Re: [PATCH v24b] RAS: Add a tracepoint for reporting memory controller events

[Borislav Petkov <borislav.petkov@amd.com>]

On Mon, May 21, 2012 at 01:40:08PM -0300, Mauro Carvalho Chehab wrote:
> That's exactly what the latest version of this patch does.

Really, where is the address field?

+       TP_PROTO(const unsigned int err_type,
+                const unsigned int mc_index,
+                const char *error_msg,
+                const char *label,
+                const char *location,
+                const char *core_detail,
+                const char *driver_detail),


[ … ]

> Huh? Nobody said to add a bit to tell about the validity of another bit.
> 
> I'm saying is that a bitfield could be used to indicate that certain integer
> values (and not bits) are filled or not. Btw, MCE has it: depending on the
> MCE status registers, reading other registers may be needed to evaluate
> fully the error.

Bullshit, there'll be no bits showing the validity of other fields.

> > We're going to have single fields for EDAC-global valid values and leave
> > the driver-specific stuff lumped in one char * string.
> > 
> > Then, if a driver cannot specify sindrome or whatever, it simply DOESN'T
> > PUT IT IN THE STRING instead of adding validity bits.
> > 
> > [ … ]
> 
> Parsing strings that can be changed from Kernel versions and drivers is a 
> maintenance nightmare. Both I and Tony pointed it with different examples.

Bullshit, stop making up reasons to prove your point.

> >>> Which is basically what I'm suggesting for a couple of days now :-)
> >>>
> >>>        TP_PROTO(const unsigned int err_type,
> >>>                 const unsigned int mc_index,
> >>>                 const char *error_msg,
> >>>                 const char *label,
> >>>                 const char *location,
> >>>                 const char *detail)
> >>>
> >>> and I'm really not sure about err_type - this is an edac-specific define
> >>> and it means nothing outside the kernel so its string representation
> >>> could very well could be merged with error_msg and we can drop the ( ? :
> >>> ) ugliness in the tracepoint definition itself.
> >>
> >> All error management systems use a mandatory field for the error severity.
> > 
> > ... I'm pretty sure the defines you're exporting are not the same as
> > your error management systems require. So some kind of mapping is
> > still needed. Which shows that you're going to need to massage that
> > information in userspace after all.
> 
> Yes, field mapping is needed at the management systems, but this is not the
> point. It is not Kernel's task to help mapping fields inside userspace apps,
> but it is the task of a properly designed API to avoid userspace to guess about
> what the kernel is reporting.
> 
> A printk/sprintk-designed API, as you're proposing, requires userspace to guess
> what the events mean, with the help of a very careful designed parsing rules
> that needs to take into account every single EDAC driver (as the (s)printk message
> message will vary among them), and needs to be reviewed on every single new kernel
> version, as the (s)printk output can change among different Kernel versions. Even
> -stable versions might require changing the parsers, as a fixup patch might be
> changing the (s)printk arguments.

Enough with the crap already - now you're really desperately looking for
bogus arguments to prove your point.

We're going to have single fields for EDAC-global valid values and leave
the driver-specific stuff lumped in one char * string.


-- 
Regards/Gruss,
Boris.

Advanced Micro Devices GmbH
Einsteinring 24, 85609 Dornach
GM: Alberto Bozzo
Reg: Dornach, Landkreis Muenchen
HRB Nr. 43632 WEEE Registernr: 129 19551