not find the relevant code. > defined either. The open file's event counter is never synchronized > to the backing node's. Before the first kernfs_notify(), it won't > trigger and after always. We'll prolly need to define that better. > What's the use case that you're interested in? I wanted to add notification to drivers/i2c/i2c-slave-eeprom.c which is a software backend when the Linux system is acting as an I2C slave device. This is basically a dual mem driver. One side is exposed to the system as a bin-file in sysfs. The other side is accessible like an EEPROM for other I2C masters connected to the bus. The notification should report to userspace when a remote master has written something to us so the memory contents changed. I know this is not the best way of notification because we don't know what changed and would need to compare. However, the whole I2C slave framework is still very young and I wanted to play with it for the case where only a few bytes are exposed.
On Wed, Mar 11, 2015 at 07:45:31AM -0400, Tejun Heo wrote: > On Wed, Mar 11, 2015 at 11:26:02AM +0100, Wolfram Sang wrote: > > From: Wolfram Sang <wsa+renesas@sang-engineering.com> > > > > Grabbing the parent is not happening anymore since 2010 (e72ceb8ccac5f7 > > "sysfs: Remove sysfs_get/put_active_two"). Remove this confusing > > comment. > > > > Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> > > Acked-by: Tejun Heo <tj@kernel.org> So, how does it go in? gkh? akpm? Thanks, Wolfram
On Thu, Apr 23, 2015 at 02:51:26PM +0200, Wolfram Sang wrote: > On Wed, Mar 11, 2015 at 07:45:31AM -0400, Tejun Heo wrote: > > On Wed, Mar 11, 2015 at 11:26:02AM +0100, Wolfram Sang wrote: > > > From: Wolfram Sang <wsa+renesas@sang-engineering.com> > > > > > > Grabbing the parent is not happening anymore since 2010 (e72ceb8ccac5f7 > > > "sysfs: Remove sysfs_get/put_active_two"). Remove this confusing > > > comment. > > > > > > Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> > > > > Acked-by: Tejun Heo <tj@kernel.org> > > So, how does it go in? gkh? akpm? Usually Greg, ah, who isn't cc'd. Can you please repost the patch w/ Greg cc'd and my ack added? Thanks.
diff --git a/fs/kernfs/file.c b/fs/kernfs/file.c index b684e8a132e625..a893c634cd9c0b 100644 --- a/fs/kernfs/file.c +++ b/fs/kernfs/file.c @@ -784,7 +784,6 @@ static unsigned int kernfs_fop_poll(struct file *filp, poll_table *wait) struct kernfs_node *kn = filp->f_path.dentry->d_fsdata; struct kernfs_open_node *on = kn->attr.open; - /* need parent for the kobj, grab both */ if (!kernfs_get_active(kn)) goto trigger;