2011/7/11 Seiji Aguchi <seiji.aguchi at hds.com>: > Hi, > Hi, > > [Patch Description] > > ?For meeting Eric/Vivek's requirements and solving issues of pstore/efivars driver, ?I propose a following patch. > > ? ?- Remove kmsg_dump(KMSG_DUMP_KEXEC) from crash_kexec() It is already removed in -mm, can you rebase your patch against -mm? > ? ?- Add kmsg_dump_kexec() to crash_kexec() > ? ? ? ?kmsg_dump_kexec() moved behind machine_crash_shutdown() so that > ? ? ? ?we can output kernel messages without getting any locks. > > ? ?- Introduce CONFIG_KMSG_DUMP_KEXEC > ? ? ? ?- CONFIG_KMSG_DUMP_KEXEC depends on CONFIG_PRINTK which is required for kmsg_dumper. > > ? ? ? ?- CONFIG_KMSG_DUMP_KEXEC=n (default) > ? ? ? ? ?Kernel message logging feature doesn't work in kdump path. > > ? ? ? ?- CONFIG_KMSG_DUMP_KEXEC=y > ? ? ? ? ?following static functions are called in kdump path. > ? ? ? ? ?- kmsg_dump_kexec(): This is based on kmsg_dump() and just removed spinlock and a function call chain from it. > ? ? ? ? ?- do_kmsg_dump_kexec(): This is based on pstore_dump() and just removed mutex lock from it. > > ? ? ? ?Some people who are not familiar with kexec may add function calls getting spin_lock/mutex_lock in kexec path. > ? ? ? ?These function calls causes failure of kexec. > ? ? ? ?So, I suggest replace a call chain with static function calls so that we can keep reliability of kexec. > > ? ? - Add efi_kmsg_dump_kexec() which outputs kernel messages into NVRAM with EFI. > ? ? ? This is called by do_kmsg_dump_kexec() statically. > ? ? ? By applying to Matthew Garret's patch below, its kernel messages in NVRAM are visible through /dev/pstore. This looks weird, kmsg_dump_kexec() calls do_kmsg_dump_kexec() which then calls efi_kmsg_dump_kexec(), so, why not just call efi_kmsg_dump_kexec()? since there is no other user of kmsg_dump_kexec().
