Skip to content

ISA support updates #3236

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Feb 9, 2024
Merged

ISA support updates #3236

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
113 changes: 23 additions & 90 deletions docs/isa-support.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,70 +30,44 @@ opcode src imm off description
0x1d any 0x00 any if dst == src goto +offset Y Y Y jeq-reg
0x1e any 0x00 any if (u32)dst == (u32)src goto +offset Y Y Y jeq32-reg
0x1f any 0x00 0 dst -= src Y Y Y alu64-arith
0x20 any any any (deprecated, implementation-specific) no no no (none)
0x21 any any any (deprecated, implementation-specific) no no no (none)
0x22 any any any (deprecated, implementation-specific) no no no (none)
0x23 any any any (deprecated, implementation-specific) no no no (none)
0x20 any any any (deprecated, implementation-specific) Y no no (none)
0x24 0x0 any 0 dst = (u32)(dst \* imm) Y Y Y mul32-imm
0x25 0x0 any any if dst > imm goto +offset Y Y Y jgt-imm
0x26 0x0 any any if (u32)dst > imm goto +offset Y Y Y jgt32-imm
0x27 0x0 any 0 dst \*= imm Y Y Y mul64-imm
0x28 any any any (deprecated, implementation-specific) no no no (none)
0x29 any any any (deprecated, implementation-specific) no no no (none)
0x2a any any any (deprecated, implementation-specific) no no no (none)
0x2b any any any (deprecated, implementation-specific) no no no (none)
0x28 any any any (deprecated, implementation-specific) Y no no (none)
0x2c any 0x00 0 dst = (u32)(dst \* src) Y Y Y mul32-reg
0x2d any 0x00 any if dst > src goto +offset Y Y Y jgt-reg
0x2e any 0x00 any if (u32)dst > (u32)src goto +offset Y Y Y jgt32-reg
0x2f any 0x00 0 dst \*= src Y Y Y mul64-reg
0x30 any any any (deprecated, implementation-specific) no no no (none)
0x31 any any any (deprecated, implementation-specific) no no no (none)
0x32 any any any (deprecated, implementation-specific) no no no (none)
0x33 any any any (deprecated, implementation-specific) no no no (none)
0x30 any any any (deprecated, implementation-specific) Y no no (none)
0x34 0x0 any 0 dst = (u32)((imm != 0) ? ((u32)dst / (u32)imm) : 0) Y Y Y alu-arith
0x34 0x0 any 1 dst = (u32)((imm != 0) ? ((s32)dst s/ imm) : 0) Y no Y sdiv32-imm
0x35 0x0 any any if dst >= imm goto +offset Y Y Y jge-imm
0x36 0x0 any any if (u32)dst >= imm goto +offset Y Y Y jge32-imm
0x37 0x0 any 0 dst = (imm != 0) ? (dst / (u32)imm) : 0 Y Y Y alu64-arith
0x37 0x0 any 1 dst = (imm != 0) ? (dst s/ imm) : 0 Y no Y sdiv64-imm
0x38 any any any (deprecated, implementation-specific) no no no (none)
0x39 any any any (deprecated, implementation-specific) no no no (none)
0x3a any any any (deprecated, implementation-specific) no no no (none)
0x3b any any any (deprecated, implementation-specific) no no no (none)
0x3c any 0x00 0 dst = (u32)((src != 0) ? ((u32)dst / (u32)src) : 0) Y Y Y alu-arith
0x3c any 0x00 1 dst = (u32)((src != 0) ? ((s32)dst s/ (s32)src) : 0) Y no Y sdiv32-reg
0x3d any 0x00 any if dst >= src goto +offset Y Y Y prime
0x3e any 0x00 any if (u32)dst >= (u32)src goto +offset Y Y Y jge32-reg
0x3f any 0x00 0 dst = (src !+ 0) ? (dst / src) : 0 Y Y Y alu64-arith
0x3f any 0x00 1 dst = (src !+ 0) ? (dst s/ src) : 0 Y no Y sdiv64-reg
0x40 any any any (deprecated, implementation-specific) no no no (none)
0x41 any any any (deprecated, implementation-specific) no no no (none)
0x42 any any any (deprecated, implementation-specific) no no no (none)
0x43 any any any (deprecated, implementation-specific) no no no (none)
0x40 any any any (deprecated, implementation-specific) Y no no (none)
0x44 0x0 any 0 dst = (u32)(dst \| imm) Y Y Y alu-bit
0x45 0x0 any any if dst & imm goto +offset Y Y Y jset-imm
0x46 0x0 any any if (u32)dst & imm goto +offset Y Y Y jset32-imm
0x47 0x0 any 0 dst \|= imm Y Y Y alu64-bit
0x48 any any any (deprecated, implementation-specific) no no no (none)
0x49 any any any (deprecated, implementation-specific) no no no (none)
0x4a any any any (deprecated, implementation-specific) no no no (none)
0x4b any any any (deprecated, implementation-specific) no no no (none)
0x48 any any any (deprecated, implementation-specific) Y no no (none)
0x4c any 0x00 0 dst = (u32)(dst \| src) Y Y Y alu-bit
0x4d any 0x00 any if dst & src goto +offset Y Y Y jset-reg
0x4e any 0x00 any if (u32)dst & (u32)src goto +offset Y Y Y jset32-reg
0x4f any 0x00 0 dst \|= src Y Y Y alu64-bit
0x50 any any any (deprecated, implementation-specific) no no no (none)
0x51 any any any (deprecated, implementation-specific) no no no (none)
0x52 any any any (deprecated, implementation-specific) no no no (none)
0x53 any any any (deprecated, implementation-specific) no no no (none)
0x50 any any any (deprecated, implementation-specific) Y no no (none)
0x54 0x0 any 0 dst = (u32)(dst & imm) Y Y Y alu-bit
0x55 0x0 any any if dst != imm goto +offset Y Y Y alu-arith
0x56 0x0 any any if (u32)dst != imm goto +offset Y Y Y jne32-imm
0x57 0x0 any 0 dst &= imm Y Y Y alu64-bit
0x58 any any any (deprecated, implementation-specific) no no no (none)
0x59 any any any (deprecated, implementation-specific) no no no (none)
0x5a any any any (deprecated, implementation-specific) no no no (none)
0x5b any any any (deprecated, implementation-specific) no no no (none)
0x5c any 0x00 0 dst = (u32)(dst & src) Y Y Y alu-bit
0x5d any 0x00 any if dst != src goto +offset Y Y Y jne-reg
0x5e any 0x00 any if (u32)dst != (u32)src goto +offset Y Y Y jne32-reg
Expand Down Expand Up @@ -162,34 +136,15 @@ opcode src imm off description
0xbf any 0x00 16 dst = (s64) (s16) src Y no Y movsx1664-reg
0xbf any 0x00 32 dst = (s64) (s32) src Y no Y movsx3264-reg
0xc3 any 0x00 any lock \*(u32 \*)(dst + offset) += src no no Y lock_add32
0xc3 any 0x01 any | lock no no Y lock_fetch_add32
| temp = \*(u32 \*)(dst + offset)
| \*(u32 \*)(dst + offset) += src
| src = temp
0xc3 any 0x01 any src = atomic_fetch_add_32((u32 \*)(dst + offset), src) no no Y lock_fetch_add32
0xc3 any 0x40 any lock \*(u32 \*)(dst + offset) \|= src no no Y lock_or32
0xc3 any 0x41 any | lock no no Y lock_fetch_or32
| temp = \*(u32 \*)(dst + offset)
| \*(u32 \*)(dst + offset) |= src
| src = temp
0xc3 any 0x41 any src = atomic_fetch_or_32((u32 \*)(dst + offset), src) no no Y lock_fetch_or32
0xc3 any 0x50 any lock \*(u32 \*)(dst + offset) &= src no no Y lock_and32
0xc3 any 0x51 any | lock no no Y lock_fetch_and32
| temp = \*(u32 \*)(dst + offset)
| \*(u32 *)(dst + offset) &= src
| src = temp
0xc3 any 0x51 any src = atomic_fetch_and_32((u32 \*)(dst + offset), src) no no Y lock_fetch_and32
0xc3 any 0xa0 any lock \*(u32 \*)(dst + offset) ^= src no no Y lock_xor32
0xc3 any 0xa1 any | lock no no Y lock_fetch_xor32
| temp = \*(u32 \*)(dst + offset)
| \*(u32 \*)(dst + offset) ^= src
| src = temp
0xc3 any 0xe1 any | lock no no Y lock_xchg32
| temp = \*(u32 \*)(dst + offset)
| \*(u32 \*)(dst + offset) = src
| src = temp
0xc3 any 0xf1 any | lock no no Y lock_cmpxchg32
| temp = \*(u32 \*)(dst + offset)
| if \*(u32)(dst + offset) == R0
| \*(u32)(dst + offset) = src
| R0 = temp
0xc3 any 0xa1 any src = atomic_fetch_xor_32((u32 \*)(dst + offset), src) no no Y lock_fetch_xor32
0xc3 any 0xe1 any src = xchg_32((u32 \*)(dst + offset), src) no no Y lock_xchg32
0xc3 any 0xf1 any r0 = cmpxchg_32((u32 \*)(dst + offset), r0, src) no no Y lock_cmpxchg32
0xc4 0x0 any 0 dst = (u32)(dst s>> imm) Y Y Y arsh
0xc5 0x0 any any if dst s< imm goto +offset Y Y Y jslt-imm
0xc6 0x0 any any if (s32)dst s< (s32)imm goto +offset Y Y Y jslt32-imm
Expand All @@ -207,46 +162,24 @@ opcode src imm off description
0xd7 0x0 0x20 0 dst = bswap32(dst) Y no Y swap32
0xd7 0x0 0x40 0 dst = bswap64(dst) Y no Y swap64
0xdb any 0x00 any lock \*(u64 \*)(dst + offset) += src no no Y lock_add
0xdb any 0x01 any | lock no no Y lock_fetch_add
| temp = \*(u64 \*)(dst + offset)
| \*(u64 \*)(dst + offset) += src
| src = temp
0xdb any 0x01 any src = atomic_fetch_add_64((u64 \*)(dst + offset), src) no no Y lock_fetch_add
0xdb any 0x40 any lock \*(u64 \*)(dst + offset) \|= src no no Y lock_or
0xdb any 0x41 any | lock no no Y lock_fetch_or
| temp = \*(u64 \*)(dst + offset)
| \*(u64 \*)(dst + offset) |= src
| src = temp
0xdb any 0x41 any src = atomic_fetch_or_64((u64 \*)(dst + offset), src) no no Y lock_fetch_or
0xdb any 0x50 any lock \*(u64 \*)(dst + offset) &= src no no Y lock_and
0xdb any 0x51 any | lock no no Y lock_fetch_and
| temp = \*(u64 \*)(dst + offset)
| \*(u64 \*)(dst + offset) &= src
| src = temp
0xdb any 0x51 any src = atomic_fetch_and_64((u64 \*)(dst + offset), src) no no Y lock_fetch_and
0xdb any 0xa0 any lock \*(u64 \*)(dst + offset) ^= src no no Y lock_xor
0xdb any 0xa1 any | lock no no Y lock_fetch_xor
| temp = \*(u64 \*)(dst + offset)
| \*(u64 \*)(dst + offset) ^= src
| src = temp
0xdb any 0xe1 any | lock no no Y lock_xchg
| temp = \*(u64 \*)(dst + offset)
| \*(u64 \*)(dst + offset) = src
| src = temp
0xdb any 0xf1 any | lock no no Y lock_cmpxchg
| temp = \*(u64 \*)(dst + offset)
| if \*(u64)(dst + offset) == R0
| \*(u64)(dst + offset) = src
| R0 = temp
0xdc 0x0 0x10 0 dst = htobe16(dst) Y Y Y be16
0xdc 0x0 0x20 0 dst = htobe32(dst) Y Y Y be32
0xdc 0x0 0x40 0 dst = htobe64(dst) Y Y Y be64
0xdd any 0x00 any if dst s<= src goto +offset Y Y Y jsle-reg
0xde any 0x00 any if (s32)dst s<= (s32)src goto +offset Y Y Y jsle32-reg
0xdb any 0xa1 any src = atomic_fetch_xor_64((u64 \*)(dst + offset), src) no no Y lock_fetch_xor
0xdb any 0xe1 any src = xchg_64((u64 \*)(dst + offset), src) no no Y lock_xchg
0xdb any 0xf1 any r0 = cmpxchg_64((u64 \*)(dst + offset), r0, src) no no Y lock_cmpxchg
0xdc 0x0 0x10 0 dst = htobe16(dst) Y Y Y be16
0xdc 0x0 0x20 0 dst = htobe32(dst) Y Y Y be32
0xdc 0x0 0x40 0 dst = htobe64(dst) Y Y Y be64
0xdd any 0x00 any if dst s<= src goto +offset Y Y Y jsle-reg
0xde any 0x00 any if (s32)dst s<= (s32)src goto +offset Y Y Y jsle32-reg
====== ==== ==== ==== ====================================================== ======= ==== ===== ======================

**Some takeaways:**

* Some ldx instruction conformance issues still exist in the PREVAIL verifier, where instructions
fail verification that shouldn't. This is not a security issue, it might simply prevent some valid
programs from being verified (https://github.com/vbpf/ebpf-verifier/issues/420).
* Atomic instructions are not supported by any of the components, though this is not a major problem
as they will not be generated by clang when an older "cpu version" is specified on the command line.
* The conformance suite does not support most 64-bit immediate instructions
Expand Down