allow console drivers to be called early
PATCH 1/1 allow console drivers to be called early
This is a simple patch to allow arches to set early_printk_ok if they've registered console drivers that support early operation. I've got an ia64 specific bit and an sn2 specific bit that I can post for reference if anyone's interested, but they're pretty straightforward, so I'm just posting this for comments. The ia64 one just adds a register_early_consoles() function to the ia64 code that gets called early on in setup_arch. All it does is call the init routines of console drivers that are setup to do early printks.
Jesse
===== kernel/printk.c 1.37 vs edited =====
kernel/printk.c +2 -1
@@ -54,6 +54,7 @@EXPORT_SYMBOL (console_printk);int oops_in_progress; +int early_printk_ok;/* * console_sem protects the console_drivers list, and also
@@ -526,7 +527,7 @@ log_level_unknown =1 ; } -if (!cpu_online (smp_processor_id ()) && +if (!early_printk_ok && !cpu_online (smp_processor_id ()) && system_state !=SYSTEM_RUNNING ) {/* * Some console drivers may assume that per-cpu resources have
===== include/linux/kernel.h 1.48 vs edited =====
include/linux/kernel.h +1 -0
@@ -107,6 +107,7 @@ }extern void bust_spinlocks (int yes); +extern int early_printk_ok;/* If set, console drivers will be called even if the system isn't up yet */ extern int oops_in_progress;/* If set, an oops, panic(), BUG() or die() is in progress */ extern int panic_on_oops;extern int system_state;/* See values below */
Quoted 10 lines · Jesse Barnes
This is a simple patch to allow arches to set
early_printk_okif they've
registered console drivers that support early operation. I've got an ia64 specific bit and an sn2 specific bit that I can post for reference if anyone's interested, but they're pretty straightforward, so I'm just posting this for comments. The ia64 one just adds aregister_early_consoles()function to the ia64 code that gets called early on insetup_arch. All it does is call the init routines of console drivers that are setup to do early printks.
Quoted 8 lines
...
*console_semprotects theconsole_driverslist, and also
@@ -526,7 +527,7 @@
log_level_unknown= 1;
}- if (!cpu_online(smp_processor_id()) &&
+ if (!early_printk_ok && !cpu_online(smp_processor_id()) &&
Is it not possible to mark this cpu as being online? It sure seems to be.
Quoted 4 lines · Andrew Morton
Quoted 2 lines
- if (!cpu_online(smp_processor_id()) &&
+ if (!early_printk_ok && !cpu_online(smp_processor_id()) &&Is it not possible to mark this cpu as being online? It sure seems to be.
But then we miss out on all the good stuff happening in setup_arch since smp_processor_id() isn't set in cpu_online_map until smp_prepare_boot_cpu. I can try upping it in arch code, but iirc, I had problems with that in older kernels due to the expectations of some of the other CPU upping code. I'll give it a try again and either give you a good answer or send a more appropriate patch to David.
Thanks,
Jesse