<div dir="ltr">On 29 June 2016 at 04:48, Dave Anderson <<a href="mailto:anderson@redhat.com">anderson@redhat.com</a>> wrote: > > > On arm64, the link register (LR) holds a return address, which is the one > > just after a branch instruction. So using a saved lr as PC for backtracing > > might cause some confusion. > > For example, in kernel/entry.S, > > work_resched: > > ... > > bl schedule > > > > ret_to_user: > > ... > > > > The current code shows "ret_o_user", instead of "work_resched", > > as a caller of schedule(). > > I see... <div class="gmail_default" style="font-family:monospace,monospace">FYI, this can potentially be seen in the kernel's ftrace-based stack tracer or</div><div class="gmail_default" style="font-family:monospace,monospace">panic dump.</div><div class="gmail_default" style="font-family:monospace,monospace">In the past, PC calculation was modified to be "LR-4", but this patch</div><div class="gmail_default" style="font-family:monospace,monospace">was reverted. See</div><div class="gmail_default" style=""> commit 9702970c7bd3(Revert "ARM64: unwind: Fix PC calculation")</div><div> </div><div class="gmail_default" style="font-family:monospace,monospace">So I said that adding this patch was a discussion topic.</div> > > > > This patch corrects a PC by decrementing it by 4. > > But please note that this change may also make people a bit confused > > because a value of LR in the stack dump of "bt -f" doesn't match with > > an address in one-line summary. > > > > #2 [ffffcc7511407eb0] schedule at ffff0000d628aee0 > > ffffcc7511407eb0: ffffcc6d22f23080 ffff0000d5b44d6c <= LR > > ffffcc7511407ec0: ffffcc7511407ed0 0000000000000000 > > #3 [ffffcc7511407ed0] work_resched at ffff0000d5b44d68 <= correcrted PC > > That's probably OK -- and at least "bt -F" would clarify it somewhat. <div class="gmail_default" style="font-family:monospace,monospace">and "bt -Fs"</div><div> </div><div class="gmail_default" style="font-family:monospace,monospace">-Takahiro AKASHI</div> > Thanks, > Dave > > > > > > Signed-off-by: AKASHI Takahiro <<a href="mailto:takahiro.akashi@linaro.org">takahiro.akashi@linaro.org</a>> > > --- > > arm64.c | 23 ++++++++++++++--------- > > 1 file changed, 14 insertions(+), 9 deletions(-) > > > > diff --git a/arm64.c b/arm64.c > > index b74d2c4..bad36aa 100644 > > --- a/arm64.c > > +++ b/arm64.c > > @@ -1504,35 +1504,40 @@ static int > > arm64_print_stackframe_entry(struct bt_info *bt, int level, struct > > arm64_stackframe *frame, FILE *ofp) > > { > > char *name, *name_plus_offset; > > - ulong symbol_offset; > > + ulong pc, symbol_offset; > > struct syment *sp; > > struct load_module *lm; > > char buf[BUFSIZE]; > > > > - name = closest_symbol(frame->pc); > > + /* > > + * if pc comes from a saved lr, it actually points to an instruction > > + * after branch. To avoid any confusion, decrement pc by 4. > > + * See, for example, "bl schedule" before ret_to_user(). > > + */ > > + pc = frame->pc - 0x4; > > + name = closest_symbol(pc); > > name_plus_offset = NULL; > > > > if (bt->flags & BT_SYMBOL_OFFSET) { > > - sp = value_search(frame->pc, &symbol_offset); > > + sp = value_search(pc, &symbol_offset); > > if (sp && symbol_offset) > > - name_plus_offset = > > - value_to_symstr(frame->pc, buf, bt->radix); > > + name_plus_offset = value_to_symstr(pc, buf, bt->radix); > > } > > > > fprintf(ofp, "%s#%d [%8lx] %s at %lx", level < 10 ? " " : "", level, > > frame->fp ? frame->fp : bt->stacktop - USER_EFRAME_OFFSET, > > - name_plus_offset ? name_plus_offset : name, frame->pc); > > + name_plus_offset ? name_plus_offset : name, pc); > > > > if (BT_REFERENCE_CHECK(bt)) > > - arm64_do_bt_reference_check(bt, frame->pc, name); > > + arm64_do_bt_reference_check(bt, pc, name); > > > > - if (module_symbol(frame->pc, NULL, &lm, NULL, 0)) > > + if (module_symbol(pc, NULL, &lm, NULL, 0)) > > fprintf(ofp, " [%s]", lm->mod_name); > > > > fprintf(ofp, "\n"); > > > > if (bt->flags & BT_LINE_NUMBERS) { > > - get_line_number(frame->pc, buf, FALSE); > > + get_line_number(pc, buf, FALSE); > > if (strlen(buf)) > > fprintf(ofp, " %s\n", buf); > > } > > -- > > 2.9.0 > > > > -- > > Crash-utility mailing list > > <a href="mailto:Crash-utility@redhat.com">Crash-utility@redhat.com</a> > > <a href="https://www.redhat.com/mailman/listinfo/crash-utility">https://www.redhat.com/mailman/listinfo/crash-utility</a> > > > > -- > Crash-utility mailing list > <a href="mailto:Crash-utility@redhat.com">Crash-utility@redhat.com</a> > <a href="https://www.redhat.com/mailman/listinfo/crash-utility">https://www.redhat.com/mailman/listinfo/crash-utility</a></div>