Abort in VCHIQ

I’ve not seen this on RC14 but am seeing the odd Abort daily from VCHIQ in 5.23 (09 & 15/10/15 builds, haven’t tested earlier). Either the IRQ stack has been collapsed prior to it trying to exit, or it’s pulled too many registers.

I can’t find the matching source code in VCHIQ to see what its done previously, so it may be C code.

R0  = 00000001 R1  = 40000193 R2  = 00000000 R3  = 3A606194
R4  = 240DE1FC R5  = 23F66CB4 R6  = 60000192 R7  = 00000000
R8  = 00000000 R9  = 202CC9B4 R10 = FA200000 R11 = 00000000
R12 = 60000192 R13 = FA102000 R14 = FC1DE04C R15 = FC2033C0
Mode IRQ32 flags set: nzCvqjggggeAIft        PSR = 20000192

FC203350 : E1A0700E : MOV     R7,R14
FC203354 : E1A0AA2D : MOV     R10,R13,LSR #20
FC203358 : E1A0AA0A : MOV     R10,R10,LSL #20
FC20335C : E89A0030 : LDMIA   R10,{R4,R5}
FC203360 : E59CC000 : LDR     R12,[R12,#0]
FC203364 : E99C1800 : LDMIB   R12,{R11,R12}
FC203368 : E88A1800 : STMIA   R10,{R11,R12}
FC20336C : E28AAF87 : ADD     R10,R10,#&021C
FC203370 : E3A0B000 : MOV     R11,#0
FC203374 : E1A0E00F : MOV     R14,PC
FC203378 : E28EE00C : ADD     R14,R14,#&0C
FC20337C : E08FF002 : ADD     PC,PC,R2
FC203380 : E1A00000 : MOV     R0,R0
FC203384 : EA0000D2 : B       &FC2036D4
FC203388 : E24AAF87 : SUB     R10,R10,#&021C
FC20338C : E88A0030 : STMIA   R10,{R4,R5}
FC203390 : E1A0E007 : MOV     R14,R7
FC203394 : E13F000F : TEQ     PC,PC
FC203398 : 0A000003 : BEQ     &FC2033AC
FC20339C : E338F000 : TEQP    R8,#0
FC2033A0 : E3300000 : TEQ     R0,#0
FC2033A4 : 08FDCFFF : LDMEQIA R13!,{R0-R11,R14,PC}^
FC2033A8 : 18FD8FFF : LDMNEIA R13!,{R0-R11,PC}^
FC2033AC : E1A0C006 : MOV     R12,R6
FC2033B0 : E121F006 : MSR     CPSR_c,R6
FC2033B4 : E3300000 : TEQ     R0,#0
FC2033B8 : E8BD4FFF : LDMIA   R13!,{R0-R11,R14}
FC2033BC : 049DE004 : LDREQ   R14,[R13],#4
FC2033C0 :>E128F00C : MSR     CPSR_f,R12
FC2033C4 : E1A0F00E : MOV     PC,R14
FC2033C8 : E10F0000 : MRS     R0,CPSR
FC2033CC : E3801080 : ORR     R1,R0,#&80

If it’s not a daft question, why does it have the 26bit check TEQ PC,PC and 26bit exit code? It will always be in 32bit on RO5, the 26bit code is redundant.

I can’t find the matching source code in VCHIQ to see what its done previously, so it may be C code.

It is generic code. It’s the stuff that CMHG generates. You’ll find it in o.modhead.

If it’s not a daft question, why does it have the 26bit check TEQ PC,PC and 26bit exit code? It will always be in 32bit on RO5, the 26bit code is redundant.

True, but CMHG’s modules may be running on a 26 bit system. This may be a defensive exit so it can fail gracefully (or work?) on 26 bit machines (remember there is a 26/32 neutral option too!), or it might be dumb and just output the same boilerplate code all the time.

[I’ve just looked at my module’s object files, compiled with the 32bit option. Yup. It does the same thing. ;) ]