mistranslation: 0.713
arm: 0.670
semantic: 0.575
architecture: 0.557
kernel: 0.545
device: 0.294
assembly: 0.255
socket: 0.171
register: 0.157
PID: 0.143
virtual: 0.140
x86: 0.140
peripherals: 0.127
graphic: 0.124
ppc: 0.120
performance: 0.116
permissions: 0.095
hypervisor: 0.091
debug: 0.087
vnc: 0.080
user-level: 0.076
risc-v: 0.073
network: 0.065
files: 0.064
boot: 0.058
VMM: 0.043
TCG: 0.032
i386: 0.029
KVM: 0.024

ARM: CORTEX M, PRIMASK does not disable interrupts

qemu version 0.15.1
but the same code is in qemu 1.0

"CPSID I" does not disable interrupts for CORTEX M3


if (interrupt_request & CPU_INTERRUPT_HARD
                        && ((IS_M(env) && env->regs[15] < 0xfffffff0)
                            || !(env->uncached_cpsr & CPSR_I))) {
                        env->exception_index = EXCP_IRQ;
                        do_interrupt(env);
                        next_tb = 0;
                    }


do_interrupt() will be executed even if (env->uncached_cpsr & CPSR_I) == 1 , disable interrupt bit set.


then changed to: 

if (interrupt_request & CPU_INTERRUPT_HARD 
                        && !(env->uncached_cpsr & CPSR_I)
                        && (IS_M(env) ? env->regs[15] < 0xfffffff0: 1) ) {
                        env->exception_index = EXCP_IRQ;
                        do_interrupt(env);
                        next_tb = 0;
                    }

works

This change changes the behaviour for non-M-profile cores, which looks wrong.

See discussion in this mailing list thread where a similar patch was suggested:
 http://lists.gnu.org/archive/html/qemu-devel/2011-06/msg00500.html

M profile interrupt handling is known-broken. I'm not accepting any patches in this area unless they come attached to a decent explanation of why they are the correct change to make and show some evidence of the whole problem having been considered.


> This change changes the behaviour for non-M-profile cores

...or maybe not: I was confused by the resemblance to that other patch. I still think one-line fixes are unlikely to be the right approach here, though.


This long-standing bug has been fixed by the rewrite of the M-profile exception handling for QEMU 2.9.