On 26/10/18 10:04, Jan Beulich wrote: >>>> On 25.10.18 at 20:36, <[email protected]> wrote: >> On 18/09/18 12:53, Jan Beulich wrote: >>> This is needed before enabling any AVX512 insns in the emulator. Change >>> the way alignment is enforced at the same time. >>> >>> Add a check that the buffer won't actually overflow, and while at it >>> also convert the check for accesses to not cross page boundaries. >>> >>> Signed-off-by: Jan Beulich <[email protected]> >>> --- >>> v3: New. >>> >>> --- a/xen/arch/x86/hvm/emulate.c >>> +++ b/xen/arch/x86/hvm/emulate.c >>> @@ -866,7 +866,18 @@ static int hvmemul_phys_mmio_access( >>> int rc = X86EMUL_OKAY; >>> >>> /* Accesses must fall within a page. */ >>> - BUG_ON((gpa & ~PAGE_MASK) + size > PAGE_SIZE); >>> + if ( (gpa & ~PAGE_MASK) + size > PAGE_SIZE ) >>> + { >>> + ASSERT_UNREACHABLE(); >>> + return X86EMUL_UNHANDLEABLE; >>> + } >>> + >>> + /* Accesses must not overflow the cache's buffer. */ >>> + if ( size > sizeof(cache->buffer) ) >>> + { >>> + ASSERT_UNREACHABLE(); >>> + return X86EMUL_UNHANDLEABLE; >>> + } >>> >>> /* >>> * hvmemul_do_io() cannot handle non-power-of-2 accesses or >>> --- a/xen/include/asm-x86/hvm/vcpu.h >>> +++ b/xen/include/asm-x86/hvm/vcpu.h >>> @@ -42,15 +42,14 @@ struct hvm_vcpu_asid { >>> }; >>> >>> /* >>> - * We may read or write up to m256 as a number of device-model >>> + * We may read or write up to m512 as a number of device-model >>> * transactions. >>> */ >>> struct hvm_mmio_cache { >>> unsigned long gla; >>> unsigned int size; >>> uint8_t dir; >>> - uint8_t pad[3]; /* make buffer[] long-aligned */ >>> - uint8_t buffer[32]; >>> + uint8_t buffer[64] __aligned(sizeof(long)); >> Don't we want it 16-byte aligned, rather than 8? > Why? We don't access the buffer via SIMD insns.
Fair point. Acked-by: Andrew Cooper <[email protected]> _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
