Re: [PATCH] ppc/fadump: collect dump if the collected size is lesser than reserved

On 29/07/26 13:59, Shivang Upadhyay wrote:

On Wed, 2026-07-29 at 13:43 +0530, Sourabh Jain wrote:

quoted

quoted

quoted

But the firmware not copying the entire Real Mode memory is also
a
problem for
the kernel, isn't it?

I think it should be fine. Firmware will report it status flags,
right?

I don't know. Does PAPR says anything about the error code
corresponds
to when memory to not copied completely?

As much as i could search, I didn't see this case metioned there. I
think if, fw is failing to copy full content it should report one of
the below errors.

- copy from [x...x+size] to [y...y+size].
- x + x_i is not valid
   - leads to invalid source address.
- y + y+i is not valid
   - leads to invalid dest address.


But anyways, As the source/proper description is not available, we
can't prove much here. Would it be appropriate, if I just make this
case for CPU_DATA area?

It would be good to first understand how the firmware (PHYP) handles
CPU register data when maxcpus differs from the number of online CPUs.

According to PAPR, the firmware should only include data for online CPUs.
However, when min=1, desired=1, and max=2 CPUs, we observed that the
firmware reported NumCpus as 16, which is the system's maximum CPU count.
This makes it unclear how the firmware handles this scenario.

- Sourabh Jain