summary refs log tree commit diff stats
path: root/results/scraper/launchpad/1748434
blob: b535bf5ef840ea49dde6e01d873312895727b3cc (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
Possibly wrong GICv3 behavior when secure enabled

I an tried arm-aarch64 interrupt routing to EL3, by SCR_EL3.FIQ=1. First I am started QEMU with secure=on and GICv3 support.
I programmed secure and non-secure timers and set-up appropriate interrupts.Secure timer to be GRP1_Secure and non-secure timer to be GRP1_NonSecure. ICC_PMR = 0xff. Then I switched CPU to EL1. 
With that setup no interrupt was delivered to PE. GIC interface showed that non secure IRQ is pending. ICC_PMR read at EL1 returns 0 (shall return value ((PMR_(el3) << 1) & 0xff) according to GIC specification.
Than I tried to increase interrupt priority mask  - so I set ICC_PMR = 0x7f (at EL3). Then I read at EL1 ICC_PMR=0xfe - (is shall be 0). With this setup IRQ of secure timer was taken at EL3, non secure timer didn't rise IRQ (as it is masked by PMR). 
I dig to qemu code and see wrong condition in file arm_gicv3_cpuif.c in function  icc_pmr_read(). This behavior is opposite of ARM specification.

see possible solution, in #if 0 is original code in #else see possible fix

static uint64_t icc_pmr_read(CPUARMState *env, const ARMCPRegInfo *ri)
....
#if 0 // KIURCHER: bug - shall be opposite; see ARM specification
        if (value & 0x80) {
            /* Secure priorities not visible to NS */
            value = 0;
        } else if (value != 0xff) {
            value = (value << 1) & 0xff;
        }
#else

        if (value & 0x80) {
            value = (value << 1) & 0xff;
        } else {
            value = 0;
        }
#endif
....


static void icc_pmr_write(CPUARMState *env, const ARMCPRegInfo *ri,
                          uint64_t value)
....
#if 0 //KIURCHER: bug
    if (arm_feature(env, ARM_FEATURE_EL3) && !arm_is_secure(env) &&
        (env->cp15.scr_el3 & SCR_FIQ)) {
        /* NS access and Group 0 is inaccessible to NS: return the
         * NS view of the current priority
         */
        if (!(cs->icc_pmr_el1 & 0x80)) {
            /* Current PMR in the secure range, don't allow NS to change it */
            return;
        }
        value = (value >> 1) & 0x80;
    }
#else
    if (arm_feature(env, ARM_FEATURE_EL3) && !arm_is_secure(env) &&
        (env->cp15.scr_el3 & SCR_FIQ)) {
        /* NS access and Group 0 is inaccessible to NS: return the
         * NS view of the current priority
         */
        if (!(cs->icc_pmr_el1 & 0x80)) {
            /* Current PMR in the secure range, don't allow NS to change it */
            return;
        }
        value = (value >> 1) | 0x80;
    }
#endif




Whoops, yes, we have the wrong condition for ICC_PMR non-secure reads and writes. We also have the same bug in ICC_RPR reads. I'll put together a patch and send it to the mailing list.


Patch which should fix this:
https://lists.gnu.org/archive/html/qemu-devel/2018-03/msg04537.html


Now fixed in master in commit a2e2d7fc46fd8be, so will be in 2.12.0.