diff options
| author | Rusty Russell <rusty@rustcorp.com.au> | 2009-06-12 22:27:02 -0600 |
|---|---|---|
| committer | Rusty Russell <rusty@rustcorp.com.au> | 2009-06-12 22:27:03 +0930 |
| commit | a32a8813d0173163ba44d8f9556e0d89fdc4fb46 (patch) | |
| tree | fddb6742338047d0219e8c2536cd39b04e643b16 /drivers/lguest/interrupts_and_traps.c | |
| parent | abd41f037e1a64543000ed73b42f616d04d92700 (diff) | |
lguest: improve interrupt handling, speed up stream networking
lguest never checked for pending interrupts when enabling interrupts, and
things still worked. However, it makes a significant difference to TCP
performance, so it's time we fixed it by introducing a pending_irq flag
and checking it on irq_restore and irq_enable.
These two routines are now too big to patch into the 8/10 bytes
patch space, so we drop that code.
Note: The high latency on interrupt delivery had a very curious
effect: once everything else was optimized, networking without GSO was
faster than networking with GSO, since more interrupts were sent and
hence a greater chance of one getting through to the Guest!
Note2: (Almost) Closing the same loophole for iret doesn't have any
measurable effect, so I'm leaving that patch for the moment.
Before:
1GB tcpblast Guest->Host: 30.7 seconds
1GB tcpblast Guest->Host (no GSO): 76.0 seconds
After:
1GB tcpblast Guest->Host: 6.8 seconds
1GB tcpblast Guest->Host (no GSO): 27.8 seconds
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Diffstat (limited to 'drivers/lguest/interrupts_and_traps.c')
| -rw-r--r-- | drivers/lguest/interrupts_and_traps.c | 16 |
1 files changed, 13 insertions, 3 deletions
diff --git a/drivers/lguest/interrupts_and_traps.c b/drivers/lguest/interrupts_and_traps.c index a8c966fee1e4..5a10754b4790 100644 --- a/drivers/lguest/interrupts_and_traps.c +++ b/drivers/lguest/interrupts_and_traps.c @@ -131,7 +131,7 @@ static void set_guest_interrupt(struct lg_cpu *cpu, u32 lo, u32 hi, * interrupt_pending() returns the first pending interrupt which isn't blocked * by the Guest. It is called before every entry to the Guest, and just before * we go to sleep when the Guest has halted itself. */ -unsigned int interrupt_pending(struct lg_cpu *cpu) +unsigned int interrupt_pending(struct lg_cpu *cpu, bool *more) { unsigned int irq; DECLARE_BITMAP(blk, LGUEST_IRQS); @@ -149,13 +149,14 @@ unsigned int interrupt_pending(struct lg_cpu *cpu) /* Find the first interrupt. */ irq = find_first_bit(blk, LGUEST_IRQS); + *more = find_next_bit(blk, LGUEST_IRQS, irq+1); return irq; } /* This actually diverts the Guest to running an interrupt handler, once an * interrupt has been identified by interrupt_pending(). */ -void try_deliver_interrupt(struct lg_cpu *cpu, unsigned int irq) +void try_deliver_interrupt(struct lg_cpu *cpu, unsigned int irq, bool more) { struct desc_struct *idt; @@ -178,8 +179,12 @@ void try_deliver_interrupt(struct lg_cpu *cpu, unsigned int irq) u32 irq_enabled; if (get_user(irq_enabled, &cpu->lg->lguest_data->irq_enabled)) irq_enabled = 0; - if (!irq_enabled) + if (!irq_enabled) { + /* Make sure they know an IRQ is pending. */ + put_user(X86_EFLAGS_IF, + &cpu->lg->lguest_data->irq_pending); return; + } } /* Look at the IDT entry the Guest gave us for this interrupt. The @@ -202,6 +207,11 @@ void try_deliver_interrupt(struct lg_cpu *cpu, unsigned int irq) * here is a compromise which means at least it gets updated every * timer interrupt. */ write_timestamp(cpu); + + /* If there are no other interrupts we want to deliver, clear + * the pending flag. */ + if (!more) + put_user(0, &cpu->lg->lguest_data->irq_pending); } /*:*/ |