stack overflow safe kdump (2.6.18-rc1-i386) try#2

Hi,

I tried to incorporate all the ideas received after the previous post (thank you!). In particular I hope the new code is handling the Voyager case properly.

---

This is a the first of a series of patch-sets aiming at making kdump more robust against stack overflows.

This patch set does the following:

* Add safe_smp_processor_id function to i386 architecture (this function was inspired by the x86_64 function of the same name).

* Substitute "smp_processor_id" with the stack overflow-safe "safe_smp_processor_id" in the reboot path to the second kernel.

List of patches (the last two should be applied in the order of appearance):

[1/4] safe_smp_processor_id: stack overflow safe implementation of smp_processor_id.

[2/4] safe_smp_processor_id_voyager: stack overflow safe implementation of smp_processor_id for i386-voyager.

[3/4] crash_use_safe_smp_processor_id: replace smp_processor_id with safe_smp_processor_id in "arch/i386/kernel/crash.c".

[4/4] safe_smp_send_nmi_allbutself: re-implement smp_send_nmi_allbutself so that calls to smp_processor_id (through send_IPI_allbutself) can be replaced with safe_smp_processor_id without affecting other parts of the kernel (as suggested by Eric Biederman).

I am looking forward to your comments and suggestions.

Regards,

Fernando

Looks all reasonable to me.

-Andi

Looks good here.

Eric