Currently all the callers of kpatch_write_output_elf() are creating
.o object files or .ko kernel modules. Neither of these filetypes are
executable on their own, so enhance kpatch_write_output_elf() to accept
file creation mode and update its callers to pass 0664 to match
the expected permissions.
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Or, to be exact, with addend values which cannot be represented by
a signed int variable.
This only applies to the old KPatch core.
Commit 15067fcd "kmod/core: apply dynrela addend for R_X86_64_64" fixed
calculation of the values for R_X86_64_64 dynrelas. This revealed
another issue, similar to https://github.com/dynup/kpatch/issues/1064.
Dynrelas are stored as 'struct kpatch_patch_dynrela' instances in the
patch module but both the patch module and kpatch.ko use
'struct kpatch_dynrela' to work with the dynrelas. 'addend' has type
'long' in kpatch_patch_dynrela but 'int' in kpatch_dynrela, so this
value can be truncated when read.
R_X86_64_64 dynrela can be created, for example, if a patch for vmlinux
refers to something like '(unsigned long)&idt_table+0x80000000' (a global
variable which is not exported, with some addend).
The addend == +0x80000000, however, effectively becomes 0xffffffff80000000
(== -0x80000000) due to this bug.
Unfortunately, 'struct kpatch_dynrela' is a part of the ABI between
kpatch.ko and patch modules. Plain changing 'int addend' into 'long addend'
there could be problematic. The patch module built using the new
'struct kpatch_dynrela' will either fail to load if kpatch.ko is using the old
'struct kpatch_dynrela' or cause crashes or data corruptions. Unloading
and reloading patch modules and kpatch.ko is not always an option
either.
Luckily, R_X86_64_64 dynrelas seem to be quite rare in the production
patch modules and R_X86_64_64 dynrelas with large addends are expected
to be even more rare.
So, instead of fixing the truncation of addends right away, I propose to
detect it, for now, when building a patch. If one never hits such conditions,
it is not worth it to fix the issue. If R_X86_64_64 dynrelas with large
addends do happen and cannot be avoided, we can try to figure out how to
fix this properly, without breaking too much.
Signed-off-by: Evgenii Shatokhin <eshatokhin@virtuozzo.com>
Make sure symtab section was found before dereferencing it.
Found by covscan, see issue #984 for full log.
Signed-off-by: Artem Savkov <asavkov@redhat.com>
- convert section/symbol indexes and rela->offset to unsigned int as I
couldn't find any way for them to become negative.
- cast a number of rela->addend comparisons to int (assuming an 64bit
system this should be enough)
- a number of simple for-loop counter conversions to the type it
compares against
Signed-off-by: Artem Savkov <asavkov@redhat.com>
The create_dynamic_rela_sections() function assumes that all dest
symbols are bundled, i.e. each symbol is located at offset 0 in its own
section.
However that may not always be the case. Unbundled symbols can occur,
for example, when combining two .o files which have the same bundled
symbol. They will be combined into the same section and will no longer
be considered "bundled".
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Add new program create-kpatch-module, that, given an intermediate object
outputted by create-diff-object, outputs an object (.o) that contains the
.kpatch.dynrelas section required by kpatch.
Joe Lawrence
Evgenii Shatokhin
Josh Poimboeuf
Kamalesh Babulal
Artem Savkov
Jessica Yu