Re: fcore status

[Phil Muldoon <pmuldoon at redhat dot com>]

I fixed the prspinfo, in the included patch. As usual, small bugs take 
the longest to find.  ;) The problem here was not asking libelf to 
calculate the note header offset in memory, then re-fetching the section 
header, and only then updating the program header offset for notes. Just 
updating the structure back to the native bits is not enough; we have to 
ask libelf to so some shuffling around, then tell use where everything is.

I'll check this in, with a number of other fixes/changes later today.

Regards

Phil

Phil Muldoon wrote:
> > But at the end of output from readelf, there are some warning infor:
> >
> > Notes at offset 0x00000000 with length 0x00000098:
> >   Owner         Data size       Description
> > readelf: Warning: corrupt note found at offset 0 into core notes
> > readelf: Warning:  type: 0, namesize: 464c457f, descsize: 00010102
> >
> > The above "type", namesize and descsize are not equal to that I record
> > in prpsinfo struct! Although there are some warning info, I checked them
> > in in order that we both can start working on it. right? :-)
> >   
>
> Right, I'll pick up where you left off and l'll run with it today. 
> Looking at it initially, for some reason there is a small offset of 
> gcore/kernel where the notes lives (it does not begin at 0, like 
> ours). I'm wondering if there is some confusion on ElfEHeader and 
> NOTES header. We might have our offsets incorrectly calculated here.
>
> For example in others:
>
> Section Headers:
> [Nr] Name                 Type         Addr     Off    Size   ES Flags 
> Lk Inf Al
> [ 0]                      NULL         00000000 000000 000000  
> 0        0   0  0
> [ 1] note0                PROGBITS     00000000 000394 00046c  0 
> A      0   0  1
>
> > I will trace down the bug. If I could get some comments from you about
> > this, it will be appreciated very much.
> >
> > 2006-10-19  Yong Zheng  <zhengyo@cn.ibm.com>
> >
> >     * frysk-core/frysk/util/FCore.java (fillENoteSection): New
> >     function to construct PT_NOTE section).
> >     (constructSectionData): New function to construct PT_NOTE data.
> >     (fillENotePrpsinfo): New function to fill elf_prpsinfo info.
> >     * frysk-imports/lib/elf/ElfNhdr.java: New file.
> >     * frysk-imports/lib/elf/ELfNhdrType.h: Ditto.
> >     * frysk-imports/lib/elf/ELfPrpsinfo.java: Ditto.
> >     * frysk-imports/lib/elf/cni/ELfNhdr.cxx: Ditto.
> >     * frysk-imports/lib/elf/cni/ELfPrpsinfo.cxx: Ditto.
> >
> >   
>
> All looks great to me. I suspect we'll need above check-in for each 
> prspinfo, prstatus, fpregset, and auxv information. Lots of structures 
> for one section ;) I'll let you know later on today what I come up with.
>
> Regards
>
> Phil

Index: frysk-core/frysk/util/FCore.java
===================================================================
RCS file: /cvs/frysk/frysk-core/frysk/util/FCore.java,v
retrieving revision 1.7
diff -u -r1.7 FCore.java
--- frysk-core/frysk/util/FCore.java	19 Oct 2006 10:45:54 -0000	1.7
+++ frysk-core/frysk/util/FCore.java	19 Oct 2006 17:30:48 -0000
@@ -458,6 +458,17 @@
 		noteSectHeader.size = noteSection.getData().getSize();
 		noteSection.update(noteSectHeader);
         
+		// Must first ask libelf to construct offset location before
+		// adding offset back to program header. Otherwise program offset
+		// will be 0.
+		if (local_elf.update(ElfCommand.ELF_C_NULL) < 0) {
+			throw new ElfException("Cannot calculate note section offset");
+		}
+		
+		// Then re-fetch the elf modified header from section. Now offset 
+		// is calculated and correct.
+		noteSectHeader = noteSection.getSectionHeader();
+		
 		// Modify PT_NOTE program header
 		noteProgramHeader = local_elf.getPHeader(0);
 		noteProgramHeader.type = ElfPHeader.PTYPE_NOTE;