#615 Update -n option to dump all note sections

[Mykola Hohsadze]

I've updated the -n option to dump the note sections and attached
dump_note_sections.patch to review

According to documentation there might be cases whendesc or name
may be missing.
So I added checks to handle cases when descz and namesz are null,
we will print 0 in normal format, without this condition we will get a
magic value.
When namesz is zero, we will print 0 for normal format and nothing
in case of Solaris format.

Also, I added type to normal format similar to freebsd version but
just number because it is consistent with the Solaris format and type
is OS specific and string of type will be different

Also updated exec1, exec2 expected output for note sections and added exec3 directory with testing -n option.

-n output example that I used in exec3:

        // No descriptor
        .section .note.freebsd.ident
        .long   0x8             // namesz
        .long   0x0             // descsz
        .long   0x1             // type
        .string "FreeBSD"       // name

        // Descriptor and name present
        .section .note.freebsd.march
        .long   0x8             // namesz
        .long   0x6             // descsz
        .long   0x2             // type
        .string "FreeBSD"       // name
        .string "arm64"         // desc
        .short  0x0             // padding

        // No name
        .section .note.my_section
        .long   0x0             // namesz
        .long   0x4             // descsz
        .long   0x5             // type
        .long   0x10            // desc

        // Multiple entries
        .section .note.multiple_entries
entry1:
        .long   0xA             // namesz
        .long   0x4             // descsz
        .long   0x3             // type
        .string "MySection"     // name
        .short  0x0             // padding
        .byte   0x7             // desc[0] (LE)
        .byte   0x0             // desc[1] (LE)
        .byte   0x0             // desc[2] (LE)
        .byte   0x0             // desc[3] (LE)
entry2:
        .long   0xB             // namesz
        .long   0x4             // descsz
        .long   0x4             // type
        .string "MySection1"    // name
        .byte   0x0             // padding
        .long   0x30            // desc

        .text
        .global main
main:
        mov     x0, #0
        ret
        .size   main, (. - main)

Normal format:

# ./elfdump -n note_sections

note (.note.tag):
        FreeBSD 1400073 (1)
        FreeBSD 0 (4)
        FreeBSD 0 (2)

note (.note.freebsd.ident):
        FreeBSD 0 (1)

note (.note.freebsd.march):
        FreeBSD 913142369 (2)

note (.note.my_section):
        0 16 (5)

note (.note.multiple_entries):
        MySection 7 (3)
        MySection1 48 (4)

Solaris format:

# ./elfdump -Sn note_sections

Note Section:  .note.tag

    type   0x1
    namesz 0x8:
FreeBSD
    descsz 0x4:
      desc[0]   09 5d 15 00

    type   0x4
    namesz 0x8:
FreeBSD
    descsz 0x4:
      desc[0]   00 00 00 00

    type   0x2
    namesz 0x8:
FreeBSD
    descsz 0x4:
      desc[0]   00 00 00 00

Note Section:  .note.freebsd.ident

    type   0x1
    namesz 0x8:
FreeBSD
    descsz 0:

Note Section:  .note.freebsd.march

    type   0x2
    namesz 0x8:
FreeBSD
    descsz 0x6:
      desc[0]   61 72 6d 36   34 00

Note Section:  .note.my_section

    type   0x5
    namesz 0:
    descsz 0x4:
      desc[0]   10 00 00 00

Note Section:  .note.multiple_entries

    type   0x3
    namesz 0xa:
MySection
    descsz 0x4:
      desc[0]   07 00 00 00

    type   0x4
    namesz 0xb:
MySection1
    descsz 0x4:
      desc[0]   30 00 00 00

ref:
https://www.sco.com/developers/gabi/latest/ch5.pheader.html#note_section

[Joseph Koshy]

Thank you for the patch.

The changes that iterate over all the note sections in the object look fine .

However, the patch also changes the output of the tool - are these changes compatible with Solaris and FreeBSD elfdump?

[Joseph Koshy]

The test data file test/elfdump/ts/exec3/note_sections file seems to be missing from the patch.

Could you please upload it to this ticket?

...
Index: test/elfdump/ts/exec3/note_sections
===================================================================
Cannot display: file marked as a binary type.
svn:mime-type = application/x-object

Property changes on: test/elfdump/ts/exec3/note_sections
___________________________________________________________________
Added: svn:executable
## -0,0 +1 ##
+*
\ No newline at end of property
Added: svn:mime-type
## -0,0 +1 ##
+application/x-object
\ No newline at end of property
...

[Mykola Hohsadze]

Sure

[Mykola Hohsadze]

looks like I've attached an old version of the binary, so I'll reattach the binary note_sections and checked output in OpenIndiana OS, see note_sections_openIndiana.txt

[Joseph Koshy]

Incidentally:

• https://docs.oracle.com/cd/E23824_01/html/819-0690/chapter6-18048.html#scrolltoc seems to specify 4 byte fields and alignment for Elf notes, irrespective of the machine word width,
• https://www.sco.com/developers/gabi/latest/ch5.pheader.html#note_section specifies that the size and alignment of note data is dependent on the machine architecture,
• https://www.man7.org/linux/man-pages/man5/elf.5.html seems to specify architecture-dependent size for the note header, but with zero padding before the description field.