Re: [PATCH] Linux Kernel Markers 0.11 for 2.6.17

[Mathieu Desnoyers <compudj at krystal dot dyndns dot org>]

* Jeremy Fitzhardinge (jeremy@goop.org) wrote:
> Mathieu Desnoyers wrote:
> >I could declare my jump_select_label directly in assembly then.
> >  
> 
> Maybe, but it could be tricky to make that label visible to C code.
> 
> >>>+call_label: \
> >>>+		asm volatile ("" : : ); \
> >>>+		MARK_CALL(name, format, ## args); \
> >>>+		asm volatile ("" : : ); \
> >>>+over_label: \
> >>>+		asm volatile ("" : : ); \
> >>> 
> >>>      
> >>These asm volatiles won't do anything at all. What are you trying to 
> >>achieve?
> >>    
> >
> >I want to make sure that the call_label's address will be exactly after 
> >the 2nd
> >byte of the jump instruction. The over_label does not really matter, as 
> >long as
> >it points to a correct spot in the execution flow. The most important is 
> >that
> >it stays near the jump instruction.
> >  
> 
> The "volatile" modifier for "asm" *only* means that the asm emitted if 
> the code is reachable at all; it doesn't make any constraints about 
> relative ordering of the various asm volatile statement with respect to 
> each other, or with respect to other code.
> 
> >I could probably do all this in assembly too.
> >  
> 
> Perhaps, though doing as much as possible visible to gcc has its 
> benefits.  Tricky either way.
> 

Would it be correct if we put dependencies on a label corresponding to the
previous asm in the read constraints for each asm ?

> >>>+#ifdef CONFIG_MARKERS
> >>>+#define MARK(name, format, args...) \
> >>>+	do { \
> >>>+		__label__ here; \
> >>>+here:   	asm volatile(	".section .markers, \"a\";\n\t" \
> >>>+				".long %0, %1;\n\t" \
> >>>+				".previous;\n\t" : : \
> >>>+			"m" (*(#name)), \
> >>>+			"m" (*&&here)); \
> >>> 
> >>>      
> >>Seems like a bad idea that MARK() can put one type of record in 
> >>.markers, but MARK_JUMP and MARK_CALL can put different records in the 
> >>same section? How do you distinguish them? Or are they certain to be 
> >>exclusive? Either way, I'd probably put different mark records in 
> >>different sections: .markers.jump, .markers.call, markers.labels. And 
> >>define appropriate structures for the record types in each section.
> >>
> >>    
> >
> >
> >struct __mark_marker {
> >        const char *name;
> >        const void *location;
> >        char *select;
> >        const void *jump_call;
> >        const void *jump_over;
> >        marker_probe_func **call;
> >        const char *format;
> >};
> >
> >is the structure which defines a complete record in the mark section. They 
> >are
> >all tied to the same marker site, so I think it makes sense to keep them 
> >in the
> >same record.
> >  
> 
> I don't understand.  Your asms put things into the marker section with 
> ".long A, B, C".  Does does that correspond to this structure?
> 

Yes, those are all pointers and a single MARK declares 7 of them. Please tell
me if I goofed up in assembly typing.

Regards,

Mathieu

OpenPGP public key:              http://krystal.dyndns.org:8080/key/compudj.gpg
Key fingerprint:     8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68