KVM: arm64: Fix DEMUX register access
cover KVM: arm64: Fix DEMUX register access
The first patch is a fix, but not one likely to ever truly be needed, as it's unlikely to find seven levels of cache. The bug was found while code reading. Writing the second patch was actually why I was reading the code. The issue being fixed for the get-reg-list test was found when running it on a different machine than what was used to develop it.
Thanks,
drew
Andrew Jones (2):
KVM: arm64: CSSELR_EL1 max is 13
KVM: selftests: Filter out DEMUX registers
arch/arm64/kvm/sys_regs.c | 2 +- .../selftests/kvm/aarch64/get-reg-list.c | 39 ++++++++++++++----- 2 files changed, 31 insertions(+), 10 deletions(-)
Signature · 6 lines
2.26.2 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH 1/2 KVM: arm64: CSSELR_EL1 max is 13
Not counting TnD, which KVM doesn't currently consider, CSSELR_EL1 can have a maximum value of 0b1101 (13), which corresponds to an instruction cache at level 7. With CSSELR_MAX set to 12 we can only select up to cache level 6. Change it to 14.
--- arch/arm64/kvm/sys_regs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/kvm/sys_regs.c +1 -1
@@ -169,7 +169,7 @@ void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg)static u32 cache_levels;/* CSSELR values; used to index KVM_REG_ARM_DEMUX_ID_CCSIDR */ -#define CSSELR_MAX 12 +#define CSSELR_MAX 14/* Which cache CCSIDR represents depends on CSSELR value. */ static u32 get_ccsidr (u32 csselr)
Signature · 6 lines
2.26.2 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH 1/2] KVM: arm64: CSSELR_EL1 max is 13
Quoted 23 lines · Andrew Jones
Not counting TnD, which KVM doesn't currently consider,
CSSELR_EL1can have a maximum value of 0b1101 (13), which corresponds to an instruction cache at level 7. WithCSSELR_MAXset to 12 we can only select up to cache level 6. Change it to 14.Signed-off-by Andrew Jones <[email protected]>--- arch/arm64/kvm/sys_regs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/arm64/kvm/sys_regs.c +1
-1@@ -169,7 +169,7 @@ void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg)static u32 cache_levels;/* CSSELR values; used to index KVM_REG_ARM_DEMUX_ID_CCSIDR */ -#define CSSELR_MAX 12 +#define CSSELR_MAX 14/* Which cache CCSIDR represents depends on CSSELR value. */
Quoted 1 line
static u32
get_ccsidr(u32 csselr)
Huh, nice catch. Do we need a CC: stable tag for this?
Thanks,
M.
Signature · 5 lines
Jazz is not dead. It just smells funny... _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 1/2] KVM: arm64: CSSELR_EL1 max is 13
Quoted 19 lines · Marc Zyngier
Quoted 18 lines · Andrew Jones
Not counting TnD, which KVM doesn't currently consider,
CSSELR_EL1can have a maximum value of 0b1101 (13), which corresponds to an instruction cache at level 7. WithCSSELR_MAXset to 12 we can only select up to cache level 6. Change it to 14.Signed-off-by Andrew Jones <[email protected]>--- arch/arm64/kvm/sys_regs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/arm64/kvm/sys_regs.c +0
-0@@ -169,7 +169,7 @@ void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg)static u32 cache_levels;
Quoted 9 lines
Quoted 6 lines
arch/arm64/kvm/sys_regs.c +1
-1/* CSSELR values; used to index KVM_REG_ARM_DEMUX_ID_CCSIDR */ -#define CSSELR_MAX 12 +#define CSSELR_MAX 14/* Which cache CCSIDR represents depends on CSSELR value. */ static u32
get_ccsidr(u32 csselr)Huh, nice catch. Do we need a CC: stable tag for this?
Hi Marc,
I wasn't thinking so, because I'm not expecting there to actually be hardware with seven cache levels in the wild any time soon. You have more knowledge about what's out there and coming, though, so feel free CC stable if needed.
Thanks,
drew
Marc Zyngier
maintainerRe: [PATCH 1/2] KVM: arm64: CSSELR_EL1 max is 13
Quoted 30 lines · Andrew Jones
Quoted 19 lines · Marc Zyngier
Quoted 18 lines · Andrew Jones
Not counting TnD, which KVM doesn't currently consider,
CSSELR_EL1can have a maximum value of 0b1101 (13), which corresponds to an instruction cache at level 7. WithCSSELR_MAXset to 12 we can only select up to cache level 6. Change it to 14.Signed-off-by Andrew Jones <[email protected]>--- arch/arm64/kvm/sys_regs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/arm64/kvm/sys_regs.c +0
-0@@ -169,7 +169,7 @@ void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg)static u32 cache_levels;Quoted 9 lines
Quoted 6 lines
arch/arm64/kvm/sys_regs.c +1
-1/* CSSELR values; used to index KVM_REG_ARM_DEMUX_ID_CCSIDR */ -#define CSSELR_MAX 12 +#define CSSELR_MAX 14/* Which cache CCSIDR represents depends on CSSELR value. */ static u32
get_ccsidr(u32 csselr)Huh, nice catch. Do we need a CC: stable tag for this?
Quoted 6 lines
Hi Marc,
I wasn't thinking so, because I'm not expecting there to actually be hardware with seven cache levels in the wild any time soon. You have more knowledge about what's out there and coming, though, so feel free CC stable if needed.
That's actually what I was wondering, whether you had seen that in the wild already. Since you haven't (and I'm not aware of such a monstrosity), I'll queue it for 5.11.
Do you want me to take the selftest stuff at the same time?
Thanks,
M.
Signature · 5 lines
Jazz is not dead. It just smells funny... _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 1/2] KVM: arm64: CSSELR_EL1 max is 13
Quoted 38 lines · Marc Zyngier
Quoted 30 lines · Andrew Jones
Quoted 19 lines · Marc Zyngier
Quoted 18 lines · Andrew Jones
Not counting TnD, which KVM doesn't currently consider,
CSSELR_EL1can have a maximum value of 0b1101 (13), which corresponds to an instruction cache at level 7. WithCSSELR_MAXset to 12 we can only select up to cache level 6. Change it to 14.Signed-off-by Andrew Jones <[email protected]>--- arch/arm64/kvm/sys_regs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/arm64/kvm/sys_regs.c +0
-0@@ -169,7 +169,7 @@ void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg)static u32 cache_levels;Quoted 9 lines
Quoted 6 lines
arch/arm64/kvm/sys_regs.c +1
-1/* CSSELR values; used to index KVM_REG_ARM_DEMUX_ID_CCSIDR */ -#define CSSELR_MAX 12 +#define CSSELR_MAX 14/* Which cache CCSIDR represents depends on CSSELR value. */ static u32
get_ccsidr(u32 csselr)Huh, nice catch. Do we need a CC: stable tag for this?
Quoted 6 lines
Hi Marc,
I wasn't thinking so, because I'm not expecting there to actually be hardware with seven cache levels in the wild any time soon. You have more knowledge about what's out there and coming, though, so feel free CC stable if needed.
Quoted 6 lines
That's actually what I was wondering, whether you had seen that in the wild already. Since you haven't (and I'm not aware of such a monstrosity), I'll queue it for 5.11.
Do you want me to take the selftest stuff at the same time?
Yes, please :)
Thanks,
drew
PATCH 2/2 KVM: selftests: Filter out DEMUX registers
In reply to Andrew JonesDEMUX register presence depends on the host's hardware (the CLIDR_EL1 register to be precise). This means there's no set of them that we can bless and that it's possible to encounter new ones when running on different hardware (which would generate "Consider adding them ..." messages, but we'll never want to add them.)
Remove the ones we have in the blessed list and filter them out of the new list, but also provide a new command line switch to list them if one so desires.
--- .../selftests/kvm/aarch64/get-reg-list.c | 39 ++++++++++++++----- 1 file changed, 30 insertions(+), 9 deletions(-)
tools/testing/selftests/kvm/aarch64/get-reg-list.c +30 -9
@@ -42,12 +42,16 @@#define for_each_reg (i) \ for ((i) = 0; (i) < reg_list->n; ++(i)) +#define for_each_reg_filtered (i) \ + for_each_reg(i) \ + if (!filter_reg(reg_list->reg[i])) +#define for_each_missing_reg (i) \ for ((i) = 0; (i) < blessed_n; ++(i)) \ if (!find_reg(reg_list->reg, reg_list->n, blessed_reg[i]))#define for_each_new_reg (i) \ - for ((i) = 0; (i) < reg_list->n; ++(i)) \ + for_each_reg_filtered(i) \ if (!find_reg(blessed_reg, blessed_n, reg_list->reg[i]))
@@ -57,6 +61,18 @@ static __u64 base_regs[], vregs[], sve_regs[], rejects_set[];static __u64 base_regs_n, vregs_n, sve_regs_n, rejects_set_n;static __u64 *blessed_reg, blessed_n; +static bool filter_reg (__u64 reg) +{ +/* +* DEMUX register presence depends on the host's CLIDR_EL1. +* This means there's no set of them that we can bless. +*/ +if ((reg &KVM_REG_ARM_COPROC_MASK ) ==KVM_REG_ARM_DEMUX ) +return true; + +return false; +} +static bool find_reg (__u64 regs[],__u64 nr_regs,__u64 reg) {int i;
@@ -325,7 +341,7 @@ int main(int ac, char **av)struct kvm_vcpu_init init = { .target =-1 , };int new_regs =0 , missing_regs =0 , i;int failed_get =0 , failed_set =0 , failed_reject =0 ; -bool print_list = false, fixup_core_regs = false; +bool print_list = false, print_filtered = false, fixup_core_regs = false;struct kvm_vm *vm;__u64 *vec_regs;
@@ -336,8 +352,10 @@ int main(int ac, char **av) fixup_core_regs = true; else if (strcmp(av[i], "--list") == 0) print_list = true; + else if (strcmp(av[i], "--list-filtered") == 0) + print_filtered = true; else - fprintf(stderr, "Ignoring unknown option: %s\n", av[i]); + TEST_FAIL("Unknown option: %s\n", av[i]); } vm = vm_create(VM_MODE_DEFAULT, DEFAULT_GUEST_PHY_PAGES, O_RDWR);
@@ -350,10 +368,14 @@ int main(int ac, char **av)if (fixup_core_regs)core_reg_fixup (); -if (print_list) { +if (print_list || print_filtered) {putchar ('\n' ); -for_each_reg (i) -print_reg (reg_list->reg[i]); +for_each_reg (i) { +__u64 id = reg_list->reg[i]; +if ((print_list && !filter_reg (id)) || + (print_filtered &&filter_reg (id))) +print_reg (id); + }putchar ('\n' );return 0 ; }
@@ -458,6 +480,8 @@ int main(int ac, char **av)/* * The current blessed list was primed with the output of kernel version * v4.15 with --core-reg-fixup and then later updated with new registers. +* +* The blessed list is up to date with kernel version v5.10-rc5 */ static __u64 base_regs[] = {KVM_REG_ARM64 |KVM_REG_SIZE_U64 |KVM_REG_ARM_CORE |KVM_REG_ARM_CORE_REG (regs.regs[0 ]),
@@ -736,9 +760,6 @@ static __u64 base_regs[] = {ARM64_SYS_REG (3 ,4 ,3 ,0 ,0 ),/* DACR32_EL2 */ ARM64_SYS_REG (3 ,4 ,5 ,0 ,1 ),/* IFSR32_EL2 */ ARM64_SYS_REG(3, 4, 5, 3, 0), /* FPEXC32_EL2 */ -KVM_REG_ARM64 |KVM_REG_SIZE_U32 |KVM_REG_ARM_DEMUX |KVM_REG_ARM_DEMUX_ID_CCSIDR |0 , -KVM_REG_ARM64 |KVM_REG_SIZE_U32 |KVM_REG_ARM_DEMUX |KVM_REG_ARM_DEMUX_ID_CCSIDR |1 , - KVM_REG_ARM64 | KVM_REG_SIZE_U32 | KVM_REG_ARM_DEMUX | KVM_REG_ARM_DEMUX_ID_CCSIDR | 2, };static __u64 base_regs_n =ARRAY_SIZE (base_regs);
Signature · 6 lines
2.26.2 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Quoted 8 lines · Andrew Jones
The first patch is a fix, but not one likely to ever truly be needed, as it's unlikely to find seven levels of cache. The bug was found while code reading. Writing the second patch was actually why I was reading the code. The issue being fixed for the get-reg-list test was found when running it on a different machine than what was used to develop it.
[...]
Applied to kvm-arm64/cache-demux, thanks!
[1/2] KVM: arm64: CSSELR_EL1 max is 13
commit: c73a44161776f6e60d933717f3b34084b0a0eba0
[2/2] KVM: arm64: selftests: Filter out DEMUX registers
commit: c6232bd40b2eda3819d108e6e3f621ec604e15d8Cheers,
M.
Signature · 7 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm