Skip to contentSkip to conversation

Increase x86 interrupt vector range

RFC PATCH Chuck Ebbert 4 participants 4 messages May 13, 2006 – May 15, 2006

lore
Share & tools
lore
#0 lore raw

In-Reply-To: <[email protected]>

Quoted 6 lines · Chris Wright

Remove the limit of 256 interrupt vectors by changing the value stored in orig_{e,r}ax to be the negated interrupt vector. The orig_{e,r}ax needs to be < 0 to allow the signal code to distinguish between return from interrupt and return from syscall. With this change applied, NR_IRQS can be > 256.

Quoted 8 lines

Xen extends the IRQ numbering space to include room for dynamically allocated virtual interrupts (in the range 256-511), which requires a more permissive interface to do_IRQ.

---

AFAIC this could go in anytime. It's simple, self-contained and makes sense even without Xen.

One minor nit: the IRQ value isn't negated, it's complemented. The comments need to be fixed.

Original patch:

 arch/i386/kernel/entry.S    |    4 ++--
 arch/i386/kernel/irq.c      |    4 ++--
 arch/x86_64/kernel/entry.S  |    2 +-
 arch/x86_64/kernel/irq.c    |    4 ++--
 arch/x86_64/kernel/smp.c    |    4 ++--
 include/asm-x86_64/hw_irq.h |    2 +-
 6 files changed, 10 insertions(+), 10 deletions(-)
arch/i386/kernel/entry.S +2 -2
@@ -464,7 +464,7 @@ vector=0
 ENTRY(irq_entries_start)
 .rept NR_IRQS
        ALIGN
-1:     pushl $vector-256
+1:     pushl $~(vector)
        jmp common_interrupt
 .data
        .long 1b
@@ -481,7 +481,7 @@ common_interrupt:
 
 #define BUILD_INTERRUPT(name, nr)      \
 ENTRY(name)                            \
-       pushl $nr-256;                  \
+       pushl $~(nr);                   \
        SAVE_ALL                        \
        movl %esp,%eax;                 \
        call smp_/**/name;              \
arch/i386/kernel/irq.c +2 -2
@@ -53,8 +53,8 @@ static union irq_ctx *softirq_ctx[NR_CPU
  */
 fastcall unsigned int do_IRQ(struct pt_regs *regs)
 {      
-       /* high bits used in ret_from_ code */
-       int irq = regs->orig_eax & 0xff;
+       /* high bit used in ret_from_ code */
+       int irq = ~regs->orig_eax;
 #ifdef CONFIG_4KSTACKS
        union irq_ctx *curctx, *irqctx;
        u32 *isp;
arch/x86_64/kernel/entry.S +1 -1
@@ -601,7 +601,7 @@ retint_kernel:      
  */            
        .macro apicinterrupt num,func
        INTR_FRAME
-       pushq $\num-256
+       pushq $~(\num)
        CFI_ADJUST_CFA_OFFSET 8
        interrupt \func
        jmp ret_from_intr
arch/x86_64/kernel/irq.c +2 -2
@@ -91,8 +91,8 @@ skip:
  */
 asmlinkage unsigned int do_IRQ(struct pt_regs *regs)
 {      
-       /* high bits used in ret_from_ code  */
-       unsigned irq = regs->orig_rax & 0xff;
+       /* high bit used in ret_from_ code  */
+       unsigned irq = ~regs->orig_rax;
 
        exit_idle();
        irq_enter();
arch/x86_64/kernel/smp.c +2 -2
@@ -135,10 +135,10 @@ asmlinkage void smp_invalidate_interrupt
 
        cpu = smp_processor_id();
        /*
-        * orig_rax contains the interrupt vector - 256.
+        * orig_rax contains the negated interrupt vector.
         * Use that to determine where the sender put the data.
         */
-       sender = regs->orig_rax + 256 - INVALIDATE_TLB_VECTOR_START;
+       sender = ~regs->orig_rax - INVALIDATE_TLB_VECTOR_START;
        f = &per_cpu(flush_state, sender);
 
        if (!cpu_isset(cpu, f->flush_cpumask))
include/asm-x86_64/hw_irq.h +1 -1
@@ -127,7 +127,7 @@ asmlinkage void IRQ_NAME(nr); \
 __asm__( \
 "\n.p2align\n" \
 "IRQ" #nr "_interrupt:\n\t" \
-       "push $" #nr "-256 ; " \
+       "push $~(" #nr ") ; " \
        "jmp common_interrupt");
 
 #if defined(CONFIG_X86_IO_APIC)
Signature · 4 lines
Chuck

"The x86 isn't all that complex -- it just doesn't make a lot of sense."
                                                        -- Mike Johnson
#1 lore raw
Quoted 5 lines · Chuck Ebbert

AFAIC this could go in anytime. It's simple, self-contained and makes sense even without Xen.

One minor nit: the IRQ value isn't negated, it's complemented. The comments need to be fixed.

If we really want it disambiguated then we should call it something like 'bitwise-negated' or 'ones-complemented'. 'Complemented' alone is worse than 'negated' imo, since negation is at least the usual name for the tilde operator in C while complementation is an ambiguous term unless you know the base/radix.

-- Keir

#2 lore raw
Quoted 5 lines · Keir Fraser

If we really want it disambiguated then we should call it something like 'bitwise-negated' or 'ones-complemented'. 'Complemented' alone is worse than 'negated' imo, since negation is at least the usual name for the tilde operator in C while complementation is an ambiguous term unless you know the base/radix.

Not really.

I've seen "negate" only be used to mean "twos complement" and "invert" only be seen with "ones complement".

So if you like to be clear, just use "negate" vs. "invert".

Regards

Ingo Oeser

#3 lore raw
In reply to Chuck Ebbert
Quoted 14 lines · * Chuck Ebbert ([email protected]) wrote:

In-Reply-To: <[email protected]>

Quoted 10 lines · Chris Wright

Remove the limit of 256 interrupt vectors by changing the value stored in orig_{e,r}ax to be the negated interrupt vector. The orig_{e,r}ax needs to be < 0 to allow the signal code to distinguish between return from interrupt and return from syscall. With this change applied, NR_IRQS can be > 256.

Xen extends the IRQ numbering space to include room for dynamically allocated virtual interrupts (in the range 256-511), which requires a more permissive interface to do_IRQ.

Quoted 7 lines
Quoted 4 lines

---

AFAIC this could go in anytime. It's simple, self-contained and makes sense even without Xen.

The plan is to split out things like this and push them ahead of the rest of the batch.

Quoted 2 lines

One minor nit: the IRQ value isn't negated, it's complemented. The comments need to be fixed.

OK, bitwise negated.

thanks,
-chris