ARM Native copro *DIS command calculates negative offset ADR addresses incorrectly?

[mincebert]

I'm not an expert on ARM but I think the *DIS command in the ARMv7 Native Coprocessor calculates the address used in an ADR instruction incorrectly, when the offset is negative.

For example, if I assemble (in BBC BASIC V) at &9018:

.data
EQUW &87654321
ADR R1,data

Then I do *DIS 9018, I get:

>*DIS 9018
00009018 87654321 STRBHI r4, [r5, -r1, LSR #6]!
0000901C E24F100C ADR r1, &9030

... the ADR instruction is actually a SUB with an offset of &C, minus 8 for the pipeline so should be &9018; I think the offset is always being added and not subtracted, when the instruction is determined to be an ADR.

By comparison, *MEMORYI on RISC OS gives (assembled at &901C):

>*MEMORYI 901C
0000901C : !Ceá : 87654321 : STRHIB  R4,[R5,-R1,LSR #6]!
00009020 : ..O‚ : E24F100C : ADR     R1,&0000901C

... same encoding but the address is correctly shown as &901C (in this case).

[hoglet67]

Disassembling this with an Online Disassembler I get:

0x0000000000009018:  87 65 43 21    strbhi r4, [r5, -r1, lsr #6]!
0x000000000000901c:  E2 4F 10 0C    sub    r1, pc, #0xc

But I agree there is something wrong here.

[dp111]

The issue is the disassembler is using banch calculations in the table [I_ADR] = {"cdb"}," in armv7-tbl.c which is only good for forward address calculations.

On Wed, 21 Jun 2023 at 07:42, David Banks @.***> wrote:

Disassembling this with an Online Disassembler https://shell-storm.org/online/Online-Assembler-and-Disassembler/?opcodes=87654321%0D%0AE24F100C&arch=arm&endianness=big&baddr=0x00009018&dis_with_addr=True&dis_with_raw=True&dis_with_ins=True#disassembly I get:

0x0000000000009018: 87 65 43 21 strbhi r4, [r5, -r1, lsr #6]! 0x000000000000901c: E2 4F 10 0C sub r1, pc, #0xc

But I agree there is something wrong here.

— Reply to this email directly, view it on GitHub https://github.com/hoglet67/PiTubeDirect/issues/172#issuecomment-1600269563, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEVVFIU7SWUOQAVEKESMEM3XMKJTTANCNFSM6AAAAAAZN56LJY . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[dp111]

So I think in the 'b' case in darm.c ( line 463) the d->U flag needs to be used to select if the (8+ d->imm) is subtracted instead of added.

On Wed, 21 Jun 2023 at 07:49, Dominic Plunkett @.***> wrote:

The issue is the disassembler is using banch calculations in the table [I_ADR] = {"cdb"}," in armv7-tbl.c which is only good for forward address calculations.

On Wed, 21 Jun 2023 at 07:42, David Banks @.***> wrote:

Disassembling this with an Online Disassembler https://shell-storm.org/online/Online-Assembler-and-Disassembler/?opcodes=87654321%0D%0AE24F100C&arch=arm&endianness=big&baddr=0x00009018&dis_with_addr=True&dis_with_raw=True&dis_with_ins=True#disassembly I get:

0x0000000000009018: 87 65 43 21 strbhi r4, [r5, -r1, lsr #6]! 0x000000000000901c: E2 4F 10 0C sub r1, pc, #0xc

But I agree there is something wrong here.

— Reply to this email directly, view it on GitHub https://github.com/hoglet67/PiTubeDirect/issues/172#issuecomment-1600269563, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEVVFIU7SWUOQAVEKESMEM3XMKJTTANCNFSM6AAAAAAZN56LJY . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[hoglet67]

OK, I'll fix this now in hognose-fixes and merge across to indigo-dev.

Dominic, can I also merge indigo-dev-dp back to indigo-dev?

[dp111]

Yes I'm happy for indigo-dev-dp to be merged into indigo-dev

On Wed, 21 Jun 2023 at 10:21, David Banks @.***> wrote:

OK, I'll fix this now in hognose-fixes and merge across to indigo-dev.

Dominic, can I also merge indigo-dev-dp back to indigo-dev?

— Reply to this email directly, view it on GitHub https://github.com/hoglet67/PiTubeDirect/issues/172#issuecomment-1600493681, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEVVFIV4PTJLMAYJJRYDGR3XMK4IFANCNFSM6AAAAAAZN56LJY . You are receiving this because you commented.Message ID: @.***>

[hoglet67]

OK, this should now be fixed in hognose-fixes: https://github.com/hoglet67/PiTubeDirect/releases/tag/hognose-fixes