RE: [PATCH 0/2] use DEFINE_DEBUGFS_ATTRIBUTE instead of DEFINE_SIMPLE_ATTRIBUTE

[Sa, Nuno]

> From: Rohit Sarkar <rohitsarkar5398@xxxxxxxxx>
> Sent: Dienstag, 31. März 2020 13:50
> To: Lars-Peter Clausen <lars@xxxxxxxxxx>
> Cc: Sa, Nuno <Nuno.Sa@xxxxxxxxxx>; Ardelean, Alexandru
> <alexandru.Ardelean@xxxxxxxxxx>; jic23@xxxxxxxxxx;
> stefan.popa@xxxxxxxxxx; zhongjiang@xxxxxxxxxx; linux-
> iio@xxxxxxxxxxxxxxx; Bogdan, Dragos <Dragos.Bogdan@xxxxxxxxxx>;
> pmeerw@xxxxxxxxxx; knaack.h@xxxxxx; Hennerich, Michael
> <Michael.Hennerich@xxxxxxxxxx>; linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH 0/2] use DEFINE_DEBUGFS_ATTRIBUTE instead of
> DEFINE_SIMPLE_ATTRIBUTE
> 
> Hey Nuno,
> > > > >
> > > > I don't have the exact parts that this patch is touching but I have other
> parts where this patch
> > > > applies and should be same. So, the idea to test this is to read the files in
> debugfs? Maybe also
> > > > some unbind + binding?
> > > >
> > > > I will try to test this still today...
> > >
> > > The stress test is to open the debugfs file, then unbind the device and then
> > > read from the still open debugfs file.
> > Yes, also just to be sure, we need to test DEFINE_DEBUGFS_ATTRIBUTE
> > along with debugfs_create_file_unsafe. I will send out another patch
> > that changes debugfs_create_file to debugfs_create_file_unsafe and then
> > that can be tested.
> Have sent out a v2 that is ready to be tested

So I tested your changes in a different HW but the principle is the same.
It worked as expected. Unbinding the device and doing a read on an opened file
descriptor afterwards returns error...

- Nuno Sá
> > > - Lars
> >
> > Thanks,
> > Rohit
> Thanks,
> Rohit