Hexagon: update and enforce hintjr slot constraints
cover Hexagon: update and enforce hintjr slot constraints
This small series basically fix an outdated hexagon slot rule for instruction hintjr and makes the decode machinery more resilient against any possible outdated slot constraints in the future.
Matheus Tavares Bernardino (2):
Hexagon (iclass): update J4_hintjumpr slot constraints Hexagon (decode): look for pkts with multiple insns at the same slot
target/hexagon/decode.c | 30 +++++++++++++++++++++++++++--- target/hexagon/iclass.c | 4 +++- tests/tcg/hexagon/Makefile.target | 10 ++++++++++ tests/tcg/hexagon/invalid_slots.c | 29 +++++++++++++++++++++++++++++ 4 files changed, 69 insertions(+), 4 deletions(-) create mode 100644 tests/tcg/hexagon/invalid_slots.c
Signature · 1 line
2.37.2
PATCH 1/2 Hexagon (iclass): update J4_hintjumpr slot constraints
The Hexagon PRM says that "The assembler automatically encodes instructions in the packet in the proper order. In the binary encoding of a packet, the instructions must be ordered from Slot 3 down to Slot 0."
Prior to the architecture version v73, the slot constraints from instruction "hintjr" only allowed it to be executed at slot 2. With that in mind, consider the packet:
{
hintjr(r0)
nop
nop
if (!p0) memd(r1+#0) = r1:0
}To satisfy the ordering rule quoted from the PRM, the assembler would, thus, move one of the nops to the first position, so that it can be assigned to slot 3 and the subsequent hintjr to slot 2.
However, since v73, hintjr can be executed at either slot 2 or 3. So there is no need to reorder that packet and the assembler will encode it as is. When QEMU tries to execute it, however, we end up hitting a "misaliged store" exception because both the store and the hintjr will be assigned to store 0, and some functions like slot_is_predicated() expect the decode machinery to assign only one instruction per slot. In particular, the mentioned function will traverse the packet until it finds the first instruction at the desired slot which, for slot 0, will be hintjr. Since hintjr is not predicated, the result is that we try to execute the store regardless of the predicate. And because the predicate is false, we had not previously loaded hex_store_addr[0] or hex_store_width[0]. As a result, the store will decide de width based on trash memory, causing it to be misaligned.
Update the slot constraints for hintjr so that QEMU can properly handle such encodings.
Note: to avoid similar-but-not-identical issues in the future, we should look for multiple instructions at the same slot during decoding time and throw an invalid packet exception. That will be done in the subsequent commit.
--- target/hexagon/iclass.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
target/hexagon/iclass.c +3 -1
@@ -51,8 +51,10 @@ SlotMask find_iclass_slots(Opcode opcode, int itype)return SLOTS_0 ; }else if ((opcode == J2_trap0) || (opcode == Y2_isync) || - (opcode == J2_pause) || (opcode == J4_hintjumpr)) { + (opcode == J2_pause)) {return SLOTS_2 ; + }else if (opcode == J4_hintjumpr) { +return SLOTS_23 ; }else if (GET_ATTRIB (opcode,A_CRSLOT23 )) {return SLOTS_23 ; }else if (GET_ATTRIB (opcode,A_RESTRICT_PREFERSLOT0 )) {
Signature · 1 line
2.37.2
RE: [PATCH 1/2] Hexagon (iclass): update J4_hintjumpr slot constraints
Quoted 45 lines
-----Original Message-----
From: Matheus Tavares Bernardino <[email protected]>
Sent: Friday, January 13, 2023 7:39 AM
To: [email protected]
Cc: Taylor Simpson <[email protected]>; Brian Cain <[email protected]>; [email protected]
Subject: [PATCH 1/2] Hexagon (iclass): updateJ4_hintjumprslot constraintsThe Hexagon PRM says that "The assembler automatically encodes instructions in the packet in the proper order. In the binary encoding of a packet, the instructions must be ordered from Slot 3 down to Slot 0."
Prior to the architecture version v73, the slot constraints from instruction "hintjr" only allowed it to be executed at slot 2.
With that in mind, consider the packet:{ hintjr(r0) nop nop if (!p0) memd(r1+#0) = r1:0 }To satisfy the ordering rule quoted from the PRM, the assembler would, thus, move one of the nops to the first position, so that it can be assigned to slot 3 and the subsequent hintjr to slot 2.
However, since v73, hintjr can be executed at either slot 2 or 3. So there is no need to reorder that packet and the assembler will encode it as is. When QEMU tries to execute it, however, we end up hitting a "misaliged store" exception because both the store and the hintjr will be assigned to store 0, and some functions like
slot_is_predicated()expect the decode machinery to assign only one instruction per slot. In particular, the mentioned function will traverse the packet until it finds the first instruction at the desired slot which, for slot 0, will be hintjr. Since hintjr is not predicated, the result is that we try to execute the store regardless of the predicate. And because the predicate is false, we had not previously loaded hex_store_addr[0] or hex_store_width[0]. As a result, the store will decide de width based on trash memory, causing it to be misaligned.Update the slot constraints for hintjr so that QEMU can properly handle such encodings.
Quoted 8 lines
Note: to avoid similar-but-not-identical issues in the future, we should look for multiple instructions at the same slot during decoding time and throw an invalid packet exception. That will be done in the subsequent commit.
Signed-off-by Matheus Tavares Bernardino <[email protected]>--- target/hexagon/iclass.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
Could you add a check-tcg test case?
Thanks,
Taylor
RE: [PATCH 1/2] Hexagon (iclass): update J4_hintjumpr slot constraints
In reply to Matheus Tavares BernardinoQuoted 60 lines
-----Original Message-----
From: Matheus Tavares Bernardino <[email protected]>
Sent: Friday, January 13, 2023 7:39 AM
To: [email protected]
Cc: Taylor Simpson <[email protected]>; Brian Cain <[email protected]>; [email protected]
Subject: [PATCH 1/2] Hexagon (iclass): updateJ4_hintjumprslot constraintsThe Hexagon PRM says that "The assembler automatically encodes instructions in the packet in the proper order. In the binary encoding of a packet, the instructions must be ordered from Slot 3 down to Slot 0."
Prior to the architecture version v73, the slot constraints from instruction "hintjr" only allowed it to be executed at slot 2.
With that in mind, consider the packet:{ hintjr(r0) nop nop if (!p0) memd(r1+#0) = r1:0 }To satisfy the ordering rule quoted from the PRM, the assembler would, thus, move one of the nops to the first position, so that it can be assigned to slot 3 and the subsequent hintjr to slot 2.
However, since v73, hintjr can be executed at either slot 2 or 3. So there is no need to reorder that packet and the assembler will encode it as is. When QEMU tries to execute it, however, we end up hitting a "misaliged store" exception because both the store and the hintjr will be assigned to store 0, and some functions like
slot_is_predicated()expect the decode machinery to assign only one instruction per slot. In particular, the mentioned function will traverse the packet until it finds the first instruction at the desired slot which, for slot 0, will be hintjr. Since hintjr is not predicated, the result is that we try to execute the store regardless of the predicate. And because the predicate is false, we had not previously loaded hex_store_addr[0] or hex_store_width[0]. As a result, the store will decide de width based on trash memory, causing it to be misaligned.Update the slot constraints for hintjr so that QEMU can properly handle such encodings.
Note: to avoid similar-but-not-identical issues in the future, we should look for multiple instructions at the same slot during decoding time and throw an invalid packet exception. That will be done in the subsequent commit.
Signed-off-by Matheus Tavares Bernardino <[email protected]>--- target/hexagon/iclass.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)target/hexagon/iclass.c index renamed from target/hexagon/iclass.c +0
-06091286993..081116fc4d 100644
target/hexagon/iclass.c +0
-0@@ -51,8 +51,10 @@ SlotMask find_iclass_slots(Opcode opcode, int itype)return SLOTS_0 ;
target/hexagon/iclass.c SlotMask find_iclass_slots(Opcode opcode, int itype) · line 60
Quoted 10 lines
target/hexagon/iclass.c +3
-1}else if ((opcode == J2_trap0) || (opcode == Y2_isync) || - (opcode == J2_pause) || (opcode == J4_hintjumpr)) { + (opcode == J2_pause)) {return SLOTS_2 ; + }else if (opcode == J4_hintjumpr) { +return SLOTS_23 ; }else if (GET_ATTRIB (opcode,A_CRSLOT23 )) {return SLOTS_23 ; }else if (GET_ATTRIB (opcode,A_RESTRICT_PREFERSLOT0 )) {
PATCH 2/2 Hexagon (decode): look for pkts with multiple insns at the same slot
In reply to Matheus Tavares BernardinoEach slot in a packet can be assigned to at most one instruction. Although the assembler generally ought to enforce this rule, we better be safe than sorry and also do some check to properly throw an "invalid packet" exception on wrong slot assignments.
This should also make it easier to debug possible future errors caused by missing updates to find_iclass_slots() rules in target/hexagon/iclass.c.
--- target/hexagon/decode.c | 30 +++++++++++++++++++++++++++--- tests/tcg/hexagon/Makefile.target | 10 ++++++++++ tests/tcg/hexagon/invalid_slots.c | 29 +++++++++++++++++++++++++++++ 3 files changed, 66 insertions(+), 3 deletions(-) create mode 100644 tests/tcg/hexagon/invalid_slots.c
target/hexagon/decode.c +27 -3
@@ -1,5 +1,5 @@/* -* Copyright(c) 2019-2022 Qualcomm Innovation Center, Inc. All Rights Reserved. +* Copyright(c) 2019-2023 Qualcomm Innovation Center, Inc. All Rights Reserved. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by
@@ -797,7 +797,26 @@ static bool decode_parsebits_is_loopend(uint32_t encoding32)return bits ==0x2 ; } -static void +static bool has_valid_slot_assignment (Packet *pkt) +{ +int used_slots =0 ; +for (int i =0 ; i < pkt->num_insns; i++) { +int slot_mask; +Insn *insn = &pkt->insn[i]; +if (decode_opcode_ends_loop (insn->opcode)) { +/* We overload slot 0 for endloop. */ +continue ; + } + slot_mask =1 << insn->slot; +if (used_slots & slot_mask) { +return false; + } + used_slots |= slot_mask; + } +return true; +} + +static bool decode_set_slot_number (Packet *pkt) {int slot;
@@ -886,6 +905,8 @@ decode_set_slot_number(Packet *pkt)/* Then push it to slot0 */ pkt->insn[slot1_iidx].slot =0 ; } + +return has_valid_slot_assignment (pkt); }/*
@@ -962,7 +983,10 @@ int decode_packet(int max_words, const uint32_t *words, Packet *pkt,if (!disas_only) {decode_remove_extenders (pkt); } -decode_set_slot_number (pkt); +if (!decode_set_slot_number (pkt)) { +/* Invalid packet */ +return 0 ; + }decode_fill_newvalue_regno (pkt);if (pkt->pkt_has_hvx) {
tests/tcg/hexagon/Makefile.target +10 -0
@@ -45,6 +45,16 @@ HEX_TESTS += fpstuff HEX_TESTS += overflow HEX_TESTS += signal_context HEX_TESTS += reg_mut +HEX_TESTS += invalid_slots + +run-invalid_slots: + $(call run-test, $<, $(QEMU) $< 2> $<.stderr, $<); \ + if [ $$? -ne 1 ] ; then \ + return 1; \ + fi + $(call quiet-command, \ + grep -q "exception 0x15" $<.stderr, \ + "GREP", "exception 0x15"); HEX_TESTS += test_abs HEX_TESTS += test_bitcnt
tests/tcg/hexagon/invalid_slots.c new file +29 -0
@@ -0,0 +1,29 @@ +/* +* Copyright(c) 2023 Qualcomm Innovation Center, Inc. All Rights Reserved. +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License as published by +* the Free Software Foundation; either version 2 of the License, or +* (at your option) any later version. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. +* +* You should have received a copy of the GNU General Public License +* along with this program; if not, see <http://www.gnu.org/licenses/>. +*/ + +char mem[8] __attribute__((aligned(8))); + +int main () +{ + asmvolatile ( +"r0 = #mem\n" +/* Invalid packet (2 instructions at slot 0): */ +".word 0xa1804100\n" /* { memw(r0) = r1; */ +".word 0x28032804\n" /* r3 = #0; r4 = #0 } */ + : : :"r0" ,"r3" ,"r4" ,"memory" ); +return 0 ; +}
Signature · 1 line
2.37.2
RE: [PATCH 2/2] Hexagon (decode): look for pkts with multiple insns at the same slot
Quoted 19 lines
-----Original Message-----
From: Matheus Tavares Bernardino <[email protected]>
Sent: Friday, January 13, 2023 7:39 AM
To: [email protected]
Cc: Taylor Simpson <[email protected]>; Brian Cain <[email protected]>; [email protected]
Subject: [PATCH 2/2] Hexagon (decode): look for pkts with multiple insns at the same slotEach slot in a packet can be assigned to at most one instruction. Although the assembler generally ought to enforce this rule, we better be safe than sorry and also do some check to properly throw an "invalid packet" exception on wrong slot assignments.
This should also make it easier to debug possible future errors caused by missing updates to
find_iclass_slots()rules intarget/hexagon/iclass.c.
Quoted 7 lines
Signed-off-by Matheus Tavares Bernardino <[email protected]>--- target/hexagon/decode.c | 30 +++++++++++++++++++++++++++--- tests/tcg/hexagon/Makefile.target | 10 ++++++++++ tests/tcg/hexagon/invalid_slots.c | 29 +++++++++++++++++++++++++++++ 3 files changed, 66 insertions(+), 3 deletions(-) create mode 100644
tests/tcg/hexagon/invalid_slots.c