[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[PATCH 01/26] target/ppc: Fix crash on machine check caused by ifetch
From: |
Nicholas Piggin |
Subject: |
[PATCH 01/26] target/ppc: Fix crash on machine check caused by ifetch |
Date: |
Fri, 19 Jan 2024 01:06:19 +1000 |
is_prefix_insn_excp() loads the first word of the instruction address
which caused an exception, to determine whether or not it was prefixed
so the prefix bit can be set in [H]SRR1.
This works if the instruction image can be loaded, but if the exception
was caused by an ifetch, this load could fail and cause a recursive
exception and crash. Machine checks caused by ifetch are not excluded
from the prefix check and can crash (see issue 2108 for an example).
Fix this by excluding machine checks caused by ifetch from the prefix
check.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2108
Fixes: 55a7fa34f89 ("target/ppc: Machine check on invalid real address access
on POWER9/10")
Fixes: 5a5d3b23cb2 ("target/ppc: Add SRR1 prefix indication to interrupt
handlers")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
target/ppc/excp_helper.c | 32 +++++++++++++++++++++-----------
1 file changed, 21 insertions(+), 11 deletions(-)
diff --git a/target/ppc/excp_helper.c b/target/ppc/excp_helper.c
index 2ec6429e36..030260e8a9 100644
--- a/target/ppc/excp_helper.c
+++ b/target/ppc/excp_helper.c
@@ -1322,6 +1322,15 @@ static bool is_prefix_insn_excp(PowerPCCPU *cpu, int
excp)
}
switch (excp) {
+ case POWERPC_EXCP_MCHECK:
+ if (!(env->error_code & PPC_BIT(42))) {
+ /*
+ * Fetch attempt caused a machine check, so attempting to fetch
+ * again would cause a recursive machine check.
+ */
+ return false;
+ }
+ break;
case POWERPC_EXCP_HDSI:
/* HDSI PRTABLE_FAULT has the originating access type in error_code */
if ((env->spr[SPR_HDSISR] & DSISR_PRTABLE_FAULT) &&
@@ -1332,10 +1341,10 @@ static bool is_prefix_insn_excp(PowerPCCPU *cpu, int
excp)
* instruction at NIP would cause recursive faults with the same
* translation).
*/
- break;
+ return false;
}
- /* fall through */
- case POWERPC_EXCP_MCHECK:
+ break;
+
case POWERPC_EXCP_DSI:
case POWERPC_EXCP_DSEG:
case POWERPC_EXCP_ALIGN:
@@ -1346,17 +1355,14 @@ static bool is_prefix_insn_excp(PowerPCCPU *cpu, int
excp)
case POWERPC_EXCP_VPU:
case POWERPC_EXCP_VSXU:
case POWERPC_EXCP_FU:
- case POWERPC_EXCP_HV_FU: {
- uint32_t insn = ppc_ldl_code(env, env->nip);
- if (is_prefix_insn(env, insn)) {
- return true;
- }
+ case POWERPC_EXCP_HV_FU:
break;
- }
default:
- break;
+ return false;
}
- return false;
+
+
+ return is_prefix_insn(env, ppc_ldl_code(env, env->nip));
}
#else
static bool is_prefix_insn_excp(PowerPCCPU *cpu, int excp)
@@ -3245,6 +3251,10 @@ void ppc_cpu_do_transaction_failed(CPUState *cs, hwaddr
physaddr,
env->error_code |= PPC_BIT(42);
} else { /* Fetch */
+ /*
+ * is_prefix_insn_excp() tests !PPC_BIT(42) to avoid fetching
+ * the instruction, so that must always be clear for fetches.
+ */
env->error_code = PPC_BIT(36) | PPC_BIT(44) | PPC_BIT(45);
}
break;
--
2.42.0
- [PATCH 00/26] target/ppc: TCG improvements and fixes, Nicholas Piggin, 2024/01/18
- [PATCH 01/26] target/ppc: Fix crash on machine check caused by ifetch,
Nicholas Piggin <=
- [PATCH 02/26] target/ppc: Prevent supervisor from modifying MSR[ME], Nicholas Piggin, 2024/01/18
- [PATCH 03/26] spapr: set MSR[ME] and MSR[FP] on client entry, Nicholas Piggin, 2024/01/18
- [PATCH 04/26] target/ppc: Rename registers to match ISA, Nicholas Piggin, 2024/01/18
- [PATCH 05/26] target/ppc: Update gdbstub to read SPR's CFAR, DEC, HDEC, TB-L/U, Nicholas Piggin, 2024/01/18
- [PATCH 06/26] target/ppc: Rename TBL to TB on 64-bit, Nicholas Piggin, 2024/01/18
- [PATCH 07/26] target/ppc: Improve timebase register defines naming, Nicholas Piggin, 2024/01/18
- [PATCH 08/26] target/ppc: Fix move-to timebase SPR access permissions, Nicholas Piggin, 2024/01/18
- [PATCH 09/26] pnv/chiptod: Add POWER9/10 chiptod model, Nicholas Piggin, 2024/01/18
- [PATCH 10/26] ppc/pnv: Wire ChipTOD model to powernv9 and powernv10 machines, Nicholas Piggin, 2024/01/18
- [PATCH 11/26] pnv/chiptod: Implement the ChipTOD to Core transfer, Nicholas Piggin, 2024/01/18