| 40,6 → 40,10 |
|---|
| #include <debug.h> |
| #include <symtab.h> |
| |
| #ifdef CONFIG_FB |
| #include <genarch/fb/fb.h> |
| #endif |
| |
| #ifdef CONFIG_SMP |
| |
| /** Initialize spinlock |
| 73,7 → 77,31 |
|---|
| |
| preemption_disable(); |
| while (test_and_set(&sl->val)) { |
| if (i++ > 300000 && sl!=&printflock) { |
| |
| /* |
| * We need to be careful about printflock and fb_lock. |
| * Both of them are used to report deadlocks via |
| * printf() and fb_putchar(). |
| * |
| * We trust our code that there is no possible deadlock |
| * caused by these two locks (except when an exception |
| * is triggered for instance by printf() or fb_putchar()). |
| * However, we encountered false positives caused by very |
| * slow VESA framebuffer interaction (especially when |
| * run in a simulator) that caused problems with both |
| * printflock and fb_lock. |
| * |
| * Possible deadlocks on both printflock and fb_lock |
| * are therefore not reported as they would cause an |
| * infinite recursion. |
| */ |
| if (sl == &printflock) |
| continue; |
| #ifdef CONFIG_FB |
| if (sl == &fb_lock) |
| continue; |
| #endif |
| if (i++ > 300000) { |
| printf("cpu%d: looping on spinlock %.*p:%s, caller=%.*p", |
| CPU->id, sizeof(__address) * 2, sl, sl->name, sizeof(__address) * 2, CALLER); |
| symbol = get_symtab_entry(CALLER); |