bug_frame related sections exclusively contain addresses that reference back to
the address where the BUG_INSTR is executed. As such, any change to the
contents of bug_frame sections (or it's relocations) will necessarily require a
change in the caller function, as the placement of the BUG_INSTR must have
changed.
Take advantage of this relocation, and unconditionally mark the bug_frame
sections themselves as not having changed, the logic in
kpatch_regenerate_special_section() will already take care of including any
bug_frame element group that references a function that has changed.
This should be a non functional change in the payload generated by
create-diff-object, but needs doing so that we can take into account changes to
.altinstructions and .ex_table sections themselves without unnecessarily also
pulling .bug_frame sections.
Signed-off-by: Roger Pau Monné <roger....@citrix.com>
---
create-diff-object.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/create-diff-object.c b/create-diff-object.c
index f4e4da063d0a..d1b1477be1cd 100644
--- a/create-diff-object.c
+++ b/create-diff-object.c
@@ -1284,6 +1284,17 @@ static void kpatch_regenerate_special_section(struct
kpatch_elf *kelf,
}
}
+ /*
+ * For bug_frame sections don't care if the section itself or the
+ * relocations have changed, as any change in the bug_frames will be
+ * accompanied by a change in the caller function text, as the
+ * BUG_INSTR will have a different placement in the caller.
+ */
+ if (!strncmp(special->name, ".bug_frames.", strlen(".bug_frames."))) {
+ sec->status = SAME;
+ sec->base->status = SAME;
+ }
+
for ( ; src_offset < sec->base->sh.sh_size; src_offset += group_size) {
group_size = special->group_size(kelf, src_offset);
--
2.44.0
