Re: [PATCH] sysfs/cpu: Make crash_hotplug attribute world-readable

From: "Petr Tesařík" <petr@tesarici.cz>
To: Sourabh Jain <sourabhjain@linux.ibm.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	"Rafael J. Wysocki" <rafael@kernel.org>,
	Eric Biederman <ebiederm@xmission.com>,
	kexec@lists.infradead.org, linux-kernel@vger.kernel.org,
	Petr Tesarik <ptesarik@suse.com>
Subject: Re: [PATCH] sysfs/cpu: Make crash_hotplug attribute world-readable
Date: Fri, 12 Jul 2024 08:21:24 +0200	[thread overview]
Message-ID: <20240712082124.272e079f@meshulam.tesarici.cz> (raw)
In-Reply-To: <59d88de7-369a-44e8-a965-109c8690869e@linux.ibm.com>

On Fri, 12 Jul 2024 10:00:09 +0530
Sourabh Jain <sourabhjain@linux.ibm.com> wrote:

> Hello Petr,
> 
> Just want to mentioned few points about the commit message. No strong
> opinion on changing the commit message.
> 
> On 11/07/24 16:04, Petr Tesarik wrote:
> > From: Petr Tesarik <ptesarik@suse.com>
> >
> > There is no reason to restrict access to this attribute, as it merely
> > reports whether crash elfcorehdr is automatically updated on CPU hot
> > plug/unplug and/or online/offline events.  
> 
> Now, it is not just elfcorehdr; there could be multiple kexec segments 
> based on the architecture's needs.
> For example, on PowerPC, it is elfcorehdr and FDT.

Right. I am even working with a PowerVM LPAR on an L922 system now. ;-)

OTOH I was just too lazy to invent my own description, so I took it
from Documentation/ABI/testing/sysfs-devices-system-cpu. That one
could be improved. And a couple of other places... Expect a patch.

> >
> > Note that since commit 79365026f8694 ("crash: add a new kexec flag
> > for hotplug support"), this maps to the same flag which is
> > world-accessible through /sys/devices/system/memory/crash_hotplug.  
> 
> The previous comment is more relevant for the kexec_load system call.
> 
> For the kexec_file_load system call, the kexec flag introduced in the 
> commit message
> mentioned above is irrelevant. Building kernel with
> CONFIG_CRASH_HOTPLUG is enough.

Yes, but then this sysfs attribute always reads "1".

> > Signedoff-by: Petr Tesarik <ptesarik@suse.com>
> > ---
> >   drivers/base/cpu.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/base/cpu.c b/drivers/base/cpu.c
> > index c61ecb0c2ae2..73d69791d0d3 100644
> > --- a/drivers/base/cpu.c
> > +++ b/drivers/base/cpu.c
> > @@ -308,7 +308,7 @@ static ssize_t crash_hotplug_show(struct device
> > *dev, {
> >   	return sysfs_emit(buf, "%d\n",
> > crash_check_hotplug_support()); }
> > -static DEVICE_ATTR_ADMIN_RO(crash_hotplug);
> > +static DEVICE_ATTR_RO(crash_hotplug);  
> 
> I agree, it doesn't have to be readable by admin only.
> 
> Acked-by: Sourabh Jain <sourabhjain@linux.ibm.com>

Thank you!

Petr T

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec