Re: [Xen-devel] [PATCH v4 21/24] tools: L2 CAT: support get HW info for L2 CAT.

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 10.01.17 at 09:00, <yi.y.sun@xxxxxxxxxxxxxxx> wrote:
> On 17-01-09 09:26:17, Wei Liu wrote:
>> On Mon, Jan 09, 2017 at 01:31:35AM -0700, Jan Beulich wrote:
>> > >>> On 09.01.17 at 02:19, <yi.y.sun@xxxxxxxxxxxxxxx> wrote:
>> > > On 17-01-06 12:04:36, Wei Liu wrote:
>> > >> On Wed, Dec 14, 2016 at 12:08:01PM +0800, Yi Sun wrote:
>> > >> > This patch implements xl/xc changes to support get HW info
>> > >> > for L2 CAT.
>> > >> > 
>> > >> > 'xl psr-hwinfo' is updated to show both L3 CAT and L2 CAT
>> > >> > info.
>> > >> > 
>> > >> > Example(on machine which only supports L2 CAT):
>> > >> > Cache Monitoring Technology (CMT):
>> > >> > Enabled         : 0
>> > >> > Cache Allocation Technology (CAT): L3
>> > >> > libxl: error: libxl_psr.c:100:libxl__psr_cat_log_err_msg: CAT is not 
>> > >> > enabled on the socket: No such file or directory
>> > >> > Failed to get l3 cat info
>> > >> 
>> > >> What is this?
>> > >> 
>> > > The psr-hwinfo prints all features out by design. This test was executed 
>> > > on a
>> > > machine which only supports L2 CAT. So, the L3 CAT not supported info 
>> > > will be
>> > > printed out.
>> > 
>> > But there is an indication of error(s), which shouldn't be the case just
>> > because of non-present hardware. The hypercall may return failure
>> > in such a case of course, but this shouldn't propagate to error/failure
>> > messages being printed (but at the same time you need to make sure
>> > you don't hide actual errors). Perhaps there's a call missing querying
>> > host capabilities (based on the result of which further info queries
>> > would be issued)?
>> > 
>> 
>> This is what I thought as well.
>> 
> I considered to add a sysctl interface to get feature mask back before.
> Then, the application can print enabled feature list out in psr-hwinfo.
> 
> But current psr-hwinfo command implementation can have same effect. It
> gets all features HW info back. For not enabled feature, such error info
> is printed. Although it is not user friendly, it reduces one hypercall.
> 
> So, I have below two options:
> 1) Create a new sysctl interface to get feature mask and get HW info
>    according to it.
> 2) Modify current psr-hwinfo codes only to show enabled feature HW info
>    but not show error logs. Of course the libxl library interface still
>    returns error to application.
> 
> Which one do you prefer? Thanks!

Well, afaic I'd prefer any variant without new hypercall, as long as
the bogus error messages go away without hiding any actual errors.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel