mbox series

[0/2] arch_dump: fix prstatus pid on s390x and ppc

Message ID cover.1718771802.git.osandov@osandov.com
Headers show
Series arch_dump: fix prstatus pid on s390x and ppc | expand

Message

Omar Sandoval June 19, 2024, 5 a.m. UTC 
Hello,

I maintain drgn [1], a debugger for the Linux kernel. I ran into a quirk
of the NT_PRSTATUS note in kernel core dumps [2], so I looked into how
QEMU's dump-guest-memory command generates NT_PRSTATUS. I noticed that
on most architectures, the note's PID field is set to the CPU ID plus 1.
There are two exceptions: on s390x, there's an endianness bug (it's not
byte swapped if the host is little endian), and on ppc, it's not set at
all (it defaults to zero). They're both easy fixes.

Thanks,
Omar

1: https://github.com/osandov/drgn
2: https://github.com/osandov/drgn/issues/404

Omar Sandoval (2):
  target/s390x/arch_dump: use correct byte order for pid
  target/ppc/arch_dump: set prstatus pid to cpuid

 target/ppc/arch_dump.c   | 21 ++++++++++++---------
 target/s390x/arch_dump.c |  2 +-
 2 files changed, 13 insertions(+), 10 deletions(-)

Comments

Thomas Huth June 24, 2024, 6:17 a.m. UTC | #1 
On 19/06/2024 07.00, Omar Sandoval wrote:
> Hello,
> 
> I maintain drgn [1], a debugger for the Linux kernel. I ran into a quirk
> of the NT_PRSTATUS note in kernel core dumps [2], so I looked into how
> QEMU's dump-guest-memory command generates NT_PRSTATUS. I noticed that
> on most architectures, the note's PID field is set to the CPU ID plus 1.
> There are two exceptions: on s390x, there's an endianness bug (it's not
> byte swapped if the host is little endian), and on ppc, it's not set at
> all (it defaults to zero). They're both easy fixes.
> 
> Thanks,
> Omar
> 
> 1: https://github.com/osandov/drgn
> 2: https://github.com/osandov/drgn/issues/404
> 
> Omar Sandoval (2):
>    target/s390x/arch_dump: use correct byte order for pid
>    target/ppc/arch_dump: set prstatus pid to cpuid
> 
>   target/ppc/arch_dump.c   | 21 ++++++++++++---------
>   target/s390x/arch_dump.c |  2 +-
>   2 files changed, 13 insertions(+), 10 deletions(-)

I'm going to pick up patch 1 for my s390x tree ... for the second patch, it 
would be good if someone of the ppc guys could have a look at it first.

  Thomas