* Re: [PATCH 2/2] arm64: Check if GMID_EL1.BS is the same on all CPUs
[not found] ` <20210511182322.3830-3-catalin.marinas@arm.com>
@ 2021-05-26 19:47 ` Jon Hunter
2021-05-26 21:08 ` Will Deacon
2021-05-26 21:48 ` Catalin Marinas
0 siblings, 2 replies; 4+ messages in thread
From: Jon Hunter @ 2021-05-26 19:47 UTC (permalink / raw)
To: Catalin Marinas, linux-arm-kernel
Cc: Will Deacon, Mark Rutland, Suzuki K Poulose, linux-tegra
On 11/05/2021 19:23, Catalin Marinas wrote:
> The GMID_EL1.BS field determines the number of tags accessed by the
> LDGM/STGM instructions (EL1 and up), used by the kernel for copying or
> zeroing page tags.
>
> Taint the kernel if GMID_EL1.BS differs between CPUs.
>
> Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
> Cc: Will Deacon <will@kernel.org>
> Cc: Mark Rutland <mark.rutland@arm.com>
> Cc: Suzuki K Poulose <Suzuki.Poulose@arm.com>
> ---
> arch/arm64/include/asm/cpu.h | 1 +
> arch/arm64/kernel/cpufeature.c | 17 +++++++++++++++++
> arch/arm64/kernel/cpuinfo.c | 1 +
> 3 files changed, 19 insertions(+)
>
> diff --git a/arch/arm64/include/asm/cpu.h b/arch/arm64/include/asm/cpu.h
> index fe5a8499ddc2..9088e72c7cf6 100644
> --- a/arch/arm64/include/asm/cpu.h
> +++ b/arch/arm64/include/asm/cpu.h
> @@ -20,6 +20,7 @@ struct cpuinfo_arm64 {
> u64 reg_dczid;
> u64 reg_midr;
> u64 reg_revidr;
> + u64 reg_gmid;
>
> u64 reg_id_aa64dfr0;
> u64 reg_id_aa64dfr1;
> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> index ca66a61bb396..3b9089ca52dc 100644
> --- a/arch/arm64/kernel/cpufeature.c
> +++ b/arch/arm64/kernel/cpufeature.c
> @@ -401,6 +401,11 @@ static const struct arm64_ftr_bits ftr_dczid[] = {
> ARM64_FTR_END,
> };
>
> +static const struct arm64_ftr_bits ftr_gmid[] = {
> + ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, SYS_GMID_EL1_BS_SHIFT, 4, 0),
> + ARM64_FTR_END,
> +};
> +
> static const struct arm64_ftr_bits ftr_id_isar0[] = {
> ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_ISAR0_DIVIDE_SHIFT, 4, 0),
> ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_ISAR0_DEBUG_SHIFT, 4, 0),
> @@ -618,6 +623,9 @@ static const struct __ftr_reg_entry {
> /* Op1 = 0, CRn = 1, CRm = 2 */
> ARM64_FTR_REG(SYS_ZCR_EL1, ftr_zcr),
>
> + /* Op1 = 1, CRn = 0, CRm = 0 */
> + ARM64_FTR_REG(SYS_GMID_EL1, ftr_gmid),
> +
> /* Op1 = 3, CRn = 0, CRm = 0 */
> { SYS_CTR_EL0, &arm64_ftr_reg_ctrel0 },
> ARM64_FTR_REG(SYS_DCZID_EL0, ftr_dczid),
> @@ -872,6 +880,7 @@ void __init init_cpu_features(struct cpuinfo_arm64 *info)
> init_cpu_ftr_reg(SYS_CTR_EL0, info->reg_ctr);
> init_cpu_ftr_reg(SYS_DCZID_EL0, info->reg_dczid);
> init_cpu_ftr_reg(SYS_CNTFRQ_EL0, info->reg_cntfrq);
> + init_cpu_ftr_reg(SYS_GMID_EL1, info->reg_gmid);
> init_cpu_ftr_reg(SYS_ID_AA64DFR0_EL1, info->reg_id_aa64dfr0);
> init_cpu_ftr_reg(SYS_ID_AA64DFR1_EL1, info->reg_id_aa64dfr1);
> init_cpu_ftr_reg(SYS_ID_AA64ISAR0_EL1, info->reg_id_aa64isar0);
> @@ -1082,6 +1091,14 @@ void update_cpu_features(int cpu,
> taint |= check_update_ftr_reg(SYS_DCZID_EL0, cpu,
> info->reg_dczid, boot->reg_dczid);
>
> + /*
> + * The kernel uses the LDGM/STGM instructions and the number of tags
> + * they read/write depends on the GMID_EL1.BS field. Check that the
> + * value is the same on all CPUs.
> + */
> + taint |= check_update_ftr_reg(SYS_GMID_EL1, cpu,
> + info->reg_gmid, boot->reg_gmid);
> +
> /* If different, timekeeping will be broken (especially with KVM) */
> taint |= check_update_ftr_reg(SYS_CNTFRQ_EL0, cpu,
> info->reg_cntfrq, boot->reg_cntfrq);
> diff --git a/arch/arm64/kernel/cpuinfo.c b/arch/arm64/kernel/cpuinfo.c
> index 4bea701117d4..cd9f2d51285b 100644
> --- a/arch/arm64/kernel/cpuinfo.c
> +++ b/arch/arm64/kernel/cpuinfo.c
> @@ -359,6 +359,7 @@ static void __cpuinfo_store_cpu(struct cpuinfo_arm64 *info)
> info->reg_dczid = read_cpuid(DCZID_EL0);
> info->reg_midr = read_cpuid_id();
> info->reg_revidr = read_cpuid(REVIDR_EL1);
> + info->reg_gmid = read_cpuid(GMID_EL1);
>
> info->reg_id_aa64dfr0 = read_cpuid(ID_AA64DFR0_EL1);
> info->reg_id_aa64dfr1 = read_cpuid(ID_AA64DFR1_EL1);
I am seeing the following undefined instruction crash on all our
ARM64 Tegra devices on today's -next and bisect is pointing to
this patch. Reverting this patch on top of -next does fix the
problem. Let me know if you have any thoughts.
Thanks!
Jon
[ 0.000000] ------------[ cut here ]------------
[ 0.000000] kernel BUG at /dvs/git/dirty/git-master_l4t-upstream/kernel/arch/arm64/kernel/traps.c:406!
[ 0.000000] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[ 0.000000] Modules linked in:
[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 5.13.0-rc3-next-20210526-gf6b46ef27317 #1
[ 0.000000] Hardware name: NVIDIA Jetson TX2 Developer Kit (DT)
[ 0.000000] pstate: 600000c5 (nZCv daIF -PAN -UAO -TCO BTYPE=--)
[ 0.000000] pc : do_undefinstr+0x298/0x2b0
[ 0.000000] lr : do_undefinstr+0x2a8/0x2b0
[ 0.000000] sp : ffff800011cb3c10
[ 0.000000] x29: ffff800011cb3c10 x28: ffff800011cc3540 x27: 0000000000000002
[ 0.000000] x26: ffff800011760008 x25: ffff0001f4deccc0 x24: ffff800011cb9000
[ 0.000000] x23: ffff800011fb50f8 x22: ffff800011fb5000 x21: 00000000d5390080
[ 0.000000] x20: ffff800011cb3c90 x19: ffff800011cb9000 x18: ffffffffffffffff
[ 0.000000] x17: 0000000000017000 x16: 0000000000000000 x15: 000000000000001e
[ 0.000000] x14: 0000000000200000 x13: 0000000275e00000 x12: 0000001000000000
[ 0.000000] x11: 00000000009fc580 x10: 0000000274e03a80 x9 : 0000001000000000
[ 0.000000] x8 : 0000000000200000 x7 : 0000000000000003 x6 : 0000000000000000
[ 0.000000] x5 : ffff800011cc5910 x4 : 0000000000000000 x3 : ffff800011fb50f8
[ 0.000000] x2 : 0000000000000000 x1 : ffff800011cc3540 x0 : 0000000000000005
[ 0.000000] Call trace:
[ 0.000000] do_undefinstr+0x298/0x2b0
[ 0.000000] el1_undef+0x2c/0x48
[ 0.000000] el1_sync_handler+0xb4/0xd0
[ 0.000000] el1_sync+0x74/0x100
[ 0.000000] __cpuinfo_store_cpu+0x5c/0x248
[ 0.000000] cpuinfo_store_boot_cpu+0x28/0x54
[ 0.000000] smp_prepare_boot_cpu+0x2c/0x38
[ 0.000000] start_kernel+0x1a4/0x62c
[ 0.000000] __primary_switched+0x8c/0x90
[ 0.000000] Code: b5fffe40 b94047b5 17ffffca d503201f (d4210000)
[ 0.000000] random: get_random_bytes called from print_oops_end_marker+0x4c/0x68 with crng_init=0
[ 0.000000] ---[ end trace 0000000000000000 ]---
--
nvpublic
^ permalink raw reply [flat|nested] 4+ messages in thread