module: set ksymtab/kcrctab* section addresses to 0x0
PATCH 1/1 module: set ksymtab/kcrctab* section addresses to 0x0
These __ksymtab*/__kcrctab* sections currently have non-zero addresses. Non-zero section addresses in a relocatable ELF confuse GDB and it ends up not relocating all symbols when add-symbol-file is used on modules which have exports. The kernel's module loader does not care about these addresses, so let's just set them to zero.
Before:
$ readelf -S lib/notifier-error-inject.ko | grep 'Name\| __ksymtab_gpl'
[Nr] Name Type Addr Off Size ES Flg Lk Inf Al
[ 8] __ksymtab_gpl PROGBITS 0000000c 0001b4 000010 00 A 0 0 4 (gdb) add-symbol-file lib/notifier-error-inject.ko 0x500000 -s .bss 0x700000 add symbol table from file "lib/notifier-error-inject.ko" at
.text_addr = 0x500000
.bss_addr = 0x700000 (gdb) p ¬ifier_err_inject_dir
$3 = (struct dentry **) 0x0
After:
$ readelf -S lib/notifier-error-inject.ko | grep 'Name\| __ksymtab_gpl'
[Nr] Name Type Addr Off Size ES Flg Lk Inf Al
[ 8] __ksymtab_gpl PROGBITS 00000000 0001b4 000010 00 A 0 0 4 (gdb) add-symbol-file lib/notifier-error-inject.ko 0x500000 -s .bss 0x700000 add symbol table from file "lib/notifier-error-inject.ko" at
.text_addr = 0x500000
.bss_addr = 0x700000 (gdb) p ¬ifier_err_inject_dir
$3 = (struct dentry **) 0x700000
--- scripts/module-common.lds | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)
scripts/module-common.lds +10 -10
@@ -6,14 +6,14 @@ SECTIONS { /DISCARD/ : { *(.discard) } - __ksymtab : { *(SORT(___ksymtab+*)) } - __ksymtab_gpl : { *(SORT(___ksymtab_gpl+*)) } - __ksymtab_unused : { *(SORT(___ksymtab_unused+*)) } - __ksymtab_unused_gpl : { *(SORT(___ksymtab_unused_gpl+*)) } - __ksymtab_gpl_future : { *(SORT(___ksymtab_gpl_future+*)) } - __kcrctab : { *(SORT(___kcrctab+*)) } - __kcrctab_gpl : { *(SORT(___kcrctab_gpl+*)) } - __kcrctab_unused : { *(SORT(___kcrctab_unused+*)) } - __kcrctab_unused_gpl : { *(SORT(___kcrctab_unused_gpl+*)) } - __kcrctab_gpl_future : { *(SORT(___kcrctab_gpl_future+*)) } + __ksymtab 0 : { *(SORT(___ksymtab+*)) } + __ksymtab_gpl 0 : { *(SORT(___ksymtab_gpl+*)) } + __ksymtab_unused 0 : { *(SORT(___ksymtab_unused+*)) } + __ksymtab_unused_gpl 0 : { *(SORT(___ksymtab_unused_gpl+*)) } + __ksymtab_gpl_future 0 : { *(SORT(___ksymtab_gpl_future+*)) } + __kcrctab 0 : { *(SORT(___kcrctab+*)) } + __kcrctab_gpl 0 : { *(SORT(___kcrctab_gpl+*)) } + __kcrctab_unused 0 : { *(SORT(___kcrctab_unused+*)) } + __kcrctab_unused_gpl 0 : { *(SORT(___kcrctab_unused_gpl+*)) } + __kcrctab_gpl_future 0 : { *(SORT(___kcrctab_gpl_future+*)) } }
Signature · 1 line
1.7.10.4
Quoted 5 lines · Rabin Vincent
These __ksymtab*/__kcrctab* sections currently have non-zero addresses. Non-zero section addresses in a relocatable ELF confuse GDB and it ends up not relocating all symbols when add-symbol-file is used on modules which have exports. The kernel's module loader does not care about these addresses, so let's just set them to zero.
Hi Rabin,
You're right, we don't use sh_addr. This should be safe, I've
applied it for the next merge window.
AFAICT this would have always been a problem; wonder if it should be CC:stable? I don't use gdb on modules so haven't got a good feel for how many people this would be hurting.
Thanks,
Rusty.
Quoted 68 lines
Before:
$ readelf -S lib/notifier-error-inject.ko | grep 'Name\| __ksymtab_gpl' [Nr] Name Type Addr Off Size ES Flg Lk Inf Al [ 8] __ksymtab_gpl PROGBITS 0000000c 0001b4 000010 00 A 0 0 4(gdb) add-symbol-file
lib/notifier-error-inject.ko0x500000-s .bss0x700000add symbol table from file "lib/notifier-error-inject.ko" at.text_addr = 0x500000 .bss_addr = 0x700000(gdb) p
¬ifier_err_inject_dir
$3 = (struct dentry **)0x0After:
$ readelf -S lib/notifier-error-inject.ko | grep 'Name\| __ksymtab_gpl' [Nr] Name Type Addr Off Size ES Flg Lk Inf Al [ 8] __ksymtab_gpl PROGBITS 00000000 0001b4 000010 00 A 0 0 4(gdb) add-symbol-file
lib/notifier-error-inject.ko0x500000-s .bss0x700000add symbol table from file "lib/notifier-error-inject.ko" at.text_addr = 0x500000 .bss_addr = 0x700000(gdb) p
¬ifier_err_inject_dir
$3 = (struct dentry **)0x700000Signed-off-by Rabin Vincent <[email protected]>--- scripts/module-common.lds | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)scripts/module-common.lds +10
-10@@ -6,14 +6,14 @@ SECTIONS { /DISCARD/ : { *(.discard) } - __ksymtab : { *(SORT(___ksymtab+*)) } - __ksymtab_gpl : { *(SORT(___ksymtab_gpl+*)) } - __ksymtab_unused : { *(SORT(___ksymtab_unused+*)) } - __ksymtab_unused_gpl : { *(SORT(___ksymtab_unused_gpl+*)) } - __ksymtab_gpl_future : { *(SORT(___ksymtab_gpl_future+*)) } - __kcrctab : { *(SORT(___kcrctab+*)) } - __kcrctab_gpl : { *(SORT(___kcrctab_gpl+*)) } - __kcrctab_unused : { *(SORT(___kcrctab_unused+*)) } - __kcrctab_unused_gpl : { *(SORT(___kcrctab_unused_gpl+*)) } - __kcrctab_gpl_future : { *(SORT(___kcrctab_gpl_future+*)) } + __ksymtab 0 : { *(SORT(___ksymtab+*)) } + __ksymtab_gpl 0 : { *(SORT(___ksymtab_gpl+*)) } + __ksymtab_unused 0 : { *(SORT(___ksymtab_unused+*)) } + __ksymtab_unused_gpl 0 : { *(SORT(___ksymtab_unused_gpl+*)) } + __ksymtab_gpl_future 0 : { *(SORT(___ksymtab_gpl_future+*)) } + __kcrctab 0 : { *(SORT(___kcrctab+*)) } + __kcrctab_gpl 0 : { *(SORT(___kcrctab_gpl+*)) } + __kcrctab_unused 0 : { *(SORT(___kcrctab_unused+*)) } + __kcrctab_unused_gpl 0 : { *(SORT(___kcrctab_unused_gpl+*)) } + __kcrctab_gpl_future 0 : { *(SORT(___kcrctab_gpl_future+*)) } }Signature · 7 lines
1.7.10.4 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Quoted 9 lines · Rusty Russell
Quoted 5 lines · Rabin Vincent
These __ksymtab*/__kcrctab* sections currently have non-zero addresses. Non-zero section addresses in a relocatable ELF confuse GDB and it ends up not relocating all symbols when add-symbol-file is used on modules which have exports. The kernel's module loader does not care about these addresses, so let's just set them to zero.
You're right, we don't use
sh_addr. This should be safe, I've
applied it for the next merge window.
Thanks.
Quoted 3 lines
AFAICT this would have always been a problem; wonder if it should be CC:stable? I don't use gdb on modules so haven't got a good feel for how many people this would be hurting.
I've no opinion on -stable. This method of debugging kernel modules is described in for example LDD3, so it would be nice if it worked in all cases, but I guess most people won't be affected even if they use it since it only affects modules with exported symbols.