Under some conditions, irq sorting procedure used by INTC can go wrong resulting in a spurious irq getting reported. If this condition is not handled, it results in endless stream of: unexpected IRQ trap at vector 00 messages from ack_bad_irq() Handle the spurious interrupt condition in omap-intc driver to prevent this. Measurements using kernel function profiler on AM335x EVM running at 720MHz show that after this patch omap_intc_handle_irq() takes about 37.4us against 34us before this patch.
Unified · Split
| @@ -47,6 +47,7 @@ · 46 unmodified lines | ||
| 47 | 47 | #define INTC_ILR0 0x0100 |
| 48 | 48 | |
| 49 | 49 | #define ACTIVEIRQ_MASK 0x7f /* omap2/3 active interrupt bits */ |
| 50 | +#define SPURIOUSIRQ_MASK (0x1ffffff << 7) | |
| 50 | 51 | #define INTCPS_NR_ILR_REGS 128 |
| 51 | 52 | #define INTCPS_NR_MIR_REGS 4 |
| 52 | 53 | |
| @@ -329,11 +330,35 @@ static int __init omap_init_irq(u32 base, struct device_node *node) · 276 unmodified lines | ||
| 329 | 330 | static asmlinkage void __exception_irq_entry |
| 330 | 331 | omap_intc_handle_irq(struct pt_regs *regs) |
| 331 | 332 | { |
| 333 | + extern unsigned long irq_err_count; | |
| 332 | 334 | u32 irqnr; |
| 333 | 335 | |
| 334 | 336 | irqnr = intc_readl(INTC_SIR); |
| 337 | + | |
| 338 | + /* | |
| 339 | + * A spurious IRQ can result if interrupt that triggered the | |
| 340 | + * sorting is no longer active during the sorting (10 INTC | |
| 341 | + * functional clock cycles after interrupt assertion). Or a | |
| 342 | + * change in interrupt mask affected the result during sorting | |
| 343 | + * time. There is no special handling required except ignoring | |
| 344 | + * the SIR register value just read and retrying. | |
| 345 | + * See section 6.2.5 of AM335x TRM Literature Number: SPRUH73K | |
| 346 | + * | |
| 347 | + * Many a times, a spurious interrupt situation has been fixed | |
| 348 | + * by adding a flush for the posted write acking the IRQ in | |
| 349 | + * the device driver. Typically, this is going be the device | |
| 350 | + * driver whose interrupt was handled just before the spurious | |
| 351 | + * IRQ occurred. Pay attention to those device drivers if you | |
| 352 | + * run into hitting the spurious IRQ condition below. | |
| 353 | + */ | |
| 354 | + if (unlikely((irqnr & SPURIOUSIRQ_MASK) == SPURIOUSIRQ_MASK)) { | |
| 355 | + pr_err_once("%s: spurious irq!\n", __func__); | |
| 356 | + irq_err_count++; | |
| 357 | + omap_ack_irq(NULL); | |
| 358 | + return; | |
| 359 | + } | |
| 360 | + | |
| 335 | 361 | irqnr &= ACTIVEIRQ_MASK; |
| 336 | − WARN_ONCE(!irqnr, "Spurious IRQ ?\n"); | |
| 337 | 362 | handle_domain_irq(domain, irqnr, regs); |
| 338 | 363 | } |
| 339 | 364 | |