Re: [PATCH 1/2] CPU, NUMA topology ABIs: clarify the overflow issue of sysfs pagebuf

["gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>]

On Fri, Jul 23, 2021 at 11:20:19AM +0000, Song Bao Hua (Barry Song) wrote:
> 
> 
> > -----Original Message-----
> > From: Dave Hansen [mailto:dave.hansen@intel.com]
> > Sent: Friday, April 30, 2021 10:39 AM
> > To: Song Bao Hua (Barry Song) <song.bao.hua@hisilicon.com>; tiantao (H)
> > <tiantao6@hisilicon.com>; corbet@lwn.net; gregkh@linuxfoundation.org
> > Cc: linux-doc@vger.kernel.org; linux-kernel@vger.kernel.org; Rafael J.
> > Wysocki <rafael@kernel.org>; Peter Zijlstra <peterz@infradead.org>; Valentin
> > Schneider <valentin.schneider@arm.com>; Dave Hansen
> > <dave.hansen@linux.intel.com>; Daniel Bristot de Oliveira <bristot@redhat.com>
> > Subject: Re: [PATCH 1/2] CPU, NUMA topology ABIs: clarify the overflow issue
> > of sysfs pagebuf
> > 
> > On 4/29/21 3:32 PM, Song Bao Hua (Barry Song) wrote:
> > > $ strace numactl --hardware  2>&1 | grep cpu
> > > openat(AT_FDCWD, "/sys/devices/system/cpu",
> > > O_RDONLY|O_NONBLOCK|O_DIRECTORY|O_CLOEXEC) = 3
> > > openat(AT_FDCWD, "/sys/devices/system/node/node0/cpumap", O_RDONLY) = 3
> > > openat(AT_FDCWD, "/sys/devices/system/node/node1/cpumap", O_RDONLY) = 3
> > > openat(AT_FDCWD, "/sys/devices/system/node/node2/cpumap", O_RDONLY) = 3
> > > openat(AT_FDCWD, "/sys/devices/system/node/node3/cpumap", O_RDONLY) = 3
> > >
> > > If we move to binary, it means we have to change those applications.
> > 
> > I thought Greg was saying to using a sysfs binary attribute using
> > something like like sysfs_create_bin_file().  Those don't have the
> > PAGE_SIZE limitation.  But, there's also nothing to keep us from spewing
> > nice human-readable text via the "binary" file.
> > 
> > We don't need to change the file format, just the internal kernel API
> > that we produce the files with.
> 
> Sorry for waking-up the old thread.
> 
> I am not sure how common a regular device_attribute will be larger than
> 4KB and have to work around by bin_attribute. But I wrote a prototype
> which can initially support large regular sysfs entry and be able to
> fill the entire buffer by only one show() entry. The other words to say,
> we don't need to call read() of bin_attribute multiple times for a large
> regular file. Right now, only read-only attribute is supported.
> 
> Subject: [RFC PATCH] sysfs: support regular device attr which can be larger than
>  PAGE_SIZE
> 
> A regular sysfs ABI could be more than 4KB, right now, we are using
> bin_attribute to workaround and break this limit. A better solution
> would be supporting long device attribute. In this case, we will
> still be able to enjoy the advantages of buffering and seeking of
> seq file and only need to fill the entire buffer of sysfs entry
> once.

No, please no.  I WANT people to run into this problem and realize that
it went totally wrong because they should not be having more than one
"value" in a sysfs file like this.

Let's not make it easy on people please, moving to a bin attribute is a
big deal, let's keep it that way.

thanks,

greg k-h