From cd155b56033ab2693464bdade23e2ccd623e9936 Mon Sep 17 00:00:00 2001 From: Don Lewis Date: Wed, 2 Aug 2017 01:43:35 +0000 Subject: [PATCH] Lower the amd64 shared page, which contains the signal trampoline, from the top of user memory to one page lower on machines with the Ryzen (AMD Family 17h) CPU. This pushes ps_strings and the stack down by one page as well. On Ryzen there is some sort of interaction between code running at the top of user memory address space and interrupts that can cause FreeBSD to either hang or silently reset. This sounds similar to the problem found with DragonFly BSD that was fixed with this commit: https://gitweb.dragonflybsd.org/dragonfly.git/commitdiff/b48dd28447fc8ef62fbc963accd301557fd9ac20 but our signal trampoline location was already lower than the address that DragonFly moved their signal trampoline to. It also does not appear to be related to SMT as described here: https://www.phoronix.com/forums/forum/hardware/processors-memory/955368-some-ryzen-linux-users-are-facing-issues-with-heavy-compilation-loads?p=955498#post955498 "Hi, Matt Dillon here. Yes, I did find what I believe to be a hardware issue with Ryzen related to concurrent operations. In a nutshell, for any given hyperthread pair, if one hyperthread is in a cpu-bound loop of any kind (can be in user mode), and the other hyperthread is returning from an interrupt via IRETQ, the hyperthread issuing the IRETQ can stall indefinitely until the other hyperthread with the cpu-bound loop pauses (aka HLT until next interrupt). After this situation occurs, the system appears to destabilize. The situation does not occur if the cpu-bound loop is on a different core than the core doing the IRETQ. The %rip the IRETQ returns to (e.g. userland %rip address) matters a *LOT*. The problem occurs more often with high %rip addresses such as near the top of the user stack, which is where DragonFly's signal trampoline traditionally resides. So a user program taking a signal on one thread while another thread is cpu-bound can cause this behavior. Changing the location of the signal trampoline makes it more difficult to reproduce the problem. I have not been because the able to completely mitigate it. When a cpu-thread stalls in this manner it appears to stall INSIDE the microcode for IRETQ. It doesn't make it to the return pc, and the cpu thread cannot take any IPIs or other hardware interrupts while in this state." since the system instability has been observed on FreeBSD with SMT disabled. Interrupts to appear to play a factor since running a signal-intensive process on the first CPU core, which handles most of the interrupts on my machine, is far more likely to trigger the problem than running such a process on any other core. Also lower sv_maxuser to prevent a malicious user from using mmap() to load and execute code in the top page of user memory that was made available when the shared page was moved down. Make the same changes to the 64-bit Linux emulator. PR: 219399 Reported by: nbe@renzel.net Reviewed by: kib Reviewed by: dchagin (previous version) Tested by: nbe@renzel.net (earlier version) MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D11780 --- sys/amd64/amd64/elf_machdep.c | 19 +++++++++++++++++++ sys/amd64/amd64/initcpu.c | 27 +++++++++++++++++++++++++++ sys/amd64/include/md_var.h | 5 ++++- sys/amd64/linux/linux_sysvec.c | 2 ++ 4 files changed, 52 insertions(+), 1 deletion(-) diff --git a/sys/amd64/amd64/elf_machdep.c b/sys/amd64/amd64/elf_machdep.c index 1c3460eb031c..0a2a85144052 100644 --- a/sys/amd64/amd64/elf_machdep.c +++ b/sys/amd64/amd64/elf_machdep.c @@ -84,6 +84,25 @@ struct sysentvec elf64_freebsd_sysvec = { }; INIT_SYSENTVEC(elf64_sysvec, &elf64_freebsd_sysvec); +void +amd64_lower_shared_page(struct sysentvec *sv) +{ + if (hw_lower_amd64_sharedpage != 0) { + sv->sv_maxuser -= PAGE_SIZE; + sv->sv_shared_page_base -= PAGE_SIZE; + sv->sv_usrstack -= PAGE_SIZE; + sv->sv_psstrings -= PAGE_SIZE; + } +} + +/* + * Do this fixup before INIT_SYSENTVEC (SI_ORDER_ANY) because the latter + * uses the value of sv_shared_page_base. + */ +SYSINIT(elf64_sysvec_fixup, SI_SUB_EXEC, SI_ORDER_FIRST, + (sysinit_cfunc_t) amd64_lower_shared_page, + &elf64_freebsd_sysvec); + static Elf64_Brandinfo freebsd_brand_info = { .brand = ELFOSABI_FREEBSD, .machine = EM_X86_64, diff --git a/sys/amd64/amd64/initcpu.c b/sys/amd64/amd64/initcpu.c index 63c0f20f4928..ff3b6be8602e 100644 --- a/sys/amd64/amd64/initcpu.c +++ b/sys/amd64/amd64/initcpu.c @@ -48,6 +48,11 @@ __FBSDID("$FreeBSD$"); static int hw_instruction_sse; SYSCTL_INT(_hw, OID_AUTO, instruction_sse, CTLFLAG_RD, &hw_instruction_sse, 0, "SIMD/MMX2 instructions available in CPU"); +static int lower_sharedpage_init; +int hw_lower_amd64_sharedpage; +SYSCTL_INT(_hw, OID_AUTO, lower_amd64_sharedpage, CTLFLAG_RDTUN, + &hw_lower_amd64_sharedpage, 0, + "Lower sharedpage to work around Ryzen issue with executing code near the top of user memory"); /* * -1: automatic (default) * 0: keep enable CLFLUSH @@ -122,6 +127,28 @@ init_amd(void) wrmsr(0xc0011020, msr); } } + + /* + * Work around a problem on Ryzen that is triggered by executing + * code near the top of user memory, in our case the signal + * trampoline code in the shared page on amd64. + * + * This function is executed once for the BSP before tunables take + * effect so the value determined here can be overridden by the + * tunable. This function is then executed again for each AP and + * also on resume. Set a flag the first time so that value set by + * the tunable is not overwritten. + * + * The stepping and/or microcode versions should be checked after + * this issue is fixed by AMD so that we don't use this mode if not + * needed. + */ + if (lower_sharedpage_init == 0) { + lower_sharedpage_init = 1; + if (CPUID_TO_FAMILY(cpu_id) == 0x17) { + hw_lower_amd64_sharedpage = 1; + } + } } /* diff --git a/sys/amd64/include/md_var.h b/sys/amd64/include/md_var.h index f161c97d8e41..3854de4a0ba3 100644 --- a/sys/amd64/include/md_var.h +++ b/sys/amd64/include/md_var.h @@ -34,11 +34,14 @@ #include -extern uint64_t *vm_page_dump; +extern uint64_t *vm_page_dump; +extern int hw_lower_amd64_sharedpage; struct savefpu; +struct sysentvec; void amd64_db_resume_dbreg(void); +void amd64_lower_shared_page(struct sysentvec *); void amd64_syscall(struct thread *td, int traced); void doreti_iret(void) __asm(__STRING(doreti_iret)); void doreti_iret_fault(void) __asm(__STRING(doreti_iret_fault)); diff --git a/sys/amd64/linux/linux_sysvec.c b/sys/amd64/linux/linux_sysvec.c index 829324aef440..b87185f1fe33 100644 --- a/sys/amd64/linux/linux_sysvec.c +++ b/sys/amd64/linux/linux_sysvec.c @@ -833,6 +833,8 @@ static void linux_vdso_install(void *param) { + amd64_lower_shared_page(&elf_linux_sysvec); + linux_szsigcode = (&_binary_linux_locore_o_end - &_binary_linux_locore_o_start);