scripts/dump-guest-memory.py | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
[Included a "Signed-off-by" line in this version.]
I recently investigated a bug in which the dump-guest-memory.py script
sees a gdb.MemoryError exception while attempting to dump memory
obtained from a QEMU core dump. (And, yes, dump-guest-core=on was
specified in the -machine option of the QEMU invocation.)
It turns out that memory region in question is not being placed in the
core dump and, after stepping through the kernel core dumping code
responsible for making this decision, it looks reasonable to me to not
include that region in the core dump. The region in question consists
of all zeros and, according to the kernel's logic, has never been
written to.
This commit makes a small change to the dump-guest-memory script to
cause inaccessible memory to be dumped as zeroes. This avoids the
exception and places the correct values in the guest memory dump.
Signed-off-by: Kevin Buettner <kevinb@redhat.com>
---
scripts/dump-guest-memory.py | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/scripts/dump-guest-memory.py b/scripts/dump-guest-memory.py
index 4177261d33..fbdfba458b 100644
--- a/scripts/dump-guest-memory.py
+++ b/scripts/dump-guest-memory.py
@@ -539,7 +539,12 @@ shape and this command should mostly work."""
while left > 0:
chunk_size = min(TARGET_PAGE_SIZE, left)
- chunk = qemu_core.read_memory(cur, chunk_size)
+ try:
+ chunk = qemu_core.read_memory(cur, chunk_size)
+ except gdb.MemoryError:
+ # Consider blocks of memory absent from a core file
+ # as being zeroed.
+ chunk = bytes(chunk_size)
vmcore.write(chunk)
cur += chunk_size
left -= chunk_size
--
2.24.1
Hi On Sat, Feb 15, 2020 at 1:34 AM Kevin Buettner <kevinb@redhat.com> wrote: > > [Included a "Signed-off-by" line in this version.] > > I recently investigated a bug in which the dump-guest-memory.py script > sees a gdb.MemoryError exception while attempting to dump memory > obtained from a QEMU core dump. (And, yes, dump-guest-core=on was > specified in the -machine option of the QEMU invocation.) > > It turns out that memory region in question is not being placed in the > core dump and, after stepping through the kernel core dumping code > responsible for making this decision, it looks reasonable to me to not > include that region in the core dump. The region in question consists > of all zeros and, according to the kernel's logic, has never been > written to. > > This commit makes a small change to the dump-guest-memory script to > cause inaccessible memory to be dumped as zeroes. This avoids the > exception and places the correct values in the guest memory dump. > > Signed-off-by: Kevin Buettner <kevinb@redhat.com> > --- > scripts/dump-guest-memory.py | 7 ++++++- > 1 file changed, 6 insertions(+), 1 deletion(-) > > diff --git a/scripts/dump-guest-memory.py b/scripts/dump-guest-memory.py > index 4177261d33..fbdfba458b 100644 > --- a/scripts/dump-guest-memory.py > +++ b/scripts/dump-guest-memory.py > @@ -539,7 +539,12 @@ shape and this command should mostly work.""" > > while left > 0: > chunk_size = min(TARGET_PAGE_SIZE, left) > - chunk = qemu_core.read_memory(cur, chunk_size) > + try: > + chunk = qemu_core.read_memory(cur, chunk_size) > + except gdb.MemoryError: > + # Consider blocks of memory absent from a core file > + # as being zeroed. > + chunk = bytes(chunk_size) That seems reasonable, but it will silently ignore any other memory error. Keith Seitz also looked at this bug, and he was wondering if BFD shouldn't treat the missing section differently: https://bugzilla.redhat.com/show_bug.cgi?id=1777751#c6 Keith, what do you think? thanks
Hi On Sat, Feb 15, 2020 at 1:34 AM Kevin Buettner <kevinb@redhat.com> wrote: > > [Included a "Signed-off-by" line in this version.] > > I recently investigated a bug in which the dump-guest-memory.py script > sees a gdb.MemoryError exception while attempting to dump memory > obtained from a QEMU core dump. (And, yes, dump-guest-core=on was > specified in the -machine option of the QEMU invocation.) > > It turns out that memory region in question is not being placed in the > core dump and, after stepping through the kernel core dumping code > responsible for making this decision, it looks reasonable to me to not > include that region in the core dump. The region in question consists > of all zeros and, according to the kernel's logic, has never been > written to. > > This commit makes a small change to the dump-guest-memory script to > cause inaccessible memory to be dumped as zeroes. This avoids the > exception and places the correct values in the guest memory dump. > > Signed-off-by: Kevin Buettner <kevinb@redhat.com> fwiw, Kevin fixed it in gdb: https://sourceware.org/ml/gdb-patches/2020-03/msg00106.html > --- > scripts/dump-guest-memory.py | 7 ++++++- > 1 file changed, 6 insertions(+), 1 deletion(-) > > diff --git a/scripts/dump-guest-memory.py b/scripts/dump-guest-memory.py > index 4177261d33..fbdfba458b 100644 > --- a/scripts/dump-guest-memory.py > +++ b/scripts/dump-guest-memory.py > @@ -539,7 +539,12 @@ shape and this command should mostly work.""" > > while left > 0: > chunk_size = min(TARGET_PAGE_SIZE, left) > - chunk = qemu_core.read_memory(cur, chunk_size) > + try: > + chunk = qemu_core.read_memory(cur, chunk_size) > + except gdb.MemoryError: > + # Consider blocks of memory absent from a core file > + # as being zeroed. > + chunk = bytes(chunk_size) > vmcore.write(chunk) > cur += chunk_size > left -= chunk_size > -- > 2.24.1 > > -- Marc-André Lureau
© 2016 - 2024 Red Hat, Inc.