Skip to contentSkip to conversation

target/i386: helper_sysret(): Check that RCX contains a canonical address when emulating an Intel CPU

PATCH Applied Andrey Polivoda 3 participants 5 messages Jun 8 – Jul 24

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 target/i386: helper_sysret(): Check that RCX contains a canonical address when emulating an Intel CPU

Intel and AMD CPUs implement SYSRETQ instruction differently.
One of these differences is whether a canonicality check of the address that will be loaded to RIP is performed: Intel CPUs do this check, AMD CPUs don't.

Currently, QEMU does not perform this check when emulating Intel CPUs. This patch corrects this by implementing the canonlicality check on a new RIP value from RCX and performing it only when emulating Intel CPUs.

Flags and segment registers' caches are updated only after checking the new RIP value to ensure that CPU state is not modified in case the #GP(0) exception is raised due to the check failure.

Cc Paolo Bonzini <[email protected]>
Cc Richard Henderson <[email protected]>
Fixes 14ce26e75513 ("x86_64 target support")

Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3223

---
 target/i386/tcg/seg_helper.c | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)
target/i386/tcg/seg_helper.c +13 -4
index 58aac72011..d5c174b7fd 100644
@@ -1096,17 +1096,22 @@ void helper_sysret(CPUX86State *env, int dflag)
     selector = (env->star >> 48) & 0xffff;
 #ifdef TARGET_X86_64
     if (env->hflags & HF_LMA_MASK) {
-        cpu_load_eflags(env, (uint32_t)(env->regs[11]), TF_MASK | AC_MASK
-                        | ID_MASK | IF_MASK | IOPL_MASK | VM_MASK | RF_MASK |
-                        NT_MASK);
         if (dflag == 2) {
+            uint64_t new_rip = env->regs[R_ECX];
+            if (IS_INTEL_CPU(env)) {
+                int shift = (get_pg_mode(env) & PG_MODE_LA57) ? 56 : 47;
+                int64_t sext = (int64_t)new_rip >> shift;
+                if (sext != 0 && sext != -1) {
+                    raise_exception_err_ra(env, EXCP0D_GPF, 0, GETPC());
+                }
+            }
             cpu_x86_load_seg_cache(env, R_CS, (selector + 16) | 3,
                                    0, 0xffffffff,
                                    DESC_G_MASK | DESC_P_MASK |
                                    DESC_S_MASK | (3 << DESC_DPL_SHIFT) |
                                    DESC_CS_MASK | DESC_R_MASK | DESC_A_MASK |
                                    DESC_L_MASK);
-            env->eip = env->regs[R_ECX];
+            env->eip = new_rip;
         } else {
             cpu_x86_load_seg_cache(env, R_CS, selector | 3,
                                    0, 0xffffffff,
@@ -1120,6 +1125,10 @@ void helper_sysret(CPUX86State *env, int dflag)
                                DESC_G_MASK | DESC_B_MASK | DESC_P_MASK |
                                DESC_S_MASK | (3 << DESC_DPL_SHIFT) |
                                DESC_W_MASK | DESC_A_MASK);
+
+        cpu_load_eflags(env, (uint32_t)(env->regs[11]), TF_MASK | AC_MASK
+                        | ID_MASK | IF_MASK | IOPL_MASK | VM_MASK | RF_MASK |
+                        NT_MASK);
     } else
 #endif
     {
Signature · 1 line
2.53.0
author
#1 lore raw

ping
https://lore.kernel.org/qemu-devel/[email protected]/T/#u

пн, 8 июн. 2026 г., 19:19 Andrey Polivoda <[email protected]>:

Quoted 75 lines

Intel and AMD CPUs implement SYSRETQ instruction differently. One of these differences is whether a canonicality check of the address that
will be loaded to RIP is performed: Intel CPUs do this check, AMD CPUs don't.

Currently, QEMU does not perform this check when emulating Intel CPUs. This patch corrects this by implementing the canonlicality check on a new RIP
value from RCX and performing it only when emulating Intel CPUs.

Flags and segment registers' caches are updated only after checking the new RIP
value to ensure that CPU state is not modified in case the #GP(0) exception is raised due to the check failure.

Cc Paolo Bonzini <[email protected]>
Cc Richard Henderson <[email protected]>
Fixes 14ce26e75513 ("x86_64 target support")

Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3223

---
 target/i386/tcg/seg_helper.c | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)
target/i386/tcg/seg_helper.c +0 -2
index 58aac72011..d5c174b7fd 100644
@@ -1096,17 +1096,22 @@ void helper_sysret(CPUX86State *env, int dflag)
     selector = (env->star >> 48) & 0xffff;
 #ifdef TARGET_X86_64
     if (env->hflags & HF_LMA_MASK) {
-        cpu_load_eflags(env, (uint32_t)(env->regs[11]), TF_MASK | AC_MASK
-                        | ID_MASK | IF_MASK | IOPL_MASK | VM_MASK |

RF_MASK |

target/i386/tcg/seg_helper.c +8 -1
-                        NT_MASK);
         if (dflag == 2) {
+            uint64_t new_rip = env->regs[R_ECX];
+            if (IS_INTEL_CPU(env)) {
+                int shift = (get_pg_mode(env) & PG_MODE_LA57) ? 56 : 47;
+                int64_t sext = (int64_t)new_rip >> shift;
+                if (sext != 0 && sext != -1) {
+                    raise_exception_err_ra(env, EXCP0D_GPF, 0, GETPC());
+                }
+            }
             cpu_x86_load_seg_cache(env, R_CS, (selector + 16) | 3,
                                    0, 0xffffffff,
                                    DESC_G_MASK | DESC_P_MASK |
                                    DESC_S_MASK | (3 << DESC_DPL_SHIFT) |
                                    DESC_CS_MASK | DESC_R_MASK |

DESC_A_MASK |

                                    DESC_L_MASK);
target/i386/tcg/seg_helper.c +4 -1
-            env->eip = env->regs[R_ECX];
+            env->eip = new_rip;
         } else {
             cpu_x86_load_seg_cache(env, R_CS, selector | 3,
                                    0, 0xffffffff,
@@ -1120,6 +1125,10 @@ void helper_sysret(CPUX86State *env, int dflag)
                                DESC_G_MASK | DESC_B_MASK | DESC_P_MASK |
                                DESC_S_MASK | (3 << DESC_DPL_SHIFT) |
                                DESC_W_MASK | DESC_A_MASK);
+
+        cpu_load_eflags(env, (uint32_t)(env->regs[11]), TF_MASK | AC_MASK
+                        | ID_MASK | IF_MASK | IOPL_MASK | VM_MASK |

RF_MASK |

+                        NT_MASK);
     } else
 #endif
     {

--
2.53.0

#4 lore raw
In reply to Andrey Polivoda
Quoted 12 lines · Andrey Polivoda

Intel and AMD CPUs implement SYSRETQ instruction differently.
One of these differences is whether a canonicality check of the address that will be loaded to RIP is performed: Intel CPUs do this check, AMD CPUs don't.

Currently, QEMU does not perform this check when emulating Intel CPUs. This patch corrects this by implementing the canonlicality check on a new RIP value from RCX and performing it only when emulating Intel CPUs.

Flags and segment registers' caches are updated only after checking the new RIP value to ensure that CPU state is not modified in case the #GP(0) exception is raised due to the check failure.

Quoted 6 lines
Cc Paolo Bonzini <[email protected]>
Cc Richard Henderson <[email protected]>
Fixes 14ce26e75513 ("x86_64 target support")

Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3223

I'm picking it up for the stable series. This issue has been with us since the day one, but since there's a bug/issue filed about it, it bothers people.

Please let me know if I shouldn't pick it up.

Thanks,

/mjt