[External] Re: [PATCH] RISC-V: cacheinfo: add init_cache_level()

Sudeep Holla sudeep.holla at arm.com
Mon Jan 22 09:02:55 PST 2024


On Mon, Jan 22, 2024 at 06:57:49PM +0800, yunhui cui wrote:
> Hi Conor,
> 
> On Mon, Jan 22, 2024 at 4:55 PM Conor Dooley <conor at kernel.org> wrote:
> >
> > On Mon, Jan 22, 2024 at 04:32:15PM +0800, yunhui cui wrote:
> > > Hi Conor,
> > >
> > > On Mon, Jan 22, 2024 at 4:09 PM Conor Dooley <conor at kernel.org> wrote:
> > > >
> > > > On Mon, Jan 22, 2024 at 09:35:10AM +0800, Yunhui Cui wrote:
> > > > > When cacheinfo_sysfs_init() is executed, the general weak function
> > > > > init_cache_level() returns -ENOENT, causing failure to add the "cache"
> > > > > node to /sys/devices/system/cpu/cpux/. Implement the init_cache_level()
> > > > > function on RISC-V to fix it.
> > > >
> > > > If you recall correctly, I asked you to explain how to reproduce this
> > > > when you sent the patch.
> > >
> > > In fact, the reason has been explained in the commit log. As for how
> > > to reproduce it, you can check whether there is a "cache" node in
> > > /sys/devices/system/cpu/cpux/ on the riscv platform.
> >
> > That's the thing - I tried to reproduce this several times and either:
> > a) The system had cache information in DT and the directory was
> >    created. If I hot unplugged and re-plugged the directory was
> >    re-created.
> > b) The system had no cache information in DT and the directory was never
> >    created.
> 
> Indeed, I verified it again, it’s because there is no cache node in
> dts, thank you for reminding me.
>

That's good. I wasn't able to figure out why the issue could occur without
issues in DT/ACPI.

-- 
Regards,
Sudeep



More information about the linux-riscv mailing list