* [PATCH v5 00/15] riscv support for control flow integrity extensions
@ 2024-08-20 0:01 Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 01/15] target/riscv: Add zicfilp extension Deepak Gupta
` (14 more replies)
0 siblings, 15 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta
v5 for riscv zicfilp and zicfiss extensions support in qemu.
Links for previous versions
[1] - v1 https://lists.nongnu.org/archive/html/qemu-devel/2024-07/msg06017.html
[2] - v2 https://lore.kernel.org/all/ed23bcbc-fdc4-4492-803c-daa95880375a@linaro.org/T/
[3] - v3 https://lists.nongnu.org/archive/html/qemu-devel/2024-08/msg01005.html
[4] - v4 https://lore.kernel.org/all/20240816010711.3055425-6-debug@rivosinc.com/T/
---
v5:
- Simplified elp tracking and lpad implementation as per suggestion by richard
- Simplified shadow stack mmu checks as per suggestion by richard
- Converged zicfiss compressed and non-comressed instructions to same translation
- Removed trace hooks. Don't need for upstream.
v4:
- elp state in cpu is true/false instead of enum and elp cleared
unconditionally on trap entry. elp in *status cleared unconditionally on
trap return.
- Moved logic for branch tracking in instruction translation from tb_start.
- fixed zicfiss dependency on 'A'
- `cpu_get_fcfien/bcfien` helpers checks fixed to check for extension first.
- removed trace hook enums. Instead added dedicated trace helpers wherever needed.
- fixed/simplified instruction format in decoder for lpad, sspush, sspopchk
- simplified tlb index logic for shadow stack instructions. Removed SUM TB_FLAG
- access to ssp CSR is gated on `cpu_get_bcfien` instead of duplicated logic
- removed vDSO related changes for now.
v3:
- Removed prctl specific patches because they need to be upstream
in kernel first.
- As suggested by Richard, added TB flag if fcfi enabled
- Re-worked translation for landing pad and shadow stack instructions
to not require helper.
- tcg helpers only for cfi violation cases so that trace hooks can be
placed.
- Style changes.
- fixes assert condition in accel/tcg
v2:
- added missed file (in v1) for shadow stack instructions implementation.
Deepak Gupta (15):
target/riscv: Add zicfilp extension
target/riscv: Introduce elp state and enabling controls for zicfilp
target/riscv: save and restore elp state on priv transitions
target/riscv: additional code information for sw check
target/riscv: tracking indirect branches (fcfi) for zicfilp
target/riscv: zicfilp `lpad` impl and branch tracking
disas/riscv: enable `lpad` disassembly
target/riscv: Add zicfiss extension
target/riscv: introduce ssp and enabling controls for zicfiss
target/riscv: tb flag for shadow stack instructions
target/riscv: mmu changes for zicfiss shadow stack protection
target/riscv: implement zicfiss instructions
target/riscv: compressed encodings for sspush and sspopchk
disas/riscv: enable disassembly for zicfiss instructions
disas/riscv: enable disassembly for compressed sspush/sspopchk
disas/riscv.c | 77 +++++++++-
disas/riscv.h | 4 +
target/riscv/cpu.c | 17 +++
target/riscv/cpu.h | 15 ++
target/riscv/cpu_bits.h | 17 +++
target/riscv/cpu_cfg.h | 2 +
target/riscv/cpu_helper.c | 140 +++++++++++++++++-
target/riscv/cpu_user.h | 1 +
target/riscv/csr.c | 84 +++++++++++
target/riscv/insn16.decode | 4 +
target/riscv/insn32.decode | 26 +++-
target/riscv/insn_trans/trans_rva.c.inc | 41 +++++
target/riscv/insn_trans/trans_rvi.c.inc | 55 +++++++
target/riscv/insn_trans/trans_rvzicfiss.c.inc | 73 +++++++++
target/riscv/internals.h | 3 +
target/riscv/op_helper.c | 18 +++
target/riscv/pmp.c | 5 +
target/riscv/pmp.h | 3 +-
target/riscv/tcg/tcg-cpu.c | 24 +++
target/riscv/translate.c | 30 +++-
20 files changed, 625 insertions(+), 14 deletions(-)
create mode 100644 target/riscv/insn_trans/trans_rvzicfiss.c.inc
--
2.44.0
^ permalink raw reply [flat|nested] 25+ messages in thread
* [PATCH v5 01/15] target/riscv: Add zicfilp extension
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 02/15] target/riscv: Introduce elp state and enabling controls for zicfilp Deepak Gupta
` (13 subsequent siblings)
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta
zicfilp [1] riscv cpu extension enables forward control flow integrity.
If enabled, all indirect calls must land on a landing pad instruction.
This patch sets up space for zicfilp extension in cpuconfig. zicfilp
is dependend on zicsr.
[1] - https://github.com/riscv/riscv-cfi
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
---
target/riscv/cpu.c | 2 ++
target/riscv/cpu_cfg.h | 1 +
target/riscv/tcg/tcg-cpu.c | 5 +++++
3 files changed, 8 insertions(+)
diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index 33ef4eb795..5dfb3f39ab 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -106,6 +106,7 @@ const RISCVIsaExtData isa_edata_arr[] = {
ISA_EXT_DATA_ENTRY(ziccif, PRIV_VERSION_1_11_0, has_priv_1_11),
ISA_EXT_DATA_ENTRY(zicclsm, PRIV_VERSION_1_11_0, has_priv_1_11),
ISA_EXT_DATA_ENTRY(ziccrse, PRIV_VERSION_1_11_0, has_priv_1_11),
+ ISA_EXT_DATA_ENTRY(zicfilp, PRIV_VERSION_1_12_0, ext_zicfilp),
ISA_EXT_DATA_ENTRY(zicond, PRIV_VERSION_1_12_0, ext_zicond),
ISA_EXT_DATA_ENTRY(zicntr, PRIV_VERSION_1_12_0, ext_zicntr),
ISA_EXT_DATA_ENTRY(zicsr, PRIV_VERSION_1_10_0, ext_zicsr),
@@ -1472,6 +1473,7 @@ const RISCVCPUMultiExtConfig riscv_cpu_extensions[] = {
/* Defaults for standard extensions */
MULTI_EXT_CFG_BOOL("sscofpmf", ext_sscofpmf, false),
MULTI_EXT_CFG_BOOL("zifencei", ext_zifencei, true),
+ MULTI_EXT_CFG_BOOL("zicfilp", ext_zicfilp, false),
MULTI_EXT_CFG_BOOL("zicsr", ext_zicsr, true),
MULTI_EXT_CFG_BOOL("zihintntl", ext_zihintntl, true),
MULTI_EXT_CFG_BOOL("zihintpause", ext_zihintpause, true),
diff --git a/target/riscv/cpu_cfg.h b/target/riscv/cpu_cfg.h
index 120905a254..88d5defbb5 100644
--- a/target/riscv/cpu_cfg.h
+++ b/target/riscv/cpu_cfg.h
@@ -67,6 +67,7 @@ struct RISCVCPUConfig {
bool ext_zicbom;
bool ext_zicbop;
bool ext_zicboz;
+ bool ext_zicfilp;
bool ext_zicond;
bool ext_zihintntl;
bool ext_zihintpause;
diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
index b8814ab753..ed19586c9d 100644
--- a/target/riscv/tcg/tcg-cpu.c
+++ b/target/riscv/tcg/tcg-cpu.c
@@ -623,6 +623,11 @@ void riscv_cpu_validate_set_extensions(RISCVCPU *cpu, Error **errp)
cpu->pmu_avail_ctrs = 0;
}
+ if (cpu->cfg.ext_zicfilp && !cpu->cfg.ext_zicsr) {
+ error_setg(errp, "zicfilp extension requires zicsr extension");
+ return;
+ }
+
/*
* Disable isa extensions based on priv spec after we
* validated and set everything we need.
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 02/15] target/riscv: Introduce elp state and enabling controls for zicfilp
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 01/15] target/riscv: Add zicfilp extension Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 5:17 ` Richard Henderson
2024-08-20 0:01 ` [PATCH v5 03/15] target/riscv: save and restore elp state on priv transitions Deepak Gupta
` (12 subsequent siblings)
14 siblings, 1 reply; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta,
Richard Henderson
zicfilp introduces a new state elp ("expected landing pad") in cpu.
During normal execution, elp is idle (NO_LP_EXPECTED) i.e not expecting
landing pad. On an indirect call, elp moves LP_EXPECTED. When elp is
LP_EXPECTED, only a subsquent landing pad instruction can set state back
to NO_LP_EXPECTED. On reset, elp is set to NO_LP_EXPECTED.
zicfilp is enabled via bit2 in *envcfg CSRs. Enabling control for M-mode
is in mseccfg CSR at bit position 10.
On trap, elp state is saved away in *status.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
target/riscv/cpu.c | 3 +++
target/riscv/cpu.h | 2 ++
target/riscv/cpu_bits.h | 6 ++++++
target/riscv/csr.c | 31 +++++++++++++++++++++++++++++++
target/riscv/pmp.c | 5 +++++
target/riscv/pmp.h | 3 ++-
6 files changed, 49 insertions(+), 1 deletion(-)
diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index 5dfb3f39ab..8e1f05e5b1 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -994,6 +994,9 @@ static void riscv_cpu_reset_hold(Object *obj, ResetType type)
/* mmte is supposed to have pm.current hardwired to 1 */
env->mmte |= (EXT_STATUS_INITIAL | MMTE_M_PM_CURRENT);
+ /* on reset elp is clear */
+ env->elp = false;
+
/*
* Bits 10, 6, 2 and 12 of mideleg are read only 1 when the Hypervisor
* extension is enabled.
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index 87742047ce..f966c36a31 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -222,6 +222,8 @@ struct CPUArchState {
target_ulong jvt;
+ /* elp state for zicfilp extension */
+ bool elp;
#ifdef CONFIG_USER_ONLY
uint32_t elf_flags;
#endif
diff --git a/target/riscv/cpu_bits.h b/target/riscv/cpu_bits.h
index c257c5ed7d..b05ebe6f29 100644
--- a/target/riscv/cpu_bits.h
+++ b/target/riscv/cpu_bits.h
@@ -545,6 +545,8 @@
#define MSTATUS_TVM 0x00100000 /* since: priv-1.10 */
#define MSTATUS_TW 0x00200000 /* since: priv-1.10 */
#define MSTATUS_TSR 0x00400000 /* since: priv-1.10 */
+#define MSTATUS_SPELP 0x00800000 /* zicfilp */
+#define MSTATUS_MPELP 0x020000000000 /* zicfilp */
#define MSTATUS_GVA 0x4000000000ULL
#define MSTATUS_MPV 0x8000000000ULL
@@ -575,6 +577,7 @@ typedef enum {
#define SSTATUS_XS 0x00018000
#define SSTATUS_SUM 0x00040000 /* since: priv-1.10 */
#define SSTATUS_MXR 0x00080000
+#define SSTATUS_SPELP MSTATUS_SPELP /* zicfilp */
#define SSTATUS64_UXL 0x0000000300000000ULL
@@ -747,6 +750,7 @@ typedef enum RISCVException {
/* Execution environment configuration bits */
#define MENVCFG_FIOM BIT(0)
+#define MENVCFG_LPE BIT(2) /* zicfilp */
#define MENVCFG_CBIE (3UL << 4)
#define MENVCFG_CBCFE BIT(6)
#define MENVCFG_CBZE BIT(7)
@@ -760,11 +764,13 @@ typedef enum RISCVException {
#define MENVCFGH_STCE BIT(31)
#define SENVCFG_FIOM MENVCFG_FIOM
+#define SENVCFG_LPE MENVCFG_LPE
#define SENVCFG_CBIE MENVCFG_CBIE
#define SENVCFG_CBCFE MENVCFG_CBCFE
#define SENVCFG_CBZE MENVCFG_CBZE
#define HENVCFG_FIOM MENVCFG_FIOM
+#define HENVCFG_LPE MENVCFG_LPE
#define HENVCFG_CBIE MENVCFG_CBIE
#define HENVCFG_CBCFE MENVCFG_CBCFE
#define HENVCFG_CBZE MENVCFG_CBZE
diff --git a/target/riscv/csr.c b/target/riscv/csr.c
index 432c59dc66..5771a14848 100644
--- a/target/riscv/csr.c
+++ b/target/riscv/csr.c
@@ -1400,6 +1400,11 @@ static RISCVException write_mstatus(CPURISCVState *env, int csrno,
}
}
+ /* If cfi lp extension is available, then apply cfi lp mask */
+ if (env_archcpu(env)->cfg.ext_zicfilp) {
+ mask |= (MSTATUS_MPELP | MSTATUS_SPELP);
+ }
+
mstatus = (mstatus & ~mask) | (val & mask);
env->mstatus = mstatus;
@@ -2101,6 +2106,10 @@ static RISCVException write_menvcfg(CPURISCVState *env, int csrno,
mask |= (cfg->ext_svpbmt ? MENVCFG_PBMTE : 0) |
(cfg->ext_sstc ? MENVCFG_STCE : 0) |
(cfg->ext_svadu ? MENVCFG_ADUE : 0);
+
+ if (env_archcpu(env)->cfg.ext_zicfilp) {
+ mask |= MENVCFG_LPE;
+ }
}
env->menvcfg = (env->menvcfg & ~mask) | (val & mask);
@@ -2153,6 +2162,10 @@ static RISCVException write_senvcfg(CPURISCVState *env, int csrno,
return ret;
}
+ if (env_archcpu(env)->cfg.ext_zicfilp) {
+ mask |= SENVCFG_LPE;
+ }
+
env->senvcfg = (env->senvcfg & ~mask) | (val & mask);
return RISCV_EXCP_NONE;
}
@@ -2190,6 +2203,10 @@ static RISCVException write_henvcfg(CPURISCVState *env, int csrno,
if (riscv_cpu_mxl(env) == MXL_RV64) {
mask |= env->menvcfg & (HENVCFG_PBMTE | HENVCFG_STCE | HENVCFG_ADUE);
+
+ if (env_archcpu(env)->cfg.ext_zicfilp) {
+ mask |= HENVCFG_LPE;
+ }
}
env->henvcfg = (env->henvcfg & ~mask) | (val & mask);
@@ -2654,6 +2671,10 @@ static RISCVException read_sstatus_i128(CPURISCVState *env, int csrno,
mask |= SSTATUS64_UXL;
}
+ if (env_archcpu(env)->cfg.ext_zicfilp) {
+ mask |= SSTATUS_SPELP;
+ }
+
*val = int128_make128(sstatus, add_status_sd(MXL_RV128, sstatus));
return RISCV_EXCP_NONE;
}
@@ -2665,6 +2686,11 @@ static RISCVException read_sstatus(CPURISCVState *env, int csrno,
if (env->xl != MXL_RV32 || env->debugger) {
mask |= SSTATUS64_UXL;
}
+
+ if (env_archcpu(env)->cfg.ext_zicfilp) {
+ mask |= SSTATUS_SPELP;
+ }
+
/* TODO: Use SXL not MXL. */
*val = add_status_sd(riscv_cpu_mxl(env), env->mstatus & mask);
return RISCV_EXCP_NONE;
@@ -2680,6 +2706,11 @@ static RISCVException write_sstatus(CPURISCVState *env, int csrno,
mask |= SSTATUS64_UXL;
}
}
+
+ if (env_archcpu(env)->cfg.ext_zicfilp) {
+ mask |= SSTATUS_SPELP;
+ }
+
target_ulong newval = (env->mstatus & ~mask) | (val & mask);
return write_mstatus(env, CSR_MSTATUS, newval);
}
diff --git a/target/riscv/pmp.c b/target/riscv/pmp.c
index 9eea397e72..1111d08d08 100644
--- a/target/riscv/pmp.c
+++ b/target/riscv/pmp.c
@@ -598,6 +598,11 @@ void mseccfg_csr_write(CPURISCVState *env, target_ulong val)
val &= ~(MSECCFG_MMWP | MSECCFG_MML | MSECCFG_RLB);
}
+ /* M-mode forward cfi to be enabled if cfi extension is implemented */
+ if (env_archcpu(env)->cfg.ext_zicfilp) {
+ val |= (val & MSECCFG_MLPE);
+ }
+
env->mseccfg = val;
}
diff --git a/target/riscv/pmp.h b/target/riscv/pmp.h
index f5c10ce85c..e0530a17a3 100644
--- a/target/riscv/pmp.h
+++ b/target/riscv/pmp.h
@@ -44,7 +44,8 @@ typedef enum {
MSECCFG_MMWP = 1 << 1,
MSECCFG_RLB = 1 << 2,
MSECCFG_USEED = 1 << 8,
- MSECCFG_SSEED = 1 << 9
+ MSECCFG_SSEED = 1 << 9,
+ MSECCFG_MLPE = 1 << 10,
} mseccfg_field_t;
typedef struct {
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 03/15] target/riscv: save and restore elp state on priv transitions
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 01/15] target/riscv: Add zicfilp extension Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 02/15] target/riscv: Introduce elp state and enabling controls for zicfilp Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 04/15] target/riscv: additional code information for sw check Deepak Gupta
` (11 subsequent siblings)
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta,
Richard Henderson
elp state is recorded in *status on trap entry (less privilege to higher
privilege) and restored in elp from *status on trap exit (higher to less
privilege).
Additionally this patch introduces a forward cfi helper function to
determine if current privilege has forward cfi is enabled or not based on
*envcfg (for U, VU, S, VU, HS) or mseccfg csr (for M). For qemu-user, a
new field `ufcfien` is introduced which is by default set to false and
helper function returns value deposited in `ufcfien` for qemu-user.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
target/riscv/cpu.c | 5 ++++
target/riscv/cpu.h | 2 ++
target/riscv/cpu_helper.c | 55 +++++++++++++++++++++++++++++++++++++++
target/riscv/op_helper.c | 18 +++++++++++++
4 files changed, 80 insertions(+)
diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index 8e1f05e5b1..083d405516 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -1022,6 +1022,11 @@ static void riscv_cpu_reset_hold(Object *obj, ResetType type)
env->load_res = -1;
set_default_nan_mode(1, &env->fp_status);
+#ifdef CONFIG_USER_ONLY
+ /* qemu-user for riscv, fcfi is off by default */
+ env->ufcfien = false;
+#endif
+
#ifndef CONFIG_USER_ONLY
if (cpu->cfg.debug) {
riscv_trigger_reset_hold(env);
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index f966c36a31..7be0fa30f7 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -226,6 +226,7 @@ struct CPUArchState {
bool elp;
#ifdef CONFIG_USER_ONLY
uint32_t elf_flags;
+ bool ufcfien;
#endif
#ifndef CONFIG_USER_ONLY
@@ -530,6 +531,7 @@ void riscv_cpu_set_geilen(CPURISCVState *env, target_ulong geilen);
bool riscv_cpu_vector_enabled(CPURISCVState *env);
void riscv_cpu_set_virt_enabled(CPURISCVState *env, bool enable);
int riscv_env_mmu_index(CPURISCVState *env, bool ifetch);
+bool cpu_get_fcfien(CPURISCVState *env);
G_NORETURN void riscv_cpu_do_unaligned_access(CPUState *cs, vaddr addr,
MMUAccessType access_type,
int mmu_idx, uintptr_t retaddr);
diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index 6709622dd3..12484ca7d2 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -33,6 +33,7 @@
#include "cpu_bits.h"
#include "debug.h"
#include "tcg/oversized-guest.h"
+#include "pmp.h"
int riscv_env_mmu_index(CPURISCVState *env, bool ifetch)
{
@@ -63,6 +64,34 @@ int riscv_env_mmu_index(CPURISCVState *env, bool ifetch)
#endif
}
+bool cpu_get_fcfien(CPURISCVState *env)
+{
+ /* no cfi extension, return false */
+ if (!env_archcpu(env)->cfg.ext_zicfilp) {
+ return false;
+ }
+#ifdef CONFIG_USER_ONLY
+ return env->ufcfien;
+#else
+ switch (env->priv) {
+ case PRV_U:
+ if (riscv_has_ext(env, RVS)) {
+ return env->senvcfg & MENVCFG_LPE;
+ }
+ return env->menvcfg & MENVCFG_LPE;
+ case PRV_S:
+ if (env->virt_enabled) {
+ return env->henvcfg & HENVCFG_LPE;
+ }
+ return env->menvcfg & MENVCFG_LPE;
+ case PRV_M:
+ return env->mseccfg & MSECCFG_MLPE;
+ default:
+ g_assert_not_reached();
+ }
+#endif
+}
+
void cpu_get_tb_cpu_state(CPURISCVState *env, vaddr *pc,
uint64_t *cs_base, uint32_t *pflags)
{
@@ -546,6 +575,15 @@ void riscv_cpu_swap_hypervisor_regs(CPURISCVState *env)
}
bool current_virt = env->virt_enabled;
+ /*
+ * If zicfilp extension available and henvcfg.LPE = 1,
+ * then apply SPELP mask on mstatus
+ */
+ if (env_archcpu(env)->cfg.ext_zicfilp &&
+ get_field(env->henvcfg, HENVCFG_LPE)) {
+ mstatus_mask |= SSTATUS_SPELP;
+ }
+
g_assert(riscv_has_ext(env, RVH));
if (current_virt) {
@@ -1754,6 +1792,11 @@ void riscv_cpu_do_interrupt(CPUState *cs)
if (env->priv <= PRV_S && cause < 64 &&
(((deleg >> cause) & 1) || s_injected || vs_injected)) {
/* handle the trap in S-mode */
+ /* save elp status */
+ if (cpu_get_fcfien(env)) {
+ env->mstatus = set_field(env->mstatus, MSTATUS_SPELP, env->elp);
+ }
+
if (riscv_has_ext(env, RVH)) {
uint64_t hdeleg = async ? env->hideleg : env->hedeleg;
@@ -1802,6 +1845,11 @@ void riscv_cpu_do_interrupt(CPUState *cs)
riscv_cpu_set_mode(env, PRV_S);
} else {
/* handle the trap in M-mode */
+ /* save elp status */
+ if (cpu_get_fcfien(env)) {
+ env->mstatus = set_field(env->mstatus, MSTATUS_MPELP, env->elp);
+ }
+
if (riscv_has_ext(env, RVH)) {
if (env->virt_enabled) {
riscv_cpu_swap_hypervisor_regs(env);
@@ -1833,6 +1881,13 @@ void riscv_cpu_do_interrupt(CPUState *cs)
riscv_cpu_set_mode(env, PRV_M);
}
+ /*
+ * Interrupt/exception/trap delivery is asynchronous event and as per
+ * zicfilp spec CPU should clear up the ELP state. No harm in clearing
+ * unconditionally.
+ */
+ env->elp = false;
+
/*
* NOTE: it is not necessary to yield load reservations here. It is only
* necessary for an SC from "another hart" to cause a load reservation
diff --git a/target/riscv/op_helper.c b/target/riscv/op_helper.c
index 2baf5bc3ca..5848aaf437 100644
--- a/target/riscv/op_helper.c
+++ b/target/riscv/op_helper.c
@@ -313,6 +313,15 @@ target_ulong helper_sret(CPURISCVState *env)
riscv_cpu_set_mode(env, prev_priv);
+ /*
+ * If forward cfi enabled for new priv, restore elp status
+ * and clear spelp in mstatus
+ */
+ if (cpu_get_fcfien(env)) {
+ env->elp = get_field(env->mstatus, MSTATUS_SPELP);
+ }
+ env->mstatus = set_field(env->mstatus, MSTATUS_SPELP, 0);
+
return retpc;
}
@@ -357,6 +366,15 @@ target_ulong helper_mret(CPURISCVState *env)
riscv_cpu_set_virt_enabled(env, prev_virt);
}
+ /*
+ * If forward cfi enabled for new priv, restore elp status
+ * and clear mpelp in mstatus
+ */
+ if (cpu_get_fcfien(env)) {
+ env->elp = get_field(env->mstatus, MSTATUS_MPELP);
+ }
+ env->mstatus = set_field(env->mstatus, MSTATUS_MPELP, 0);
+
return retpc;
}
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 04/15] target/riscv: additional code information for sw check
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (2 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 03/15] target/riscv: save and restore elp state on priv transitions Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 05/15] target/riscv: tracking indirect branches (fcfi) for zicfilp Deepak Gupta
` (10 subsequent siblings)
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta,
Richard Henderson
sw check exception support was recently added. This patch further augments
sw check exception by providing support for additional code which is
provided in *tval. Adds `sw_check_code` field in cpuarchstate. Whenever
sw check exception is raised *tval gets the value deposited in
`sw_check_code`.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
target/riscv/cpu.h | 2 ++
target/riscv/cpu_helper.c | 2 ++
target/riscv/csr.c | 1 +
3 files changed, 5 insertions(+)
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index 7be0fa30f7..11c6513a90 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -224,6 +224,8 @@ struct CPUArchState {
/* elp state for zicfilp extension */
bool elp;
+ /* sw check code for sw check exception */
+ target_ulong sw_check_code;
#ifdef CONFIG_USER_ONLY
uint32_t elf_flags;
bool ufcfien;
diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index 12484ca7d2..9f08a67a9e 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -1761,6 +1761,8 @@ void riscv_cpu_do_interrupt(CPUState *cs)
cs->watchpoint_hit = NULL;
}
break;
+ case RISCV_EXCP_SW_CHECK:
+ tval = env->sw_check_code;
default:
break;
}
diff --git a/target/riscv/csr.c b/target/riscv/csr.c
index 5771a14848..a5a969a377 100644
--- a/target/riscv/csr.c
+++ b/target/riscv/csr.c
@@ -1179,6 +1179,7 @@ static const uint64_t all_ints = M_MODE_INTERRUPTS | S_MODE_INTERRUPTS |
(1ULL << (RISCV_EXCP_INST_PAGE_FAULT)) | \
(1ULL << (RISCV_EXCP_LOAD_PAGE_FAULT)) | \
(1ULL << (RISCV_EXCP_STORE_PAGE_FAULT)) | \
+ (1ULL << (RISCV_EXCP_SW_CHECK)) | \
(1ULL << (RISCV_EXCP_INST_GUEST_PAGE_FAULT)) | \
(1ULL << (RISCV_EXCP_LOAD_GUEST_ACCESS_FAULT)) | \
(1ULL << (RISCV_EXCP_VIRT_INSTRUCTION_FAULT)) | \
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 05/15] target/riscv: tracking indirect branches (fcfi) for zicfilp
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (3 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 04/15] target/riscv: additional code information for sw check Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 5:24 ` Richard Henderson
2024-08-20 0:01 ` [PATCH v5 06/15] target/riscv: zicfilp `lpad` impl and branch tracking Deepak Gupta
` (9 subsequent siblings)
14 siblings, 1 reply; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta,
Richard Henderson
zicfilp protects forward control flow (if enabled) by enforcing all
indirect call and jmp must land on a landing pad instruction `lpad`. If
target of an indirect call or jmp is not `lpad` then cpu/hart must raise
a sw check exception with tval = 2.
This patch implements the mechanism using TCG. Target architecture branch
instruction must define the end of a TB. Using this property, during
translation of branch instruction, TB flag = FCFI_LP_EXPECTED can be set.
Translation of target TB can check if FCFI_LP_EXPECTED flag is set and a
flag (fcfi_lp_expected) can be set in DisasContext. If `lpad` gets
translated, fcfi_lp_expected flag in DisasContext can be cleared. Else
it'll fault.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
---
target/riscv/cpu.h | 3 +++
target/riscv/cpu_bits.h | 3 +++
target/riscv/cpu_helper.c | 12 ++++++++++++
target/riscv/translate.c | 21 ++++++++++++++++++++-
4 files changed, 38 insertions(+), 1 deletion(-)
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index 11c6513a90..edf540339a 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -606,6 +606,9 @@ FIELD(TB_FLAGS, ITRIGGER, 22, 1)
FIELD(TB_FLAGS, VIRT_ENABLED, 23, 1)
FIELD(TB_FLAGS, PRIV, 24, 2)
FIELD(TB_FLAGS, AXL, 26, 2)
+/* zicfilp needs a TB flag to track indirect branches */
+FIELD(TB_FLAGS, FCFI_ENABLED, 28, 1)
+FIELD(TB_FLAGS, FCFI_LP_EXPECTED, 29, 1)
#ifdef TARGET_RISCV32
#define riscv_cpu_mxl(env) ((void)(env), MXL_RV32)
diff --git a/target/riscv/cpu_bits.h b/target/riscv/cpu_bits.h
index b05ebe6f29..900769ce60 100644
--- a/target/riscv/cpu_bits.h
+++ b/target/riscv/cpu_bits.h
@@ -685,6 +685,9 @@ typedef enum RISCVException {
RISCV_EXCP_SEMIHOST = 0x3f,
} RISCVException;
+/* zicfilp defines lp violation results in sw check with tval = 2*/
+#define RISCV_EXCP_SW_CHECK_FCFI_TVAL 2
+
#define RISCV_EXCP_INT_FLAG 0x80000000
#define RISCV_EXCP_INT_MASK 0x7fffffff
diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index 9f08a67a9e..3a56bea8b9 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -133,6 +133,18 @@ void cpu_get_tb_cpu_state(CPURISCVState *env, vaddr *pc,
flags = FIELD_DP32(flags, TB_FLAGS, VILL, 1);
}
+ if (cpu_get_fcfien(env)) {
+ /*
+ * For Forward CFI, only the expectation of a lpcll at
+ * the start of the block is tracked (which can only happen
+ * when FCFI is enabled for the current processor mode). A jump
+ * or call at the end of the previous TB will have updated
+ * env->elp to indicate the expectation.
+ */
+ flags = FIELD_DP32(flags, TB_FLAGS, FCFI_LP_EXPECTED, env->elp);
+ flags = FIELD_DP32(flags, TB_FLAGS, FCFI_ENABLED, 1);
+ }
+
#ifdef CONFIG_USER_ONLY
fs = EXT_STATUS_DIRTY;
vs = EXT_STATUS_DIRTY;
diff --git a/target/riscv/translate.c b/target/riscv/translate.c
index acba90f170..6c1c868a4a 100644
--- a/target/riscv/translate.c
+++ b/target/riscv/translate.c
@@ -116,6 +116,9 @@ typedef struct DisasContext {
bool frm_valid;
bool insn_start_updated;
const GPtrArray *decoders;
+ /* zicfilp extension. fcfi_enabled, lp expected or not */
+ bool fcfi_enabled;
+ bool fcfi_lp_expected;
} DisasContext;
static inline bool has_ext(DisasContext *ctx, uint32_t ext)
@@ -1238,6 +1241,8 @@ static void riscv_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs)
ctx->pm_base_enabled = FIELD_EX32(tb_flags, TB_FLAGS, PM_BASE_ENABLED);
ctx->ztso = cpu->cfg.ext_ztso;
ctx->itrigger = FIELD_EX32(tb_flags, TB_FLAGS, ITRIGGER);
+ ctx->fcfi_lp_expected = FIELD_EX32(tb_flags, TB_FLAGS, FCFI_LP_EXPECTED);
+ ctx->fcfi_enabled = FIELD_EX32(tb_flags, TB_FLAGS, FCFI_ENABLED);
ctx->zero = tcg_constant_tl(0);
ctx->virt_inst_excp = false;
ctx->decoders = cpu->decoders;
@@ -1265,11 +1270,25 @@ static void riscv_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
DisasContext *ctx = container_of(dcbase, DisasContext, base);
CPURISCVState *env = cpu_env(cpu);
uint16_t opcode16 = translator_lduw(env, &ctx->base, ctx->base.pc_next);
-
ctx->ol = ctx->xl;
decode_opc(env, ctx, opcode16);
ctx->base.pc_next += ctx->cur_insn_len;
+ /*
+ * instr decode of `lpad` should have lowered fcfi_lp_expected.
+ * If not `lpad`, raise sw check exception.
+ */
+ if (ctx->fcfi_lp_expected) {
+ /* Emit after insn_start, i.e. before the op following insn_start. */
+ tcg_ctx->emit_before_op = QTAILQ_NEXT(ctx->base.insn_start, link);
+ tcg_gen_st_tl(tcg_constant_tl(RISCV_EXCP_SW_CHECK_FCFI_TVAL),
+ tcg_env, offsetof(CPURISCVState, sw_check_code));
+ gen_helper_raise_exception(tcg_env,
+ tcg_constant_i32(RISCV_EXCP_SW_CHECK));
+ tcg_ctx->emit_before_op = NULL;
+ ctx->base.is_jmp = DISAS_NORETURN;
+ }
+
/* Only the first insn within a TB is allowed to cross a page boundary. */
if (ctx->base.is_jmp == DISAS_NEXT) {
if (ctx->itrigger || !is_same_page(&ctx->base, ctx->base.pc_next)) {
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 06/15] target/riscv: zicfilp `lpad` impl and branch tracking
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (4 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 05/15] target/riscv: tracking indirect branches (fcfi) for zicfilp Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 5:29 ` Richard Henderson
2024-08-20 0:01 ` [PATCH v5 07/15] disas/riscv: enable `lpad` disassembly Deepak Gupta
` (8 subsequent siblings)
14 siblings, 1 reply; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta
Implements setting lp expected when `jalr` is encountered and implements
`lpad` instruction of zicfilp. `lpad` instruction is taken out of
auipc x0, <imm_20>. This is an existing HINTNOP space. If `lpad` is
target of an indirect branch, cpu checks for 20 bit value in x7 upper
with 20 bit value embedded in `lpad`. If they don't match, cpu raises a
sw check exception with tval = 2.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
---
target/riscv/cpu_user.h | 1 +
target/riscv/insn32.decode | 5 ++-
target/riscv/insn_trans/trans_rvi.c.inc | 55 +++++++++++++++++++++++++
3 files changed, 60 insertions(+), 1 deletion(-)
diff --git a/target/riscv/cpu_user.h b/target/riscv/cpu_user.h
index 02afad608b..e6927ff847 100644
--- a/target/riscv/cpu_user.h
+++ b/target/riscv/cpu_user.h
@@ -15,5 +15,6 @@
#define xA6 16
#define xA7 17 /* syscall number for RVI ABI */
#define xT0 5 /* syscall number for RVE ABI */
+#define xT2 7
#endif
diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode
index c45b8fa1d8..6533cb0758 100644
--- a/target/riscv/insn32.decode
+++ b/target/riscv/insn32.decode
@@ -123,7 +123,10 @@ sfence_vm 0001000 00100 ..... 000 00000 1110011 @sfence_vm
# *** RV32I Base Instruction Set ***
lui .................... ..... 0110111 @u
-auipc .................... ..... 0010111 @u
+{
+ lpad label:20 00000 0010111
+ auipc .................... ..... 0010111 @u
+}
jal .................... ..... 1101111 @j
jalr ............ ..... 000 ..... 1100111 @i
beq ....... ..... ..... 000 ..... 1100011 @b
diff --git a/target/riscv/insn_trans/trans_rvi.c.inc b/target/riscv/insn_trans/trans_rvi.c.inc
index 98e3806d5e..b427f3a939 100644
--- a/target/riscv/insn_trans/trans_rvi.c.inc
+++ b/target/riscv/insn_trans/trans_rvi.c.inc
@@ -36,6 +36,49 @@ static bool trans_lui(DisasContext *ctx, arg_lui *a)
return true;
}
+static bool trans_lpad(DisasContext *ctx, arg_lpad *a)
+{
+ /*
+ * fcfi_lp_expected can set only if fcfi was eanbled.
+ * translate further only if fcfi_lp_expected set.
+ * lpad comes from NOP space anyways, so return true if
+ * fcfi_lp_expected is false.
+ */
+ if (!ctx->fcfi_lp_expected) {
+ return true;
+ }
+
+ ctx->fcfi_lp_expected = false;
+ if ((ctx->base.pc_next) & 0x3) {
+ /*
+ * misaligned, according to spec we should raise sw check exception
+ */
+ tcg_gen_st_tl(tcg_constant_tl(RISCV_EXCP_SW_CHECK_FCFI_TVAL),
+ tcg_env, offsetof(CPURISCVState, sw_check_code));
+ gen_helper_raise_exception(tcg_env,
+ tcg_constant_i32(RISCV_EXCP_SW_CHECK));
+ return true;
+ }
+
+ /* per spec, label check performed only when embedded label non-zero */
+ if (a->label != 0) {
+ TCGLabel *skip = gen_new_label();
+ TCGv tmp = tcg_temp_new();
+ tcg_gen_extract_tl(tmp, get_gpr(ctx, xT2, EXT_NONE), 12, 20);
+ tcg_gen_brcondi_tl(TCG_COND_EQ, tmp, a->label, skip);
+ tcg_gen_st_tl(tcg_constant_tl(RISCV_EXCP_SW_CHECK_FCFI_TVAL),
+ tcg_env, offsetof(CPURISCVState, sw_check_code));
+ gen_helper_raise_exception(tcg_env,
+ tcg_constant_i32(RISCV_EXCP_SW_CHECK));
+ gen_set_label(skip);
+ }
+
+ tcg_gen_st8_tl(tcg_constant_tl(0), tcg_env,
+ offsetof(CPURISCVState, elp));
+
+ return true;
+}
+
static bool trans_auipc(DisasContext *ctx, arg_auipc *a)
{
TCGv target_pc = dest_gpr(ctx, a->rd);
@@ -75,6 +118,18 @@ static bool trans_jalr(DisasContext *ctx, arg_jalr *a)
gen_set_gpr(ctx, a->rd, succ_pc);
tcg_gen_mov_tl(cpu_pc, target_pc);
+ if (ctx->fcfi_enabled) {
+ /*
+ * return from functions (i.e. rs1 == xRA || rs1 == xT0) are not
+ * tracked. zicfilp introduces sw guarded branch as well. sw guarded
+ * branch are not tracked. rs1 == xT2 is a sw guarded branch.
+ */
+ if (a->rs1 != xRA && a->rs1 != xT0 && a->rs1 != xT2) {
+ tcg_gen_st8_tl(tcg_constant_tl(1),
+ tcg_env, offsetof(CPURISCVState, elp));
+ }
+ }
+
lookup_and_goto_ptr(ctx);
if (misaligned) {
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 07/15] disas/riscv: enable `lpad` disassembly
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (5 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 06/15] target/riscv: zicfilp `lpad` impl and branch tracking Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 08/15] target/riscv: Add zicfiss extension Deepak Gupta
` (7 subsequent siblings)
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta,
Richard Henderson
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
disas/riscv.c | 18 +++++++++++++++++-
disas/riscv.h | 2 ++
2 files changed, 19 insertions(+), 1 deletion(-)
diff --git a/disas/riscv.c b/disas/riscv.c
index c8364c2b07..c7c92acef7 100644
--- a/disas/riscv.c
+++ b/disas/riscv.c
@@ -974,6 +974,7 @@ typedef enum {
rv_op_amomaxu_h = 943,
rv_op_amocas_b = 944,
rv_op_amocas_h = 945,
+ rv_op_lpad = 946,
} rv_op;
/* register names */
@@ -2232,6 +2233,7 @@ const rv_opcode_data rvi_opcode_data[] = {
{ "amomaxu.h", rv_codec_r_a, rv_fmt_aqrl_rd_rs2_rs1, NULL, 0, 0, 0 },
{ "amocas.b", rv_codec_r_a, rv_fmt_aqrl_rd_rs2_rs1, NULL, 0, 0, 0 },
{ "amocas.h", rv_codec_r_a, rv_fmt_aqrl_rd_rs2_rs1, NULL, 0, 0, 0 },
+ { "lpad", rv_codec_lp, rv_fmt_imm, NULL, 0, 0, 0 },
};
/* CSR names */
@@ -2925,7 +2927,13 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
case 7: op = rv_op_andi; break;
}
break;
- case 5: op = rv_op_auipc; break;
+ case 5:
+ op = rv_op_auipc;
+ if (dec->cfg->ext_zicfilp &&
+ (((inst >> 7) & 0b11111) == 0b00000)) {
+ op = rv_op_lpad;
+ }
+ break;
case 6:
switch ((inst >> 12) & 0b111) {
case 0: op = rv_op_addiw; break;
@@ -4482,6 +4490,11 @@ static uint32_t operand_tbl_index(rv_inst inst)
return ((inst << 54) >> 56);
}
+static uint32_t operand_lpl(rv_inst inst)
+{
+ return inst >> 12;
+}
+
/* decode operands */
static void decode_inst_operands(rv_decode *dec, rv_isa isa)
@@ -4869,6 +4882,9 @@ static void decode_inst_operands(rv_decode *dec, rv_isa isa)
dec->imm = sextract32(operand_rs2(inst), 0, 5);
dec->imm1 = operand_imm2(inst);
break;
+ case rv_codec_lp:
+ dec->imm = operand_lpl(inst);
+ break;
};
}
diff --git a/disas/riscv.h b/disas/riscv.h
index 16a08e4895..1182457aff 100644
--- a/disas/riscv.h
+++ b/disas/riscv.h
@@ -166,6 +166,7 @@ typedef enum {
rv_codec_r2_immhl,
rv_codec_r2_imm2_imm5,
rv_codec_fli,
+ rv_codec_lp,
} rv_codec;
/* structures */
@@ -228,6 +229,7 @@ enum {
#define rv_fmt_rs1_rs2 "O\t1,2"
#define rv_fmt_rd_imm "O\t0,i"
#define rv_fmt_rd_uimm "O\t0,Ui"
+#define rv_fmt_imm "O\ti"
#define rv_fmt_rd_offset "O\t0,o"
#define rv_fmt_rd_uoffset "O\t0,Uo"
#define rv_fmt_rd_rs1_rs2 "O\t0,1,2"
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 08/15] target/riscv: Add zicfiss extension
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (6 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 07/15] disas/riscv: enable `lpad` disassembly Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 09/15] target/riscv: introduce ssp and enabling controls for zicfiss Deepak Gupta
` (6 subsequent siblings)
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta
zicfiss [1] riscv cpu extension enables backward control flow integrity.
This patch sets up space for zicfiss extension in cpuconfig. And imple-
ments dependency on A, zicsr, zimop and zcmop extensions.
[1] - https://github.com/riscv/riscv-cfi
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
---
target/riscv/cpu.c | 2 ++
target/riscv/cpu_cfg.h | 1 +
target/riscv/tcg/tcg-cpu.c | 19 +++++++++++++++++++
3 files changed, 22 insertions(+)
diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index 083d405516..10a2a32345 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -107,6 +107,7 @@ const RISCVIsaExtData isa_edata_arr[] = {
ISA_EXT_DATA_ENTRY(zicclsm, PRIV_VERSION_1_11_0, has_priv_1_11),
ISA_EXT_DATA_ENTRY(ziccrse, PRIV_VERSION_1_11_0, has_priv_1_11),
ISA_EXT_DATA_ENTRY(zicfilp, PRIV_VERSION_1_12_0, ext_zicfilp),
+ ISA_EXT_DATA_ENTRY(zicfiss, PRIV_VERSION_1_13_0, ext_zicfiss),
ISA_EXT_DATA_ENTRY(zicond, PRIV_VERSION_1_12_0, ext_zicond),
ISA_EXT_DATA_ENTRY(zicntr, PRIV_VERSION_1_12_0, ext_zicntr),
ISA_EXT_DATA_ENTRY(zicsr, PRIV_VERSION_1_10_0, ext_zicsr),
@@ -1482,6 +1483,7 @@ const RISCVCPUMultiExtConfig riscv_cpu_extensions[] = {
MULTI_EXT_CFG_BOOL("sscofpmf", ext_sscofpmf, false),
MULTI_EXT_CFG_BOOL("zifencei", ext_zifencei, true),
MULTI_EXT_CFG_BOOL("zicfilp", ext_zicfilp, false),
+ MULTI_EXT_CFG_BOOL("zicfiss", ext_zicfiss, false),
MULTI_EXT_CFG_BOOL("zicsr", ext_zicsr, true),
MULTI_EXT_CFG_BOOL("zihintntl", ext_zihintntl, true),
MULTI_EXT_CFG_BOOL("zihintpause", ext_zihintpause, true),
diff --git a/target/riscv/cpu_cfg.h b/target/riscv/cpu_cfg.h
index 88d5defbb5..2499f38407 100644
--- a/target/riscv/cpu_cfg.h
+++ b/target/riscv/cpu_cfg.h
@@ -68,6 +68,7 @@ struct RISCVCPUConfig {
bool ext_zicbop;
bool ext_zicboz;
bool ext_zicfilp;
+ bool ext_zicfiss;
bool ext_zicond;
bool ext_zihintntl;
bool ext_zihintpause;
diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
index ed19586c9d..4da26cb926 100644
--- a/target/riscv/tcg/tcg-cpu.c
+++ b/target/riscv/tcg/tcg-cpu.c
@@ -618,6 +618,25 @@ void riscv_cpu_validate_set_extensions(RISCVCPU *cpu, Error **errp)
cpu->cfg.ext_zihpm = false;
}
+ if (cpu->cfg.ext_zicfiss) {
+ if (!cpu->cfg.ext_zicsr) {
+ error_setg(errp, "zicfiss extension requires zicsr extension");
+ return;
+ }
+ if (!riscv_has_ext(env, RVA)) {
+ error_setg(errp, "zicfiss extension requires A extension");
+ return;
+ }
+ if (!cpu->cfg.ext_zimop) {
+ error_setg(errp, "zicfiss extension requires zimop extension");
+ return;
+ }
+ if (cpu->cfg.ext_zca && !cpu->cfg.ext_zcmop) {
+ error_setg(errp, "zicfiss with zca requires zcmop extension");
+ return;
+ }
+ }
+
if (!cpu->cfg.ext_zihpm) {
cpu->cfg.pmu_mask = 0;
cpu->pmu_avail_ctrs = 0;
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 09/15] target/riscv: introduce ssp and enabling controls for zicfiss
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (7 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 08/15] target/riscv: Add zicfiss extension Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 5:34 ` Richard Henderson
2024-08-20 0:01 ` [PATCH v5 10/15] target/riscv: tb flag for shadow stack instructions Deepak Gupta
` (5 subsequent siblings)
14 siblings, 1 reply; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta
zicfiss introduces a new state ssp ("shadow stack register") in cpu.
ssp is expressed as a new unprivileged csr (CSR_SSP=0x11) and holds
virtual address for shadow stack as programmed by software.
Shadow stack (for each mode) is enabled via bit3 in *envcfg CSRs.
Shadow stack can be enabled for a mode only if it's higher privileged
mode had it enabled for itself. M mode doesn't need enabling control,
it's always available if extension is available on cpu.
This patch also implements helper bcfi function which determines if bcfi
is enabled at current privilege or not. qemu-user also gets field
`ubcfien` indicating whether qemu user has shadow stack enabled or not.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
---
target/riscv/cpu.c | 5 ++++
target/riscv/cpu.h | 4 +++
target/riscv/cpu_bits.h | 6 +++++
target/riscv/cpu_helper.c | 25 +++++++++++++++++++
target/riscv/csr.c | 52 +++++++++++++++++++++++++++++++++++++++
5 files changed, 92 insertions(+)
diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index 10a2a32345..76f1edd15a 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -998,6 +998,9 @@ static void riscv_cpu_reset_hold(Object *obj, ResetType type)
/* on reset elp is clear */
env->elp = false;
+ /* on reset ssp is set to 0 */
+ env->ssp = 0;
+
/*
* Bits 10, 6, 2 and 12 of mideleg are read only 1 when the Hypervisor
* extension is enabled.
@@ -1026,6 +1029,8 @@ static void riscv_cpu_reset_hold(Object *obj, ResetType type)
#ifdef CONFIG_USER_ONLY
/* qemu-user for riscv, fcfi is off by default */
env->ufcfien = false;
+ /* qemu-user for riscv, bcfi is off by default */
+ env->ubcfien = false;
#endif
#ifndef CONFIG_USER_ONLY
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index edf540339a..5a57099d59 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -224,11 +224,14 @@ struct CPUArchState {
/* elp state for zicfilp extension */
bool elp;
+ /* shadow stack register for zicfiss extension */
+ target_ulong ssp;
/* sw check code for sw check exception */
target_ulong sw_check_code;
#ifdef CONFIG_USER_ONLY
uint32_t elf_flags;
bool ufcfien;
+ bool ubcfien;
#endif
#ifndef CONFIG_USER_ONLY
@@ -534,6 +537,7 @@ bool riscv_cpu_vector_enabled(CPURISCVState *env);
void riscv_cpu_set_virt_enabled(CPURISCVState *env, bool enable);
int riscv_env_mmu_index(CPURISCVState *env, bool ifetch);
bool cpu_get_fcfien(CPURISCVState *env);
+bool cpu_get_bcfien(CPURISCVState *env);
G_NORETURN void riscv_cpu_do_unaligned_access(CPUState *cs, vaddr addr,
MMUAccessType access_type,
int mmu_idx, uintptr_t retaddr);
diff --git a/target/riscv/cpu_bits.h b/target/riscv/cpu_bits.h
index 900769ce60..48ce24dc32 100644
--- a/target/riscv/cpu_bits.h
+++ b/target/riscv/cpu_bits.h
@@ -34,6 +34,9 @@
/* Control and Status Registers */
+/* zicfiss user ssp csr */
+#define CSR_SSP 0x011
+
/* User Trap Setup */
#define CSR_USTATUS 0x000
#define CSR_UIE 0x004
@@ -754,6 +757,7 @@ typedef enum RISCVException {
/* Execution environment configuration bits */
#define MENVCFG_FIOM BIT(0)
#define MENVCFG_LPE BIT(2) /* zicfilp */
+#define MENVCFG_SSE BIT(3) /* zicfiss */
#define MENVCFG_CBIE (3UL << 4)
#define MENVCFG_CBCFE BIT(6)
#define MENVCFG_CBZE BIT(7)
@@ -768,12 +772,14 @@ typedef enum RISCVException {
#define SENVCFG_FIOM MENVCFG_FIOM
#define SENVCFG_LPE MENVCFG_LPE
+#define SENVCFG_SSE MENVCFG_SSE
#define SENVCFG_CBIE MENVCFG_CBIE
#define SENVCFG_CBCFE MENVCFG_CBCFE
#define SENVCFG_CBZE MENVCFG_CBZE
#define HENVCFG_FIOM MENVCFG_FIOM
#define HENVCFG_LPE MENVCFG_LPE
+#define HENVCFG_SSE MENVCFG_SSE
#define HENVCFG_CBIE MENVCFG_CBIE
#define HENVCFG_CBCFE MENVCFG_CBCFE
#define HENVCFG_CBZE MENVCFG_CBZE
diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index 3a56bea8b9..96571177ef 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -92,6 +92,31 @@ bool cpu_get_fcfien(CPURISCVState *env)
#endif
}
+bool cpu_get_bcfien(CPURISCVState *env)
+{
+ /* no cfi extension, return false */
+ if (!env_archcpu(env)->cfg.ext_zicfiss) {
+ return false;
+ }
+#ifdef CONFIG_USER_ONLY
+ return env->ubcfien;
+#else
+ switch (env->priv) {
+ case PRV_U:
+ return env->senvcfg & SENVCFG_SSE;
+ case PRV_S:
+ if (env->virt_enabled) {
+ return env->henvcfg & HENVCFG_SSE;
+ }
+ return env->menvcfg & MENVCFG_SSE;
+ case PRV_M: /* M-mode shadow stack is always on if hart implements */
+ return true;
+ default:
+ g_assert_not_reached();
+ }
+#endif
+}
+
void cpu_get_tb_cpu_state(CPURISCVState *env, vaddr *pc,
uint64_t *cs_base, uint32_t *pflags)
{
diff --git a/target/riscv/csr.c b/target/riscv/csr.c
index a5a969a377..ec04b2b32b 100644
--- a/target/riscv/csr.c
+++ b/target/riscv/csr.c
@@ -185,6 +185,25 @@ static RISCVException zcmt(CPURISCVState *env, int csrno)
return RISCV_EXCP_NONE;
}
+static RISCVException cfi_ss(CPURISCVState *env, int csrno)
+{
+ if (!env_archcpu(env)->cfg.ext_zicfiss) {
+ return RISCV_EXCP_ILLEGAL_INST;
+ }
+
+ /* if bcfi not active for current env, access to csr is illegal */
+ if (!cpu_get_bcfien(env)) {
+#if !defined(CONFIG_USER_ONLY)
+ if (env->debugger) {
+ return RISCV_EXCP_NONE;
+ }
+#endif
+ return RISCV_EXCP_ILLEGAL_INST;
+ }
+
+ return RISCV_EXCP_NONE;
+}
+
#if !defined(CONFIG_USER_ONLY)
static RISCVException mctr(CPURISCVState *env, int csrno)
{
@@ -596,6 +615,19 @@ static RISCVException seed(CPURISCVState *env, int csrno)
#endif
}
+/* zicfiss CSR_SSP read and write */
+static int read_ssp(CPURISCVState *env, int csrno, target_ulong *val)
+{
+ *val = env->ssp;
+ return RISCV_EXCP_NONE;
+}
+
+static int write_ssp(CPURISCVState *env, int csrno, target_ulong val)
+{
+ env->ssp = val;
+ return RISCV_EXCP_NONE;
+}
+
/* User Floating-Point CSRs */
static RISCVException read_fflags(CPURISCVState *env, int csrno,
target_ulong *val)
@@ -2111,6 +2143,10 @@ static RISCVException write_menvcfg(CPURISCVState *env, int csrno,
if (env_archcpu(env)->cfg.ext_zicfilp) {
mask |= MENVCFG_LPE;
}
+
+ if (env_archcpu(env)->cfg.ext_zicfiss) {
+ mask |= MENVCFG_SSE;
+ }
}
env->menvcfg = (env->menvcfg & ~mask) | (val & mask);
@@ -2167,6 +2203,13 @@ static RISCVException write_senvcfg(CPURISCVState *env, int csrno,
mask |= SENVCFG_LPE;
}
+ /* Higher mode SSE must be ON for next-less mode SSE to be ON */
+ if (env_archcpu(env)->cfg.ext_zicfiss &&
+ get_field(env->menvcfg, MENVCFG_SSE) &&
+ (env->virt_enabled ? get_field(env->henvcfg, HENVCFG_SSE) : true)) {
+ mask |= SENVCFG_SSE;
+ }
+
env->senvcfg = (env->senvcfg & ~mask) | (val & mask);
return RISCV_EXCP_NONE;
}
@@ -2208,6 +2251,12 @@ static RISCVException write_henvcfg(CPURISCVState *env, int csrno,
if (env_archcpu(env)->cfg.ext_zicfilp) {
mask |= HENVCFG_LPE;
}
+
+ /* H can light up SSE for VS only if HS had it from menvcfg */
+ if (env_archcpu(env)->cfg.ext_zicfiss &&
+ get_field(env->menvcfg, MENVCFG_SSE)) {
+ mask |= HENVCFG_SSE;
+ }
}
env->henvcfg = (env->henvcfg & ~mask) | (val & mask);
@@ -4663,6 +4712,9 @@ riscv_csr_operations csr_ops[CSR_TABLE_SIZE] = {
/* Zcmt Extension */
[CSR_JVT] = {"jvt", zcmt, read_jvt, write_jvt},
+ /* zicfiss Extension, shadow stack register */
+ [CSR_SSP] = { "ssp", cfi_ss, read_ssp, write_ssp },
+
#if !defined(CONFIG_USER_ONLY)
/* Machine Timers and Counters */
[CSR_MCYCLE] = { "mcycle", any, read_hpmcounter,
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 10/15] target/riscv: tb flag for shadow stack instructions
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (8 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 09/15] target/riscv: introduce ssp and enabling controls for zicfiss Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection Deepak Gupta
` (4 subsequent siblings)
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta,
Richard Henderson
Shadow stack instructions can be decoded as zimop / zcmop or shadow stack
instructions depending on whether shadow stack are enabled at current
privilege. This requires a TB flag so that correct TB generation and correct
TB lookup happens. `DisasContext` gets a field indicating whether bcfi is
enabled or not.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
target/riscv/cpu.h | 2 ++
target/riscv/cpu_helper.c | 4 ++++
target/riscv/translate.c | 4 ++++
3 files changed, 10 insertions(+)
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index 5a57099d59..dcc3bc9d93 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -613,6 +613,8 @@ FIELD(TB_FLAGS, AXL, 26, 2)
/* zicfilp needs a TB flag to track indirect branches */
FIELD(TB_FLAGS, FCFI_ENABLED, 28, 1)
FIELD(TB_FLAGS, FCFI_LP_EXPECTED, 29, 1)
+/* zicfiss needs a TB flag so that correct TB is located based on tb flags */
+FIELD(TB_FLAGS, BCFI_ENABLED, 30, 1)
#ifdef TARGET_RISCV32
#define riscv_cpu_mxl(env) ((void)(env), MXL_RV32)
diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index 96571177ef..d3115da28d 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -170,6 +170,10 @@ void cpu_get_tb_cpu_state(CPURISCVState *env, vaddr *pc,
flags = FIELD_DP32(flags, TB_FLAGS, FCFI_ENABLED, 1);
}
+ if (cpu_get_bcfien(env)) {
+ flags = FIELD_DP32(flags, TB_FLAGS, BCFI_ENABLED, 1);
+ }
+
#ifdef CONFIG_USER_ONLY
fs = EXT_STATUS_DIRTY;
vs = EXT_STATUS_DIRTY;
diff --git a/target/riscv/translate.c b/target/riscv/translate.c
index 6c1c868a4a..42dc8d3600 100644
--- a/target/riscv/translate.c
+++ b/target/riscv/translate.c
@@ -119,6 +119,8 @@ typedef struct DisasContext {
/* zicfilp extension. fcfi_enabled, lp expected or not */
bool fcfi_enabled;
bool fcfi_lp_expected;
+ /* zicfiss extension, if shadow stack was enabled during TB gen */
+ bool bcfi_enabled;
} DisasContext;
static inline bool has_ext(DisasContext *ctx, uint32_t ext)
@@ -1241,6 +1243,8 @@ static void riscv_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs)
ctx->pm_base_enabled = FIELD_EX32(tb_flags, TB_FLAGS, PM_BASE_ENABLED);
ctx->ztso = cpu->cfg.ext_ztso;
ctx->itrigger = FIELD_EX32(tb_flags, TB_FLAGS, ITRIGGER);
+ ctx->bcfi_enabled = cpu_get_bcfien(env) &&
+ FIELD_EX32(tb_flags, TB_FLAGS, BCFI_ENABLED);
ctx->fcfi_lp_expected = FIELD_EX32(tb_flags, TB_FLAGS, FCFI_LP_EXPECTED);
ctx->fcfi_enabled = FIELD_EX32(tb_flags, TB_FLAGS, FCFI_ENABLED);
ctx->zero = tcg_constant_tl(0);
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (9 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 10/15] target/riscv: tb flag for shadow stack instructions Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 7:35 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 12/15] target/riscv: implement zicfiss instructions Deepak Gupta
` (3 subsequent siblings)
14 siblings, 1 reply; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta,
Richard Henderson
zicfiss protects shadow stack using new page table encodings PTE.W=0,
PTE.R=0 and PTE.X=0. This encoding is reserved if zicfiss is not
implemented or if shadow stack are not enabled.
Loads on shadow stack memory are allowed while stores to shadow stack
memory leads to access faults. Shadow stack accesses to RO memory
leads to store page fault.
To implement special nature of shadow stack memory where only selected
stores (shadow stack stores from sspush) have to be allowed while rest
of regular stores disallowed, new MMU TLB index is created for shadow
stack.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
---
target/riscv/cpu_helper.c | 42 +++++++++++++++++++++++++++++++++------
target/riscv/internals.h | 3 +++
2 files changed, 39 insertions(+), 6 deletions(-)
diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index d3115da28d..f74a1216b1 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -894,6 +894,8 @@ static int get_physical_address(CPURISCVState *env, hwaddr *physical,
hwaddr ppn;
int napot_bits = 0;
target_ulong napot_mask;
+ bool is_sstack_idx = ((mmu_idx & MMU_IDX_SS_WRITE) == MMU_IDX_SS_WRITE);
+ bool sstack_page = false;
/*
* Check if we should use the background registers for the two
@@ -1102,21 +1104,36 @@ restart:
return TRANSLATE_FAIL;
}
+ target_ulong rwx = pte & (PTE_R | PTE_W | PTE_X);
/* Check for reserved combinations of RWX flags. */
- switch (pte & (PTE_R | PTE_W | PTE_X)) {
- case PTE_W:
+ switch (rwx) {
case PTE_W | PTE_X:
return TRANSLATE_FAIL;
+ case PTE_W:
+ /* if bcfi enabled, PTE_W is not reserved and shadow stack page */
+ if (cpu_get_bcfien(env) && first_stage) {
+ sstack_page = true;
+ /* if ss index, read and write allowed. else only read allowed */
+ rwx = is_sstack_idx ? PTE_R | PTE_W : PTE_R;
+ break;
+ }
+ return TRANSLATE_FAIL;
+ case PTE_R:
+ /* shadow stack writes to readonly memory are page faults */
+ if (is_sstack_idx && access_type == MMU_DATA_STORE) {
+ return TRANSLATE_FAIL;
+ }
+ break;
}
int prot = 0;
- if (pte & PTE_R) {
+ if (rwx & PTE_R) {
prot |= PAGE_READ;
}
- if (pte & PTE_W) {
+ if (rwx & PTE_W) {
prot |= PAGE_WRITE;
}
- if (pte & PTE_X) {
+ if (rwx & PTE_X) {
bool mxr = false;
/*
@@ -1161,7 +1178,7 @@ restart:
if (!((prot >> access_type) & 1)) {
/* Access check failed */
- return TRANSLATE_FAIL;
+ return sstack_page ? TRANSLATE_PMP_FAIL : TRANSLATE_FAIL;
}
target_ulong updated_pte = pte;
@@ -1348,9 +1365,17 @@ void riscv_cpu_do_unaligned_access(CPUState *cs, vaddr addr,
break;
case MMU_DATA_LOAD:
cs->exception_index = RISCV_EXCP_LOAD_ADDR_MIS;
+ /* shadow stack mis aligned accesses are access faults */
+ if (mmu_idx & MMU_IDX_SS_WRITE) {
+ cs->exception_index = RISCV_EXCP_LOAD_ACCESS_FAULT;
+ }
break;
case MMU_DATA_STORE:
cs->exception_index = RISCV_EXCP_STORE_AMO_ADDR_MIS;
+ /* shadow stack mis aligned accesses are access faults */
+ if (mmu_idx & MMU_IDX_SS_WRITE) {
+ cs->exception_index = RISCV_EXCP_STORE_AMO_ACCESS_FAULT;
+ }
break;
default:
g_assert_not_reached();
@@ -1406,6 +1431,11 @@ bool riscv_cpu_tlb_fill(CPUState *cs, vaddr address, int size,
qemu_log_mask(CPU_LOG_MMU, "%s ad %" VADDR_PRIx " rw %d mmu_idx %d\n",
__func__, address, access_type, mmu_idx);
+ /* If shadow stack instruction initiated this access, treat it as store */
+ if (mmu_idx & MMU_IDX_SS_WRITE) {
+ access_type = MMU_DATA_STORE;
+ }
+
pmu_tlb_fill_incr_ctr(cpu, access_type);
if (two_stage_lookup) {
/* Two stage lookup */
diff --git a/target/riscv/internals.h b/target/riscv/internals.h
index 0ac17bc5ad..ddbdee885b 100644
--- a/target/riscv/internals.h
+++ b/target/riscv/internals.h
@@ -30,12 +30,15 @@
* - U+2STAGE 0b100
* - S+2STAGE 0b101
* - S+SUM+2STAGE 0b110
+ * - Shadow stack+U 0b1000
+ * - Shadow stack+S 0b1001
*/
#define MMUIdx_U 0
#define MMUIdx_S 1
#define MMUIdx_S_SUM 2
#define MMUIdx_M 3
#define MMU_2STAGE_BIT (1 << 2)
+#define MMU_IDX_SS_WRITE (1 << 3)
static inline int mmuidx_priv(int mmu_idx)
{
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 12/15] target/riscv: implement zicfiss instructions
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (10 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 13/15] target/riscv: compressed encodings for sspush and sspopchk Deepak Gupta
` (2 subsequent siblings)
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta
zicfiss has following instructions
- sspopchk: pops a value from shadow stack and compares with x1/x5.
If they dont match, reports a sw check exception with tval = 3.
- sspush: pushes value in x1/x5 on shadow stack
- ssrdp: reads current shadow stack
- ssamoswap: swaps contents of shadow stack atomically
sspopchk/sspush/ssrdp default to zimop if zimop implemented and SSE=0
If SSE=0, ssamoswap is illegal instruction exception.
This patch implements shadow stack operations for qemu-user and shadow
stack is not protected.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
---
target/riscv/cpu_bits.h | 2 +
target/riscv/insn32.decode | 21 +++++-
target/riscv/insn_trans/trans_rva.c.inc | 41 +++++++++++
target/riscv/insn_trans/trans_rvzicfiss.c.inc | 73 +++++++++++++++++++
target/riscv/translate.c | 5 ++
5 files changed, 140 insertions(+), 2 deletions(-)
create mode 100644 target/riscv/insn_trans/trans_rvzicfiss.c.inc
diff --git a/target/riscv/cpu_bits.h b/target/riscv/cpu_bits.h
index 48ce24dc32..bb62fbe9ec 100644
--- a/target/riscv/cpu_bits.h
+++ b/target/riscv/cpu_bits.h
@@ -690,6 +690,8 @@ typedef enum RISCVException {
/* zicfilp defines lp violation results in sw check with tval = 2*/
#define RISCV_EXCP_SW_CHECK_FCFI_TVAL 2
+/* zicfiss defines ss violation results in sw check with tval = 3*/
+#define RISCV_EXCP_SW_CHECK_BCFI_TVAL 3
#define RISCV_EXCP_INT_FLAG 0x80000000
#define RISCV_EXCP_INT_MASK 0x7fffffff
diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode
index 6533cb0758..676a7ea98e 100644
--- a/target/riscv/insn32.decode
+++ b/target/riscv/insn32.decode
@@ -246,6 +246,7 @@ remud 0000001 ..... ..... 111 ..... 1111011 @r
lr_w 00010 . . 00000 ..... 010 ..... 0101111 @atom_ld
sc_w 00011 . . ..... ..... 010 ..... 0101111 @atom_st
amoswap_w 00001 . . ..... ..... 010 ..... 0101111 @atom_st
+ssamoswap_w 01001 . . ..... ..... 010 ..... 0101111 @atom_st
amoadd_w 00000 . . ..... ..... 010 ..... 0101111 @atom_st
amoxor_w 00100 . . ..... ..... 010 ..... 0101111 @atom_st
amoand_w 01100 . . ..... ..... 010 ..... 0101111 @atom_st
@@ -259,6 +260,7 @@ amomaxu_w 11100 . . ..... ..... 010 ..... 0101111 @atom_st
lr_d 00010 . . 00000 ..... 011 ..... 0101111 @atom_ld
sc_d 00011 . . ..... ..... 011 ..... 0101111 @atom_st
amoswap_d 00001 . . ..... ..... 011 ..... 0101111 @atom_st
+ssamoswap_d 01001 . . ..... ..... 011 ..... 0101111 @atom_st
amoadd_d 00000 . . ..... ..... 011 ..... 0101111 @atom_st
amoxor_d 00100 . . ..... ..... 011 ..... 0101111 @atom_st
amoand_d 01100 . . ..... ..... 011 ..... 0101111 @atom_st
@@ -1022,8 +1024,23 @@ amocas_d 00101 . . ..... ..... 011 ..... 0101111 @atom_st
amocas_q 00101 . . ..... ..... 100 ..... 0101111 @atom_st
# *** Zimop may-be-operation extension ***
-mop_r_n 1 . 00 .. 0111 .. ..... 100 ..... 1110011 @mop5
-mop_rr_n 1 . 00 .. 1 ..... ..... 100 ..... 1110011 @mop3
+{
+ # zicfiss instructions carved out of mop.r
+ [
+ ssrdp 1100110 11100 00000 100 rd:5 1110011
+ sspopchk 1100110 11100 00001 100 00000 1110011 &r2 rs1=1 rd=0
+ sspopchk 1100110 11100 00101 100 00000 1110011 &r2 rs1=5 rd=0
+ ]
+ mop_r_n 1 . 00 .. 0111 .. ..... 100 ..... 1110011 @mop5
+}
+{
+ # zicfiss instruction carved out of mop.rr
+ [
+ sspush 1100111 00001 00000 100 00000 1110011 &r2_s rs2=1 rs1=0
+ sspush 1100111 00101 00000 100 00000 1110011 &r2_s rs2=5 rs1=0
+ ]
+ mop_rr_n 1 . 00 .. 1 ..... ..... 100 ..... 1110011 @mop3
+}
# *** Zabhb Standard Extension ***
amoswap_b 00001 . . ..... ..... 000 ..... 0101111 @atom_st
diff --git a/target/riscv/insn_trans/trans_rva.c.inc b/target/riscv/insn_trans/trans_rva.c.inc
index 39bbf60f3c..a2d9c186fd 100644
--- a/target/riscv/insn_trans/trans_rva.c.inc
+++ b/target/riscv/insn_trans/trans_rva.c.inc
@@ -18,6 +18,8 @@
* this program. If not, see <http://www.gnu.org/licenses/>.
*/
+#include "exec/memop.h"
+
#define REQUIRE_A_OR_ZAAMO(ctx) do { \
if (!ctx->cfg_ptr->ext_zaamo && !has_ext(ctx, RVA)) { \
return false; \
@@ -114,6 +116,25 @@ static bool trans_amoswap_w(DisasContext *ctx, arg_amoswap_w *a)
return gen_amo(ctx, a, &tcg_gen_atomic_xchg_tl, MO_TESL);
}
+static bool trans_ssamoswap_w(DisasContext *ctx, arg_amoswap_w *a)
+{
+ REQUIRE_A_OR_ZAAMO(ctx);
+ if (!ctx->bcfi_enabled) {
+ return false;
+ }
+
+ TCGv dest = dest_gpr(ctx, a->rd);
+ TCGv src1, src2 = get_gpr(ctx, a->rs2, EXT_NONE);
+
+ decode_save_opc(ctx);
+ src1 = get_address(ctx, a->rs1, 0);
+
+ tcg_gen_atomic_xchg_tl(dest, src1, src2, SS_MMU_INDEX(ctx),
+ (MO_ALIGN | MO_TESL));
+ gen_set_gpr(ctx, a->rd, dest);
+ return true;
+}
+
static bool trans_amoadd_w(DisasContext *ctx, arg_amoadd_w *a)
{
REQUIRE_A_OR_ZAAMO(ctx);
@@ -183,6 +204,26 @@ static bool trans_amoswap_d(DisasContext *ctx, arg_amoswap_d *a)
return gen_amo(ctx, a, &tcg_gen_atomic_xchg_tl, MO_TEUQ);
}
+static bool trans_ssamoswap_d(DisasContext *ctx, arg_amoswap_w *a)
+{
+ REQUIRE_64BIT(ctx);
+ REQUIRE_A_OR_ZAAMO(ctx);
+ if (!ctx->bcfi_enabled) {
+ return false;
+ }
+
+ TCGv dest = dest_gpr(ctx, a->rd);
+ TCGv src1, src2 = get_gpr(ctx, a->rs2, EXT_NONE);
+
+ decode_save_opc(ctx);
+ src1 = get_address(ctx, a->rs1, 0);
+
+ tcg_gen_atomic_xchg_tl(dest, src1, src2, SS_MMU_INDEX(ctx),
+ (MO_ALIGN | MO_TESQ));
+ gen_set_gpr(ctx, a->rd, dest);
+ return true;
+}
+
static bool trans_amoadd_d(DisasContext *ctx, arg_amoadd_d *a)
{
REQUIRE_64BIT(ctx);
diff --git a/target/riscv/insn_trans/trans_rvzicfiss.c.inc b/target/riscv/insn_trans/trans_rvzicfiss.c.inc
new file mode 100644
index 0000000000..bbe52ee648
--- /dev/null
+++ b/target/riscv/insn_trans/trans_rvzicfiss.c.inc
@@ -0,0 +1,73 @@
+/*
+ * RISC-V translation routines for the Control-Flow Integrity Extension
+ *
+ * Copyright (c) 2024 Rivos Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2 or later, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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/>.
+ */
+static bool trans_sspopchk(DisasContext *ctx, arg_sspopchk *a)
+{
+ if (!ctx->bcfi_enabled) {
+ return false;
+ }
+
+ TCGv addr = tcg_temp_new();
+ TCGLabel *skip = gen_new_label();
+ uint32_t tmp = (get_xl(ctx) == MXL_RV64) ? 8 : 4;
+ TCGv data = tcg_temp_new();
+ tcg_gen_ld_tl(addr, tcg_env, offsetof(CPURISCVState, ssp));
+ tcg_gen_qemu_ld_tl(data, addr, SS_MMU_INDEX(ctx),
+ mxl_memop(ctx) | MO_ALIGN);
+ TCGv rs1 = get_gpr(ctx, a->rs1, EXT_NONE);
+ tcg_gen_brcond_tl(TCG_COND_EQ, data, rs1, skip);
+ tcg_gen_st_tl(tcg_constant_tl(RISCV_EXCP_SW_CHECK_BCFI_TVAL),
+ tcg_env, offsetof(CPURISCVState, sw_check_code));
+ gen_helper_raise_exception(tcg_env,
+ tcg_constant_i32(RISCV_EXCP_SW_CHECK));
+ gen_set_label(skip);
+ tcg_gen_addi_tl(addr, addr, tmp);
+ tcg_gen_st_tl(addr, tcg_env, offsetof(CPURISCVState, ssp));
+
+ return true;
+}
+
+static bool trans_sspush(DisasContext *ctx, arg_sspush *a)
+{
+ if (!ctx->bcfi_enabled) {
+ return false;
+ }
+
+ TCGv addr = tcg_temp_new();
+ int tmp = (get_xl(ctx) == MXL_RV64) ? -8 : -4;
+ TCGv data = get_gpr(ctx, a->rs2, EXT_NONE);
+ tcg_gen_ld_tl(addr, tcg_env, offsetof(CPURISCVState, ssp));
+ tcg_gen_addi_tl(addr, addr, tmp);
+ tcg_gen_qemu_st_tl(data, addr, SS_MMU_INDEX(ctx),
+ mxl_memop(ctx) | MO_ALIGN);
+ tcg_gen_st_tl(addr, tcg_env, offsetof(CPURISCVState, ssp));
+
+ return true;
+}
+
+static bool trans_ssrdp(DisasContext *ctx, arg_ssrdp *a)
+{
+ if (!ctx->bcfi_enabled || a->rd == 0) {
+ return false;
+ }
+
+ TCGv dest = tcg_temp_new();
+ tcg_gen_ld_tl(dest, tcg_env, offsetof(CPURISCVState, ssp));
+ gen_set_gpr(ctx, a->rd, dest);
+
+ return true;
+}
diff --git a/target/riscv/translate.c b/target/riscv/translate.c
index 42dc8d3600..cc427c00ac 100644
--- a/target/riscv/translate.c
+++ b/target/riscv/translate.c
@@ -144,6 +144,8 @@ static inline bool has_ext(DisasContext *ctx, uint32_t ext)
#define get_address_xl(ctx) ((ctx)->address_xl)
#endif
+#define mxl_memop(ctx) ((get_xl(ctx) + 1) | MO_TE)
+
/* The word size for this machine mode. */
static inline int __attribute__((unused)) get_xlen(DisasContext *ctx)
{
@@ -1126,6 +1128,8 @@ static uint32_t opcode_at(DisasContextBase *dcbase, target_ulong pc)
return translator_ldl(env, &ctx->base, pc);
}
+#define SS_MMU_INDEX(ctx) (ctx->mem_idx | MMU_IDX_SS_WRITE)
+
/* Include insn module translation function */
#include "insn_trans/trans_rvi.c.inc"
#include "insn_trans/trans_rvm.c.inc"
@@ -1156,6 +1160,7 @@ static uint32_t opcode_at(DisasContextBase *dcbase, target_ulong pc)
#include "decode-insn16.c.inc"
#include "insn_trans/trans_rvzce.c.inc"
#include "insn_trans/trans_rvzcmop.c.inc"
+#include "insn_trans/trans_rvzicfiss.c.inc"
/* Include decoders for factored-out extensions */
#include "decode-XVentanaCondOps.c.inc"
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 13/15] target/riscv: compressed encodings for sspush and sspopchk
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (11 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 12/15] target/riscv: implement zicfiss instructions Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 14/15] disas/riscv: enable disassembly for zicfiss instructions Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 15/15] disas/riscv: enable disassembly for compressed sspush/sspopchk Deepak Gupta
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta
sspush/sspopchk have compressed encodings carved out of zcmops.
compressed sspush is designated as c.mop.1 while compressed sspopchk
is designated as c.mop.5.
Note that c.sspush x1 exists while c.sspush x5 doesn't. Similarly
c.sspopchk x5 exists while c.sspopchk x1 doesn't.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
---
target/riscv/insn16.decode | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/target/riscv/insn16.decode b/target/riscv/insn16.decode
index 3953bcf82d..bf893d1c2e 100644
--- a/target/riscv/insn16.decode
+++ b/target/riscv/insn16.decode
@@ -140,6 +140,10 @@ sw 110 ... ... .. ... 00 @cs_w
addi 000 . ..... ..... 01 @ci
addi 010 . ..... ..... 01 @c_li
{
+ # c.sspush x1 carving out of zcmops
+ sspush 011 0 00001 00000 01 &r2_s rs2=1 rs1=0
+ # c.sspopchk x5 carving out of zcmops
+ sspopchk 011 0 00101 00000 01 &r2 rs1=5 rd=0
c_mop_n 011 0 0 n:3 1 00000 01
illegal 011 0 ----- 00000 01 # c.addi16sp and c.lui, RES nzimm=0
addi 011 . 00010 ..... 01 @c_addi16sp
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 14/15] disas/riscv: enable disassembly for zicfiss instructions
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (12 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 13/15] target/riscv: compressed encodings for sspush and sspopchk Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 15/15] disas/riscv: enable disassembly for compressed sspush/sspopchk Deepak Gupta
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta
Enable disassembly for sspush, sspopchk, ssrdp & ssamoswap.
Disasembly is only enabled if zimop and zicfiss ext is set to true.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
---
disas/riscv.c | 40 +++++++++++++++++++++++++++++++++++++++-
disas/riscv.h | 1 +
2 files changed, 40 insertions(+), 1 deletion(-)
diff --git a/disas/riscv.c b/disas/riscv.c
index c7c92acef7..f1f4ffc50a 100644
--- a/disas/riscv.c
+++ b/disas/riscv.c
@@ -975,6 +975,11 @@ typedef enum {
rv_op_amocas_b = 944,
rv_op_amocas_h = 945,
rv_op_lpad = 946,
+ rv_op_sspush = 947,
+ rv_op_sspopchk = 948,
+ rv_op_ssrdp = 949,
+ rv_op_ssamoswap_w = 950,
+ rv_op_ssamoswap_d = 951,
} rv_op;
/* register names */
@@ -2234,6 +2239,11 @@ const rv_opcode_data rvi_opcode_data[] = {
{ "amocas.b", rv_codec_r_a, rv_fmt_aqrl_rd_rs2_rs1, NULL, 0, 0, 0 },
{ "amocas.h", rv_codec_r_a, rv_fmt_aqrl_rd_rs2_rs1, NULL, 0, 0, 0 },
{ "lpad", rv_codec_lp, rv_fmt_imm, NULL, 0, 0, 0 },
+ { "sspush", rv_codec_r, rv_fmt_rs2, NULL, 0, 0, 0 },
+ { "sspopchk", rv_codec_r, rv_fmt_rs1, NULL, 0, 0, 0 },
+ { "ssrdp", rv_codec_r, rv_fmt_rd, NULL, 0, 0, 0 },
+ { "ssamoswap.w", rv_codec_r_a, rv_fmt_aqrl_rd_rs2_rs1, NULL, 0, 0, 0 },
+ { "ssamoswap.d", rv_codec_r_a, rv_fmt_aqrl_rd_rs2_rs1, NULL, 0, 0, 0 },
};
/* CSR names */
@@ -2251,6 +2261,7 @@ static const char *csr_name(int csrno)
case 0x0009: return "vxsat";
case 0x000a: return "vxrm";
case 0x000f: return "vcsr";
+ case 0x0011: return "ssp";
case 0x0015: return "seed";
case 0x0017: return "jvt";
case 0x0040: return "uscratch";
@@ -3077,6 +3088,8 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
case 66: op = rv_op_amoor_w; break;
case 67: op = rv_op_amoor_d; break;
case 68: op = rv_op_amoor_q; break;
+ case 74: op = rv_op_ssamoswap_w; break;
+ case 75: op = rv_op_ssamoswap_d; break;
case 96: op = rv_op_amoand_b; break;
case 97: op = rv_op_amoand_h; break;
case 98: op = rv_op_amoand_w; break;
@@ -4028,7 +4041,7 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
case 3: op = rv_op_csrrc; break;
case 4:
if (dec->cfg->ext_zimop) {
- int imm_mop5, imm_mop3;
+ int imm_mop5, imm_mop3, reg_num;
if ((extract32(inst, 22, 10) & 0b1011001111)
== 0b1000000111) {
imm_mop5 = deposit32(deposit32(extract32(inst, 20, 2),
@@ -4036,11 +4049,36 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
extract32(inst, 26, 2)),
4, 1, extract32(inst, 30, 1));
op = rv_mop_r_0 + imm_mop5;
+ /* if zicfiss enabled and mop5 is shadow stack */
+ if (dec->cfg->ext_zicfiss &&
+ ((imm_mop5 & 0b11100) == 0b11100)) {
+ /* rs1=0 means ssrdp */
+ if ((inst & (0b011111 << 15)) == 0) {
+ op = rv_op_ssrdp;
+ }
+ /* rd=0 means sspopchk */
+ reg_num = (inst >> 15) & 0b011111;
+ if (((inst & (0b011111 << 7)) == 0) &&
+ ((reg_num == 1) || (reg_num == 5))) {
+ op = rv_op_sspopchk;
+ }
+ }
} else if ((extract32(inst, 25, 7) & 0b1011001)
== 0b1000001) {
imm_mop3 = deposit32(extract32(inst, 26, 2),
2, 1, extract32(inst, 30, 1));
op = rv_mop_rr_0 + imm_mop3;
+ /* if zicfiss enabled and mop3 is shadow stack */
+ if (dec->cfg->ext_zicfiss &&
+ ((imm_mop3 & 0b111) == 0b111)) {
+ /* rs1=0 and rd=0 means sspush */
+ reg_num = (inst >> 20) & 0b011111;
+ if (((inst & (0b011111 << 15)) == 0) &&
+ ((inst & (0b011111 << 7)) == 0) &&
+ ((reg_num == 1) || (reg_num == 5))) {
+ op = rv_op_sspush;
+ }
+ }
}
}
break;
diff --git a/disas/riscv.h b/disas/riscv.h
index 1182457aff..4895c5a301 100644
--- a/disas/riscv.h
+++ b/disas/riscv.h
@@ -224,6 +224,7 @@ enum {
#define rv_fmt_none "O\t"
#define rv_fmt_rs1 "O\t1"
+#define rv_fmt_rs2 "O\t2"
#define rv_fmt_offset "O\to"
#define rv_fmt_pred_succ "O\tp,s"
#define rv_fmt_rs1_rs2 "O\t1,2"
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [PATCH v5 15/15] disas/riscv: enable disassembly for compressed sspush/sspopchk
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
` (13 preceding siblings ...)
2024-08-20 0:01 ` [PATCH v5 14/15] disas/riscv: enable disassembly for zicfiss instructions Deepak Gupta
@ 2024-08-20 0:01 ` Deepak Gupta
14 siblings, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 0:01 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Deepak Gupta
sspush and sspopchk have equivalent compressed encoding taken from zcmop.
cmop.1 is sspush x1 while cmop.5 is sspopchk x5. Due to unusual encoding
for both rs1 and rs2 from space bitfield, this required a new codec.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
---
disas/riscv.c | 19 ++++++++++++++++++-
disas/riscv.h | 1 +
2 files changed, 19 insertions(+), 1 deletion(-)
diff --git a/disas/riscv.c b/disas/riscv.c
index f1f4ffc50a..8eeaa9f843 100644
--- a/disas/riscv.c
+++ b/disas/riscv.c
@@ -980,6 +980,8 @@ typedef enum {
rv_op_ssrdp = 949,
rv_op_ssamoswap_w = 950,
rv_op_ssamoswap_d = 951,
+ rv_op_c_sspush = 952,
+ rv_op_c_sspopchk = 953,
} rv_op;
/* register names */
@@ -2244,6 +2246,10 @@ const rv_opcode_data rvi_opcode_data[] = {
{ "ssrdp", rv_codec_r, rv_fmt_rd, NULL, 0, 0, 0 },
{ "ssamoswap.w", rv_codec_r_a, rv_fmt_aqrl_rd_rs2_rs1, NULL, 0, 0, 0 },
{ "ssamoswap.d", rv_codec_r_a, rv_fmt_aqrl_rd_rs2_rs1, NULL, 0, 0, 0 },
+ { "c.sspush", rv_codec_cmop_ss, rv_fmt_rs2, NULL, rv_op_sspush,
+ rv_op_sspush, 0 },
+ { "c.sspopchk", rv_codec_cmop_ss, rv_fmt_rs1, NULL, rv_op_sspopchk,
+ rv_op_sspopchk, 0 },
};
/* CSR names */
@@ -2604,7 +2610,13 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
if (dec->cfg->ext_zcmop) {
if ((((inst >> 2) & 0b111111) == 0b100000) &&
(((inst >> 11) & 0b11) == 0b0)) {
- op = rv_c_mop_1 + ((inst >> 8) & 0b111);
+ unsigned int cmop_code = 0;
+ cmop_code = ((inst >> 8) & 0b111);
+ op = rv_c_mop_1 + cmop_code;
+ if (dec->cfg->ext_zicfiss) {
+ op = (cmop_code == 0) ? rv_op_c_sspush : op;
+ op = (cmop_code == 2) ? rv_op_c_sspopchk : op;
+ }
break;
}
}
@@ -4923,6 +4935,11 @@ static void decode_inst_operands(rv_decode *dec, rv_isa isa)
case rv_codec_lp:
dec->imm = operand_lpl(inst);
break;
+ case rv_codec_cmop_ss:
+ dec->rd = rv_ireg_zero;
+ dec->rs1 = dec->rs2 = operand_crs1(inst);
+ dec->imm = 0;
+ break;
};
}
diff --git a/disas/riscv.h b/disas/riscv.h
index 4895c5a301..6a3b371cd3 100644
--- a/disas/riscv.h
+++ b/disas/riscv.h
@@ -167,6 +167,7 @@ typedef enum {
rv_codec_r2_imm2_imm5,
rv_codec_fli,
rv_codec_lp,
+ rv_codec_cmop_ss,
} rv_codec;
/* structures */
--
2.44.0
^ permalink raw reply related [flat|nested] 25+ messages in thread
* Re: [PATCH v5 02/15] target/riscv: Introduce elp state and enabling controls for zicfilp
2024-08-20 0:01 ` [PATCH v5 02/15] target/riscv: Introduce elp state and enabling controls for zicfilp Deepak Gupta
@ 2024-08-20 5:17 ` Richard Henderson
0 siblings, 0 replies; 25+ messages in thread
From: Richard Henderson @ 2024-08-20 5:17 UTC (permalink / raw)
To: Deepak Gupta, qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng
On 8/20/24 10:01, Deepak Gupta wrote:
> zicfilp introduces a new state elp ("expected landing pad") in cpu.
> During normal execution, elp is idle (NO_LP_EXPECTED) i.e not expecting
> landing pad. On an indirect call, elp moves LP_EXPECTED. When elp is
> LP_EXPECTED, only a subsquent landing pad instruction can set state back
> to NO_LP_EXPECTED. On reset, elp is set to NO_LP_EXPECTED.
>
> zicfilp is enabled via bit2 in *envcfg CSRs. Enabling control for M-mode
> is in mseccfg CSR at bit position 10.
>
> On trap, elp state is saved away in *status.
>
> Signed-off-by: Deepak Gupta<debug@rivosinc.com>
> Co-developed-by: Jim Shu<jim.shu@sifive.com>
> Co-developed-by: Andy Chiu<andy.chiu@sifive.com>
> Reviewed-by: Richard Henderson<richard.henderson@linaro.org>
> ---
> target/riscv/cpu.c | 3 +++
> target/riscv/cpu.h | 2 ++
> target/riscv/cpu_bits.h | 6 ++++++
> target/riscv/csr.c | 31 +++++++++++++++++++++++++++++++
> target/riscv/pmp.c | 5 +++++
> target/riscv/pmp.h | 3 ++-
> 6 files changed, 49 insertions(+), 1 deletion(-)
One thing missing here is to add
> + /* elp state for zicfilp extension */
> + bool elp;
this to the migration state, as a subsection under vmstate_riscv_cpu.
Sorry I missed this before...
r~
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [PATCH v5 05/15] target/riscv: tracking indirect branches (fcfi) for zicfilp
2024-08-20 0:01 ` [PATCH v5 05/15] target/riscv: tracking indirect branches (fcfi) for zicfilp Deepak Gupta
@ 2024-08-20 5:24 ` Richard Henderson
0 siblings, 0 replies; 25+ messages in thread
From: Richard Henderson @ 2024-08-20 5:24 UTC (permalink / raw)
To: Deepak Gupta, qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng
On 8/20/24 10:01, Deepak Gupta wrote:
> zicfilp protects forward control flow (if enabled) by enforcing all
> indirect call and jmp must land on a landing pad instruction `lpad`. If
> target of an indirect call or jmp is not `lpad` then cpu/hart must raise
> a sw check exception with tval = 2.
>
> This patch implements the mechanism using TCG. Target architecture branch
> instruction must define the end of a TB. Using this property, during
> translation of branch instruction, TB flag = FCFI_LP_EXPECTED can be set.
> Translation of target TB can check if FCFI_LP_EXPECTED flag is set and a
> flag (fcfi_lp_expected) can be set in DisasContext. If `lpad` gets
> translated, fcfi_lp_expected flag in DisasContext can be cleared. Else
> it'll fault.
>
> Signed-off-by: Deepak Gupta<debug@rivosinc.com>
> Co-developed-by: Jim Shu<jim.shu@sifive.com>
> Co-developed-by: Andy Chiu<andy.chiu@sifive.com>
> Suggested-by: Richard Henderson<richard.henderson@linaro.org>
> ---
> target/riscv/cpu.h | 3 +++
> target/riscv/cpu_bits.h | 3 +++
> target/riscv/cpu_helper.c | 12 ++++++++++++
> target/riscv/translate.c | 21 ++++++++++++++++++++-
> 4 files changed, 38 insertions(+), 1 deletion(-)
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> @@ -1265,11 +1270,25 @@ static void riscv_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
> DisasContext *ctx = container_of(dcbase, DisasContext, base);
> CPURISCVState *env = cpu_env(cpu);
> uint16_t opcode16 = translator_lduw(env, &ctx->base, ctx->base.pc_next);
> -
> ctx->ol = ctx->xl;
> decode_opc(env, ctx, opcode16);
> ctx->base.pc_next += ctx->cur_insn_len;
Watch the unrelated whitespace changes.
r~
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [PATCH v5 06/15] target/riscv: zicfilp `lpad` impl and branch tracking
2024-08-20 0:01 ` [PATCH v5 06/15] target/riscv: zicfilp `lpad` impl and branch tracking Deepak Gupta
@ 2024-08-20 5:29 ` Richard Henderson
0 siblings, 0 replies; 25+ messages in thread
From: Richard Henderson @ 2024-08-20 5:29 UTC (permalink / raw)
To: qemu-devel
On 8/20/24 10:01, Deepak Gupta wrote:
> Implements setting lp expected when `jalr` is encountered and implements
> `lpad` instruction of zicfilp. `lpad` instruction is taken out of
> auipc x0, <imm_20>. This is an existing HINTNOP space. If `lpad` is
> target of an indirect branch, cpu checks for 20 bit value in x7 upper
> with 20 bit value embedded in `lpad`. If they don't match, cpu raises a
> sw check exception with tval = 2.
>
> Signed-off-by: Deepak Gupta <debug@rivosinc.com>
> Co-developed-by: Jim Shu <jim.shu@sifive.com>
> Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
> ---
> target/riscv/cpu_user.h | 1 +
> target/riscv/insn32.decode | 5 ++-
> target/riscv/insn_trans/trans_rvi.c.inc | 55 +++++++++++++++++++++++++
> 3 files changed, 60 insertions(+), 1 deletion(-)
>
> diff --git a/target/riscv/cpu_user.h b/target/riscv/cpu_user.h
> index 02afad608b..e6927ff847 100644
> --- a/target/riscv/cpu_user.h
> +++ b/target/riscv/cpu_user.h
> @@ -15,5 +15,6 @@
> #define xA6 16
> #define xA7 17 /* syscall number for RVI ABI */
> #define xT0 5 /* syscall number for RVE ABI */
> +#define xT2 7
>
> #endif
> diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode
> index c45b8fa1d8..6533cb0758 100644
> --- a/target/riscv/insn32.decode
> +++ b/target/riscv/insn32.decode
> @@ -123,7 +123,10 @@ sfence_vm 0001000 00100 ..... 000 00000 1110011 @sfence_vm
>
> # *** RV32I Base Instruction Set ***
> lui .................... ..... 0110111 @u
> -auipc .................... ..... 0010111 @u
> +{
> + lpad label:20 00000 0010111
> + auipc .................... ..... 0010111 @u
> +}
> jal .................... ..... 1101111 @j
Again, best to line up the decode bits.
While you are required to add two spaces within { },
you can remove two spaces before the first '.':
lui .................... ..... 0110111 @u
{
lpad label:20 00000 0010111
auipc .................... ..... 0010111 @u
}
jal .................... ..... 1101111 @j
Otherwise,
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
r~
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [PATCH v5 09/15] target/riscv: introduce ssp and enabling controls for zicfiss
2024-08-20 0:01 ` [PATCH v5 09/15] target/riscv: introduce ssp and enabling controls for zicfiss Deepak Gupta
@ 2024-08-20 5:34 ` Richard Henderson
0 siblings, 0 replies; 25+ messages in thread
From: Richard Henderson @ 2024-08-20 5:34 UTC (permalink / raw)
To: Deepak Gupta, qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng
On 8/20/24 10:01, Deepak Gupta wrote:
> + /* shadow stack register for zicfiss extension */
> + target_ulong ssp;
This will also require migration.
r~
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection
2024-08-20 0:01 ` [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection Deepak Gupta
@ 2024-08-20 7:35 ` Deepak Gupta
2024-08-20 9:20 ` Richard Henderson
0 siblings, 1 reply; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 7:35 UTC (permalink / raw)
To: qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng, Richard Henderson
On Mon, Aug 19, 2024 at 05:01:25PM -0700, Deepak Gupta wrote:
>zicfiss protects shadow stack using new page table encodings PTE.W=0,
>PTE.R=0 and PTE.X=0. This encoding is reserved if zicfiss is not
>implemented or if shadow stack are not enabled.
>Loads on shadow stack memory are allowed while stores to shadow stack
>memory leads to access faults. Shadow stack accesses to RO memory
>leads to store page fault.
>
>To implement special nature of shadow stack memory where only selected
>stores (shadow stack stores from sspush) have to be allowed while rest
>of regular stores disallowed, new MMU TLB index is created for shadow
>stack.
>
>Signed-off-by: Deepak Gupta <debug@rivosinc.com>
>Suggested-by: Richard Henderson <richard.henderson@linaro.org>
>---
>@@ -1406,6 +1431,11 @@ bool riscv_cpu_tlb_fill(CPUState *cs, vaddr address, int size,
> qemu_log_mask(CPU_LOG_MMU, "%s ad %" VADDR_PRIx " rw %d mmu_idx %d\n",
> __func__, address, access_type, mmu_idx);
>
>+ /* If shadow stack instruction initiated this access, treat it as store */
>+ if (mmu_idx & MMU_IDX_SS_WRITE) {
>+ access_type = MMU_DATA_STORE;
>+ }
>+
I think I forgot to address this. Do you still want me to fix this up like you
had suggested?
IIRC, you mentioned to use TARGET_INSN_START_EXTRA_WORDS=2. Honestly I don't know
what it means and how its used. Based on git grep and some readup, are you expecting something
along the below lines?
"""
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index fee31b8037..dfd2efa941 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -47,7 +47,7 @@ typedef struct CPUArchState CPURISCVState;
* RISC-V-specific extra insn start words:
* 1: Original instruction opcode
*/
-#define TARGET_INSN_START_EXTRA_WORDS 1
+#define TARGET_INSN_START_EXTRA_WORDS 2
#define RV(x) ((target_ulong)1 << (x - 'A'))
diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index f74a1216b1..b266177e46 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -1271,6 +1271,11 @@ static void raise_mmu_exception(CPURISCVState *env, target_ulong address,
{
CPUState *cs = env_cpu(env);
+ if (!pmp_violation &&
+ tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] & 1) {
+ tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] &= ~1;
+ access_type = MMU_DATA_STORE;
+ }
+
switch (access_type) {
case MMU_INST_FETCH:
if (pmp_violation) {
@@ -1433,7 +1438,7 @@ bool riscv_cpu_tlb_fill(CPUState *cs, vaddr address, int size,
/* If shadow stack instruction initiated this access, treat it as store */
if (mmu_idx & MMU_IDX_SS_WRITE) {
- access_type = MMU_DATA_STORE;
+ tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] |= 1;
}
pmu_tlb_fill_incr_ctr(cpu, access_type);
@@ -1529,6 +1534,7 @@ bool riscv_cpu_tlb_fill(CPUState *cs, vaddr address, int size,
if (ret == TRANSLATE_SUCCESS) {
tlb_set_page(cs, address & ~(tlb_size - 1), pa & ~(tlb_size - 1),
prot, mmu_idx, tlb_size);
+ tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] &= ~1;
return true;
} else if (probe) {
"""
> pmu_tlb_fill_incr_ctr(cpu, access_type);
> if (two_stage_lookup) {
> /* Two stage lookup */
>diff --git a/target/riscv/internals.h b/target/riscv/internals.h
>index 0ac17bc5ad..ddbdee885b 100644
>--- a/target/riscv/internals.h
>+++ b/target/riscv/internals.h
>@@ -30,12 +30,15 @@
> * - U+2STAGE 0b100
> * - S+2STAGE 0b101
> * - S+SUM+2STAGE 0b110
>+ * - Shadow stack+U 0b1000
>+ * - Shadow stack+S 0b1001
> */
> #define MMUIdx_U 0
> #define MMUIdx_S 1
> #define MMUIdx_S_SUM 2
> #define MMUIdx_M 3
> #define MMU_2STAGE_BIT (1 << 2)
>+#define MMU_IDX_SS_WRITE (1 << 3)
>
> static inline int mmuidx_priv(int mmu_idx)
> {
>--
>2.44.0
>
^ permalink raw reply related [flat|nested] 25+ messages in thread
* Re: [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection
2024-08-20 7:35 ` Deepak Gupta
@ 2024-08-20 9:20 ` Richard Henderson
2024-08-20 18:55 ` Deepak Gupta
0 siblings, 1 reply; 25+ messages in thread
From: Richard Henderson @ 2024-08-20 9:20 UTC (permalink / raw)
To: Deepak Gupta, qemu-riscv, qemu-devel
Cc: palmer, Alistair.Francis, bmeng.cn, liwei1518, dbarboza,
zhiwei_liu, jim.shu, andy.chiu, kito.cheng
On 8/20/24 17:35, Deepak Gupta wrote:
>> + /* If shadow stack instruction initiated this access, treat it as store */
>> + if (mmu_idx & MMU_IDX_SS_WRITE) {
>> + access_type = MMU_DATA_STORE;
>> + }
>> +
>
> I think I forgot to address this. Do you still want me to fix this up like you
> had suggested?
Yes, this still needs fixing.
> IIRC, you mentioned to use TARGET_INSN_START_EXTRA_WORDS=2. Honestly I don't know
> what it means and how its used. Based on git grep and some readup, are you expecting
> something
> along the below lines?
>
>
> """
>
> diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
> index fee31b8037..dfd2efa941 100644
> --- a/target/riscv/cpu.h
> +++ b/target/riscv/cpu.h
> @@ -47,7 +47,7 @@ typedef struct CPUArchState CPURISCVState;
> * RISC-V-specific extra insn start words:
> * 1: Original instruction opcode
> */
> -#define TARGET_INSN_START_EXTRA_WORDS 1
> +#define TARGET_INSN_START_EXTRA_WORDS 2
>
> #define RV(x) ((target_ulong)1 << (x - 'A'))
>
> diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
> index f74a1216b1..b266177e46 100644
> --- a/target/riscv/cpu_helper.c
> +++ b/target/riscv/cpu_helper.c
> @@ -1271,6 +1271,11 @@ static void raise_mmu_exception(CPURISCVState *env, target_ulong
> address,
> {
> CPUState *cs = env_cpu(env);
>
> + if (!pmp_violation &&
> + tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] & 1) {
> + tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] &= ~1;
> + access_type = MMU_DATA_STORE;
> + }
The first thing to understand is that the unwind data is stored by the compiler and
recovered by the unwinder.
The unwind data is exposed to the target via one of two methods:
(1) TCGCPUOps.restore_state_to_opc, i.e. riscv_restore_state_to_opc.
The data[] argument contains the extra words.
With this method, the extra words are restored to env and are
available in a later call to riscv_cpu_do_interrupt.
Compare env->bins from the first extra word, which is used exactly so.
This is probably the easiest and best option.
You'd promote LOAD* to STORE_AMO* while dispatching the interrupt.
(2) cpu_unwind_state_data()
With this method, you have immediate access to the extra words,
and don't need to store them anywhere else.
This is supposed to be used when we are *not* going to raise
an exception, merely look something up and continue execution.
Otherwise, we'd be performing the unwind operation twice,
and it's not cheap.
So, tcg_ctx->gen_insn_data[] is not something you'd ever touch,
and this is the wrong spot to do anything.
r~
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection
2024-08-20 9:20 ` Richard Henderson
@ 2024-08-20 18:55 ` Deepak Gupta
2024-08-20 19:45 ` Deepak Gupta
2024-08-20 22:33 ` Richard Henderson
0 siblings, 2 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 18:55 UTC (permalink / raw)
To: Richard Henderson
Cc: qemu-riscv, qemu-devel, palmer, Alistair.Francis, bmeng.cn,
liwei1518, dbarboza, zhiwei_liu, jim.shu, andy.chiu, kito.cheng
On Tue, Aug 20, 2024 at 07:20:48PM +1000, Richard Henderson wrote:
>On 8/20/24 17:35, Deepak Gupta wrote:
>>>+ /* If shadow stack instruction initiated this access, treat it as store */
>>>+ if (mmu_idx & MMU_IDX_SS_WRITE) {
>>>+ access_type = MMU_DATA_STORE;
>>>+ }
>>>+
>>
>>I think I forgot to address this. Do you still want me to fix this up like you
>>had suggested?
>
>Yes, this still needs fixing.
>
>
>>IIRC, you mentioned to use TARGET_INSN_START_EXTRA_WORDS=2. Honestly I don't know
>>what it means and how its used. Based on git grep and some readup,
>>are you expecting something
>>along the below lines?
>>
>>
>>"""
>>
>>diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
>>index fee31b8037..dfd2efa941 100644
>>--- a/target/riscv/cpu.h
>>+++ b/target/riscv/cpu.h
>>@@ -47,7 +47,7 @@ typedef struct CPUArchState CPURISCVState;
>> * RISC-V-specific extra insn start words:
>> * 1: Original instruction opcode
>> */
>>-#define TARGET_INSN_START_EXTRA_WORDS 1
>>+#define TARGET_INSN_START_EXTRA_WORDS 2
>>
>> #define RV(x) ((target_ulong)1 << (x - 'A'))
>>
>>diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
>>index f74a1216b1..b266177e46 100644
>>--- a/target/riscv/cpu_helper.c
>>+++ b/target/riscv/cpu_helper.c
>>@@ -1271,6 +1271,11 @@ static void raise_mmu_exception(CPURISCVState
>>*env, target_ulong address,
>> {
>> CPUState *cs = env_cpu(env);
>>
>>+ if (!pmp_violation &&
>>+ tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] & 1) {
>>+ tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] &= ~1;
>>+ access_type = MMU_DATA_STORE;
>>+ }
>
>The first thing to understand is that the unwind data is stored by the
>compiler and recovered by the unwinder.
>
>The unwind data is exposed to the target via one of two methods:
>
>(1) TCGCPUOps.restore_state_to_opc, i.e. riscv_restore_state_to_opc.
> The data[] argument contains the extra words.
>
> With this method, the extra words are restored to env and are
> available in a later call to riscv_cpu_do_interrupt.
> Compare env->bins from the first extra word, which is used exactly so.
>
> This is probably the easiest and best option.
> You'd promote LOAD* to STORE_AMO* while dispatching the interrupt.
>
>(2) cpu_unwind_state_data()
>
> With this method, you have immediate access to the extra words,
> and don't need to store them anywhere else.
>
> This is supposed to be used when we are *not* going to raise
> an exception, merely look something up and continue execution.
> Otherwise, we'd be performing the unwind operation twice,
> and it's not cheap.
>
>So, tcg_ctx->gen_insn_data[] is not something you'd ever touch,
>and this is the wrong spot to do anything.
Thanks for more information and guiding me through this.
Not going to say that I still understand everything. But I looked
at one arm example. Before I do something more with it. I wanted to run
it by you.
Something on the below lines? I've one question as well for you in comment.
""""
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index fee31b8037..b4e04fe849 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -46,8 +46,14 @@ typedef struct CPUArchState CPURISCVState;
/*
* RISC-V-specific extra insn start words:
* 1: Original instruction opcode
+ * 2: more information about instruction
*/
-#define TARGET_INSN_START_EXTRA_WORDS 1
+#define TARGET_INSN_START_EXTRA_WORDS 2
+
+/*
+ * b0: Whether a shadow stack operation/instruction or not.
+ */
+#define RISCV_INSN_START_WORD2_SS_OP 1
#define RV(x) ((target_ulong)1 << (x - 'A'))
@@ -226,6 +232,7 @@ struct CPUArchState {
bool elp;
/* shadow stack register for zicfiss extension */
target_ulong ssp;
+ bool ss_op;
/* sw check code for sw check exception */
target_ulong sw_check_code;
#ifdef CONFIG_USER_ONLY
diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index f74a1216b1..c28b13d42c 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -1777,6 +1777,10 @@ void riscv_cpu_do_interrupt(CPUState *cs)
target_ulong mtval2 = 0;
if (!async) {
+ /* shadow stack op, promote load page fault to store page fault */
+ if (env->ss_op && cause == RISCV_EXCP_LOAD_PAGE_FAULT) {
+ cause = RISCV_EXCP_STORE_PAGE_FAULT;
+ }
/* set tval to badaddr for traps with address information */
switch (cause) {
case RISCV_EXCP_SEMIHOST:
diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
index 4da26cb926..c0f21fe3db 100644
--- a/target/riscv/tcg/tcg-cpu.c
+++ b/target/riscv/tcg/tcg-cpu.c
@@ -129,6 +129,7 @@ static void riscv_restore_state_to_opc(CPUState *cs,
env->pc = pc;
}
env->bins = data[1];
+ env->ss_op = data[2] & RISCV_INSN_START_WORD2_SS_OP;
}
static const TCGCPUOps riscv_tcg_ops = {
diff --git a/target/riscv/translate.c b/target/riscv/translate.c
index 580aa23c5b..6f952db823 100644
--- a/target/riscv/translate.c
+++ b/target/riscv/translate.c
@@ -1271,7 +1271,7 @@ static void riscv_tr_insn_start(DisasContextBase *dcbase, CPUState *cpu)
pc_next &= ~TARGET_PAGE_MASK;
}
- tcg_gen_insn_start(pc_next, 0);
+ tcg_gen_insn_start(pc_next, 0, 0);
ctx->insn_start_updated = false;
}
@@ -1301,6 +1301,14 @@ static void riscv_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
ctx->base.is_jmp = DISAS_NORETURN;
}
+ /* shadow stack index means shadow stack instruction is translated */
+ if (ctx->mem_idx & MMU_IDX_SS_WRITE) {
+ /* Is this needed to set true? */
+ ctx->insn_start_updated = true;
+ tcg_set_insn_start_param(ctx->base.insn_start, 2,
+ RISCV_INSN_START_WORD2_SS_OP);
+ }
+
/* Only the first insn within a TB is allowed to cross a page boundary. */
if (ctx->base.is_jmp == DISAS_NEXT) {
if (ctx->itrigger || !is_same_page(&ctx->base, ctx->base.pc_next)) {
"""
>
>
>r~
^ permalink raw reply related [flat|nested] 25+ messages in thread
* Re: [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection
2024-08-20 18:55 ` Deepak Gupta
@ 2024-08-20 19:45 ` Deepak Gupta
2024-08-20 22:33 ` Richard Henderson
1 sibling, 0 replies; 25+ messages in thread
From: Deepak Gupta @ 2024-08-20 19:45 UTC (permalink / raw)
To: Richard Henderson
Cc: qemu-riscv, qemu-devel, palmer, Alistair.Francis, bmeng.cn,
liwei1518, dbarboza, zhiwei_liu, jim.shu, andy.chiu, kito.cheng
On Tue, Aug 20, 2024 at 11:55 AM Deepak Gupta <debug@rivosinc.com> wrote:
>
> On Tue, Aug 20, 2024 at 07:20:48PM +1000, Richard Henderson wrote:
> >On 8/20/24 17:35, Deepak Gupta wrote:
> >>>+ /* If shadow stack instruction initiated this access, treat it as store */
> >>>+ if (mmu_idx & MMU_IDX_SS_WRITE) {
> >>>+ access_type = MMU_DATA_STORE;
> >>>+ }
> >>>+
> >>
> >>I think I forgot to address this. Do you still want me to fix this up like you
> >>had suggested?
> >
> >Yes, this still needs fixing.
> >
> >
> >>IIRC, you mentioned to use TARGET_INSN_START_EXTRA_WORDS=2. Honestly I don't know
> >>what it means and how its used. Based on git grep and some readup,
> >>are you expecting something
> >>along the below lines?
> >>
> >>
> >>"""
> >>
> >>diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
> >>index fee31b8037..dfd2efa941 100644
> >>--- a/target/riscv/cpu.h
> >>+++ b/target/riscv/cpu.h
> >>@@ -47,7 +47,7 @@ typedef struct CPUArchState CPURISCVState;
> >> * RISC-V-specific extra insn start words:
> >> * 1: Original instruction opcode
> >> */
> >>-#define TARGET_INSN_START_EXTRA_WORDS 1
> >>+#define TARGET_INSN_START_EXTRA_WORDS 2
> >>
> >> #define RV(x) ((target_ulong)1 << (x - 'A'))
> >>
> >>diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
> >>index f74a1216b1..b266177e46 100644
> >>--- a/target/riscv/cpu_helper.c
> >>+++ b/target/riscv/cpu_helper.c
> >>@@ -1271,6 +1271,11 @@ static void raise_mmu_exception(CPURISCVState
> >>*env, target_ulong address,
> >> {
> >> CPUState *cs = env_cpu(env);
> >>
> >>+ if (!pmp_violation &&
> >>+ tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] & 1) {
> >>+ tcg_ctx->gen_insn_data[TARGET_INSN_START_EXTRA_WORDS-1] &= ~1;
> >>+ access_type = MMU_DATA_STORE;
> >>+ }
> >
> >The first thing to understand is that the unwind data is stored by the
> >compiler and recovered by the unwinder.
> >
> >The unwind data is exposed to the target via one of two methods:
> >
> >(1) TCGCPUOps.restore_state_to_opc, i.e. riscv_restore_state_to_opc.
> > The data[] argument contains the extra words.
> >
> > With this method, the extra words are restored to env and are
> > available in a later call to riscv_cpu_do_interrupt.
> > Compare env->bins from the first extra word, which is used exactly so.
> >
> > This is probably the easiest and best option.
> > You'd promote LOAD* to STORE_AMO* while dispatching the interrupt.
> >
> >(2) cpu_unwind_state_data()
> >
> > With this method, you have immediate access to the extra words,
> > and don't need to store them anywhere else.
> >
> > This is supposed to be used when we are *not* going to raise
> > an exception, merely look something up and continue execution.
> > Otherwise, we'd be performing the unwind operation twice,
> > and it's not cheap.
> >
> >So, tcg_ctx->gen_insn_data[] is not something you'd ever touch,
> >and this is the wrong spot to do anything.
>
> Thanks for more information and guiding me through this.
>
> Not going to say that I still understand everything. But I looked
> at one arm example. Before I do something more with it. I wanted to run
> it by you.
>
> Something on the below lines? I've one question as well for you in comment.
>
> """"
> diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
> index fee31b8037..b4e04fe849 100644
> --- a/target/riscv/cpu.h
> +++ b/target/riscv/cpu.h
> @@ -46,8 +46,14 @@ typedef struct CPUArchState CPURISCVState;
> /*
> * RISC-V-specific extra insn start words:
> * 1: Original instruction opcode
> + * 2: more information about instruction
> */
> -#define TARGET_INSN_START_EXTRA_WORDS 1
> +#define TARGET_INSN_START_EXTRA_WORDS 2
> +
> +/*
> + * b0: Whether a shadow stack operation/instruction or not.
> + */
> +#define RISCV_INSN_START_WORD2_SS_OP 1
>
> #define RV(x) ((target_ulong)1 << (x - 'A'))
>
> @@ -226,6 +232,7 @@ struct CPUArchState {
> bool elp;
> /* shadow stack register for zicfiss extension */
> target_ulong ssp;
> + bool ss_op;
> /* sw check code for sw check exception */
> target_ulong sw_check_code;
> #ifdef CONFIG_USER_ONLY
> diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
> index f74a1216b1..c28b13d42c 100644
> --- a/target/riscv/cpu_helper.c
> +++ b/target/riscv/cpu_helper.c
> @@ -1777,6 +1777,10 @@ void riscv_cpu_do_interrupt(CPUState *cs)
> target_ulong mtval2 = 0;
>
> if (!async) {
> + /* shadow stack op, promote load page fault to store page fault */
> + if (env->ss_op && cause == RISCV_EXCP_LOAD_PAGE_FAULT) {
> + cause = RISCV_EXCP_STORE_PAGE_FAULT;
> + }
> /* set tval to badaddr for traps with address information */
> switch (cause) {
> case RISCV_EXCP_SEMIHOST:
> diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
> index 4da26cb926..c0f21fe3db 100644
> --- a/target/riscv/tcg/tcg-cpu.c
> +++ b/target/riscv/tcg/tcg-cpu.c
> @@ -129,6 +129,7 @@ static void riscv_restore_state_to_opc(CPUState *cs,
> env->pc = pc;
> }
> env->bins = data[1];
> + env->ss_op = data[2] & RISCV_INSN_START_WORD2_SS_OP;
> }
>
> static const TCGCPUOps riscv_tcg_ops = {
> diff --git a/target/riscv/translate.c b/target/riscv/translate.c
> index 580aa23c5b..6f952db823 100644
> --- a/target/riscv/translate.c
> +++ b/target/riscv/translate.c
> @@ -1271,7 +1271,7 @@ static void riscv_tr_insn_start(DisasContextBase *dcbase, CPUState *cpu)
> pc_next &= ~TARGET_PAGE_MASK;
> }
>
> - tcg_gen_insn_start(pc_next, 0);
> + tcg_gen_insn_start(pc_next, 0, 0);
> ctx->insn_start_updated = false;
> }
>
> @@ -1301,6 +1301,14 @@ static void riscv_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
> ctx->base.is_jmp = DISAS_NORETURN;
> }
>
> + /* shadow stack index means shadow stack instruction is translated */
> + if (ctx->mem_idx & MMU_IDX_SS_WRITE) {
This is a wrong check because we never store MMU_IDX_SS_WRITE in `ctx->mem_idx`
But I can place an indicator in `DisasContext` and set it up in
`trans_xxx` for shadow stack instr.
> + /* Is this needed to set true? */
> + ctx->insn_start_updated = true;
> + tcg_set_insn_start_param(ctx->base.insn_start, 2,
> + RISCV_INSN_START_WORD2_SS_OP);
> + }
> +
> /* Only the first insn within a TB is allowed to cross a page boundary. */
> if (ctx->base.is_jmp == DISAS_NEXT) {
> if (ctx->itrigger || !is_same_page(&ctx->base, ctx->base.pc_next)) {
>
> """
> >
> >
> >r~
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection
2024-08-20 18:55 ` Deepak Gupta
2024-08-20 19:45 ` Deepak Gupta
@ 2024-08-20 22:33 ` Richard Henderson
1 sibling, 0 replies; 25+ messages in thread
From: Richard Henderson @ 2024-08-20 22:33 UTC (permalink / raw)
To: Deepak Gupta
Cc: qemu-riscv, qemu-devel, palmer, Alistair.Francis, bmeng.cn,
liwei1518, dbarboza, zhiwei_liu, jim.shu, andy.chiu, kito.cheng
On 8/21/24 04:55, Deepak Gupta wrote:
> Something on the below lines? I've one question as well for you in comment.
>
> """"
> diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
> index fee31b8037..b4e04fe849 100644
> --- a/target/riscv/cpu.h
> +++ b/target/riscv/cpu.h
> @@ -46,8 +46,14 @@ typedef struct CPUArchState CPURISCVState;
> /*
> * RISC-V-specific extra insn start words:
> * 1: Original instruction opcode
> + * 2: more information about instruction
> */
> -#define TARGET_INSN_START_EXTRA_WORDS 1
> +#define TARGET_INSN_START_EXTRA_WORDS 2
> +
> +/*
> + * b0: Whether a shadow stack operation/instruction or not.
> + */
> +#define RISCV_INSN_START_WORD2_SS_OP 1
Ah, here: not shadow-stack specific. Set for any insn which should always generate
STORE_AMO, including the actual AMO instructions. It's a current emulation error, IIRC.
> @@ -226,6 +232,7 @@ struct CPUArchState {
> bool elp;
> /* shadow stack register for zicfiss extension */
> target_ulong ssp;
> + bool ss_op;
For generality, maybe just store the whole word as excp_uw2?
> if (!async) {
> + /* shadow stack op, promote load page fault to store page fault */
> + if (env->ss_op && cause == RISCV_EXCP_LOAD_PAGE_FAULT) {
> + cause = RISCV_EXCP_STORE_PAGE_FAULT;
> + }
> /* set tval to badaddr for traps with address information */
> switch (cause) {
> case RISCV_EXCP_SEMIHOST:
case RISCV_EXCP_LOAD_GUEST_ACCESS_FAULT:
if (env->excp_uw2 & RISCV_UW2_ALWAYS_STORE_AMO) {
cause = RISCV_EXCP_STORE_GUEST_AMO_ACCESS_FAULT;
}
goto load_store_fault;
case RISCV_EXCP_LOAD_ACCESS_FAULT:
...
case RISCV_EXCP_LOAD_PAGE_FAULT:
...
case RISCV_EXCP_STORE_PAGE_FAULT:
load_store_fault:
> @@ -1301,6 +1301,14 @@ static void riscv_tr_translate_insn(DisasContextBase *dcbase,
> CPUState *cpu)
> ctx->base.is_jmp = DISAS_NORETURN;
> }
>
> + /* shadow stack index means shadow stack instruction is translated */
> + if (ctx->mem_idx & MMU_IDX_SS_WRITE) {
> + /* Is this needed to set true? */
> + ctx->insn_start_updated = true;
> + tcg_set_insn_start_param(ctx->base.insn_start, 2,
> + RISCV_INSN_START_WORD2_SS_OP);
> + }
No, SS_WRITE is never part of mem_idx, and setting insn_start_updated here would break things.
You'll want to change decode_save_opcode() to take the second parameter (or introduce a
new helper for the second parameter, leaving decode_save_opcode alone). But you do have
to handle the update on a per-insn basis.
r~
^ permalink raw reply [flat|nested] 25+ messages in thread
end of thread, other threads:[~2024-08-20 22:34 UTC | newest]
Thread overview: 25+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2024-08-20 0:01 [PATCH v5 00/15] riscv support for control flow integrity extensions Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 01/15] target/riscv: Add zicfilp extension Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 02/15] target/riscv: Introduce elp state and enabling controls for zicfilp Deepak Gupta
2024-08-20 5:17 ` Richard Henderson
2024-08-20 0:01 ` [PATCH v5 03/15] target/riscv: save and restore elp state on priv transitions Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 04/15] target/riscv: additional code information for sw check Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 05/15] target/riscv: tracking indirect branches (fcfi) for zicfilp Deepak Gupta
2024-08-20 5:24 ` Richard Henderson
2024-08-20 0:01 ` [PATCH v5 06/15] target/riscv: zicfilp `lpad` impl and branch tracking Deepak Gupta
2024-08-20 5:29 ` Richard Henderson
2024-08-20 0:01 ` [PATCH v5 07/15] disas/riscv: enable `lpad` disassembly Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 08/15] target/riscv: Add zicfiss extension Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 09/15] target/riscv: introduce ssp and enabling controls for zicfiss Deepak Gupta
2024-08-20 5:34 ` Richard Henderson
2024-08-20 0:01 ` [PATCH v5 10/15] target/riscv: tb flag for shadow stack instructions Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 11/15] target/riscv: mmu changes for zicfiss shadow stack protection Deepak Gupta
2024-08-20 7:35 ` Deepak Gupta
2024-08-20 9:20 ` Richard Henderson
2024-08-20 18:55 ` Deepak Gupta
2024-08-20 19:45 ` Deepak Gupta
2024-08-20 22:33 ` Richard Henderson
2024-08-20 0:01 ` [PATCH v5 12/15] target/riscv: implement zicfiss instructions Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 13/15] target/riscv: compressed encodings for sspush and sspopchk Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 14/15] disas/riscv: enable disassembly for zicfiss instructions Deepak Gupta
2024-08-20 0:01 ` [PATCH v5 15/15] disas/riscv: enable disassembly for compressed sspush/sspopchk Deepak Gupta
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).