On 11/12/2018 14:02, Bartosz Golaszewski wrote: > pt., 9 lis 2018 o 18:21 Bartosz Golaszewski <[email protected]> napisał(a): >> >> Two threads can try to fire the irq_sim with different offsets and will >> end up fighting for the irq_work asignment. Thomas Gleixner suggested a >> solution based on a bitfield where we set a bit for every offset >> associated with an interrupt that should be fired and then iterate over >> all set bits in the interrupt handler. >> >> This is a slightly modified solution using a bitmap so that we don't >> impose a limit on the number of interrupts one can allocate with >> irq_sim. >> >> Suggested-by: Thomas Gleixner <[email protected]> >> Signed-off-by: Bartosz Golaszewski <[email protected]> >> --- >> include/linux/irq_sim.h | 2 +- >> kernel/irq/irq_sim.c | 23 +++++++++++++++++++++-- >> 2 files changed, 22 insertions(+), 3 deletions(-) >> >> diff --git a/include/linux/irq_sim.h b/include/linux/irq_sim.h >> index 630a57e55db6..4500d453a63e 100644 >> --- a/include/linux/irq_sim.h >> +++ b/include/linux/irq_sim.h >> @@ -16,7 +16,7 @@ >> >> struct irq_sim_work_ctx { >> struct irq_work work; >> - int irq; >> + unsigned long *pending; >> }; >> >> struct irq_sim_irq_ctx { >> diff --git a/kernel/irq/irq_sim.c b/kernel/irq/irq_sim.c >> index dd20d0d528d4..98a20e1594ce 100644 >> --- a/kernel/irq/irq_sim.c >> +++ b/kernel/irq/irq_sim.c >> @@ -34,9 +34,20 @@ static struct irq_chip irq_sim_irqchip = { >> static void irq_sim_handle_irq(struct irq_work *work) >> { >> struct irq_sim_work_ctx *work_ctx; >> + unsigned int offset = 0; >> + struct irq_sim *sim; >> + int irqnum; >> >> work_ctx = container_of(work, struct irq_sim_work_ctx, work); >> - handle_simple_irq(irq_to_desc(work_ctx->irq)); >> + sim = container_of(work_ctx, struct irq_sim, work_ctx); >> + >> + while (!bitmap_empty(work_ctx->pending, sim->irq_count)) { >> + offset = find_next_bit(work_ctx->pending, >> + sim->irq_count, offset); >> + clear_bit(offset, work_ctx->pending); >> + irqnum = irq_sim_irqnum(sim, offset); >> + handle_simple_irq(irq_to_desc(irqnum)); >> + } >> } >> >> /** >> @@ -63,6 +74,13 @@ int irq_sim_init(struct irq_sim *sim, unsigned int >> num_irqs) >> return sim->irq_base; >> } >> >> + sim->work_ctx.pending = bitmap_zalloc(num_irqs, GFP_KERNEL); >> + if (!sim->work_ctx.pending) { >> + kfree(sim->irqs); >> + irq_free_descs(sim->irq_base, num_irqs); >> + return -ENOMEM; >> + } >> + >> for (i = 0; i < num_irqs; i++) { >> sim->irqs[i].irqnum = sim->irq_base + i; >> sim->irqs[i].enabled = false; >> @@ -89,6 +107,7 @@ EXPORT_SYMBOL_GPL(irq_sim_init); >> void irq_sim_fini(struct irq_sim *sim) >> { >> irq_work_sync(&sim->work_ctx.work); >> + bitmap_free(sim->work_ctx.pending); >> irq_free_descs(sim->irq_base, sim->irq_count); >> kfree(sim->irqs); >> } >> @@ -143,7 +162,7 @@ EXPORT_SYMBOL_GPL(devm_irq_sim_init); >> void irq_sim_fire(struct irq_sim *sim, unsigned int offset) >> { >> if (sim->irqs[offset].enabled) { >> - sim->work_ctx.irq = irq_sim_irqnum(sim, offset); >> + set_bit(offset, sim->work_ctx.pending); >> irq_work_queue(&sim->work_ctx.work); >> } >> } >> -- >> 2.19.1 >> > > Ping and Cc'ing Marc Zyngier. Any chance of getting this in for 4.21?
Bizarrely, I can't find any trace of this patch in my Inbox. I've now cherry-picked from https://lore.kernel.org/patchwork/patch/1009297/ . Thanks, M. -- Jazz is not dead. It just smells funny...
