| Message ID | 20240702163515.1964784-3-alex.bennee@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Some fixes for running under -cpu max on QEMU | expand |
Hi Alex, [ Please don't send patches to the old kvmarm@lists.cs.columbia.edu as it had been dropped since early 2023. [1] ] On 2024/7/3 0:35, Alex Bennée wrote: > If FEAT_LPA2 is enabled there are different valid TGran values > possible to indicate the granule is supported for 52 bit addressing. > This will cause most tests to abort on QEMU's -cpu max with the error: > > lib/arm/mmu.c:216: assert failed: system_supports_granule(PAGE_SIZE): Unsupported translation granule 4096 > > Expand the test to tale this into account. > > Signed-off-by: Alex Bennée <alex.bennee@linaro.org> > Cc: Anders Roxell <anders.roxell@linaro.org> There's a similar patch on the list [2], haven't been merged in master though. [1] https://git.kernel.org/torvalds/c/960c3028a1d5 [2] https://lore.kernel.org/all/20240402132739.201939-6-andrew.jones@linux.dev
On Wed, Jul 03, 2024 at 11:52:05AM GMT, Zenghui Yu wrote: > Hi Alex, > > [ Please don't send patches to the old kvmarm@lists.cs.columbia.edu as > it had been dropped since early 2023. [1] ] > > On 2024/7/3 0:35, Alex Bennée wrote: > > If FEAT_LPA2 is enabled there are different valid TGran values > > possible to indicate the granule is supported for 52 bit addressing. > > This will cause most tests to abort on QEMU's -cpu max with the error: > > > > lib/arm/mmu.c:216: assert failed: system_supports_granule(PAGE_SIZE): Unsupported translation granule 4096 > > > > Expand the test to tale this into account. > > > > Signed-off-by: Alex Bennée <alex.bennee@linaro.org> > > Cc: Anders Roxell <anders.roxell@linaro.org> > > There's a similar patch on the list [2], haven't been merged in master > though. Drat, I queued that, and several other patches, and then, for whatever reason, I delayed the merge (I was probably just waiting for the gitlab pipeline to finish...) and then forgot to actually merge... I've merged now. Please don't hesitate to ping me on patches that linger too long. I sometimes need that interrupt to trigger my context switch! Thanks, drew > > [1] https://git.kernel.org/torvalds/c/960c3028a1d5 > [2] > https://lore.kernel.org/all/20240402132739.201939-6-andrew.jones@linux.dev
diff --git a/lib/arm64/asm/processor.h b/lib/arm64/asm/processor.h index 1c73ba32..4a213aec 100644 --- a/lib/arm64/asm/processor.h +++ b/lib/arm64/asm/processor.h @@ -110,31 +110,30 @@ static inline unsigned long get_id_aa64mmfr0_el1(void) #define ID_AA64MMFR0_TGRAN64_SHIFT 24 #define ID_AA64MMFR0_TGRAN16_SHIFT 20 -#define ID_AA64MMFR0_TGRAN4_SUPPORTED 0x0 -#define ID_AA64MMFR0_TGRAN64_SUPPORTED 0x0 -#define ID_AA64MMFR0_TGRAN16_SUPPORTED 0x1 +#define ID_AA64MMFR0_TGRAN4_OK 0x0 +#define ID_AA64MMFR0_TGRAN4_52_OK 0x1 +#define ID_AA64MMFR0_TGRAN64_OK 0x0 +#define ID_AA64MMFR0_TGRAN16_OK 0x1 +#define ID_AA64MMFR0_TGRAN16_52_OK 0x2 static inline bool system_supports_granule(size_t granule) { - u32 shift; u32 val; - u64 mmfr0; + u64 mmfr0 = get_id_aa64mmfr0_el1(); if (granule == SZ_4K) { - shift = ID_AA64MMFR0_TGRAN4_SHIFT; - val = ID_AA64MMFR0_TGRAN4_SUPPORTED; + val = ((mmfr0 >> ID_AA64MMFR0_TGRAN4_SHIFT) & 0xf); + return (val == ID_AA64MMFR0_TGRAN4_OK) || + (val == ID_AA64MMFR0_TGRAN4_52_OK); } else if (granule == SZ_16K) { - shift = ID_AA64MMFR0_TGRAN16_SHIFT; - val = ID_AA64MMFR0_TGRAN16_SUPPORTED; + val = ((mmfr0 >> ID_AA64MMFR0_TGRAN16_SHIFT) & 0xf); + return val == ID_AA64MMFR0_TGRAN16_OK; } else { assert(granule == SZ_64K); - shift = ID_AA64MMFR0_TGRAN64_SHIFT; - val = ID_AA64MMFR0_TGRAN64_SUPPORTED; + val = ((mmfr0 >> ID_AA64MMFR0_TGRAN64_SHIFT) & 0xf); + return (val == ID_AA64MMFR0_TGRAN64_OK) || + (val == ID_AA64MMFR0_TGRAN4_52_OK); } - - mmfr0 = get_id_aa64mmfr0_el1(); - - return ((mmfr0 >> shift) & 0xf) == val; } #endif /* !__ASSEMBLY__ */
If FEAT_LPA2 is enabled there are different valid TGran values possible to indicate the granule is supported for 52 bit addressing. This will cause most tests to abort on QEMU's -cpu max with the error: lib/arm/mmu.c:216: assert failed: system_supports_granule(PAGE_SIZE): Unsupported translation granule 4096 Expand the test to tale this into account. Signed-off-by: Alex Bennée <alex.bennee@linaro.org> Cc: Anders Roxell <anders.roxell@linaro.org> --- lib/arm64/asm/processor.h | 29 ++++++++++++++--------------- 1 file changed, 14 insertions(+), 15 deletions(-)