x86/insn-eval: Add support for 64-bit kernel mode
Subsystems (6)
PATCH v5 1/4 x86/insn-eval: Add support for 64-bit kernel mode
To support evaluating 64-bit kernel mode instructions:
Replace existing checks for user_64bit_mode() with a new helper that checks whether code is being executed in either 64-bit kernel mode or 64-bit user mode.
Select the GS base depending on whether the instruction is being evaluated in kernel mode.
---
Notes:
v2-v5:
no changesarch/x86/include/asm/ptrace.h | 13 +++++++++++++ arch/x86/lib/insn-eval.c | 26 +++++++++++++++----------- 2 files changed, 28 insertions(+), 11 deletions(-)
arch/x86/include/asm/ptrace.h +13 -0
@@ -159,6 +159,19 @@ static inline bool user_64bit_mode(struct pt_regs *regs)#endif } +/* +* Determine whether the register set came from any context that is running in +* 64-bit mode. +*/ +static inline bool any_64bit_mode (struct pt_regs *regs) +{ +#ifdef CONFIG_X86_64 +return !user_mode (regs) ||user_64bit_mode (regs); +#else +return false; +#endif +} +#ifdef CONFIG_X86_64 #define current_user_stack_pointer () current_pt_regs()->sp#define compat_user_stack_pointer () current_pt_regs()->sp
arch/x86/lib/insn-eval.c +15 -11
@@ -155,7 +155,7 @@ static bool check_seg_overrides(struct insn *insn, int regoff)*/ static int resolve_default_seg (struct insn *insn,struct pt_regs *regs,int off) { -if (user_64bit_mode (regs)) +if (any_64bit_mode (regs))return INAT_SEG_REG_IGNORE ;/* * Resolve the default segment register as described in Section 3.7.4
@@ -266,7 +266,7 @@ static int resolve_seg_reg(struct insn *insn, struct pt_regs *regs, int regoff)* which may be invalid at this point. */ if (regoff == offsetof(struct pt_regs , ip)) { -if (user_64bit_mode (regs)) +if (any_64bit_mode (regs))return INAT_SEG_REG_IGNORE ;else return INAT_SEG_REG_CS ;
@@ -289,7 +289,7 @@ static int resolve_seg_reg(struct insn *insn, struct pt_regs *regs, int regoff)* In long mode, segment override prefixes are ignored, except for * overrides for FS and GS. */ -if (user_64bit_mode (regs)) { +if (any_64bit_mode (regs)) {if (idx !=INAT_SEG_REG_FS && idx !=INAT_SEG_REG_GS ) idx =INAT_SEG_REG_IGNORE ;
@@ -646,23 +646,27 @@ unsigned long insn_get_seg_base(struct pt_regs *regs, int seg_reg_idx)*/ return (unsigned long )(sel <<4 ); -if (user_64bit_mode (regs)) { +if (any_64bit_mode (regs)) {/* * Only FS or GS will have a base address, the rest of * the segments' bases are forced to 0. */ unsigned long base; -if (seg_reg_idx ==INAT_SEG_REG_FS ) +if (seg_reg_idx ==INAT_SEG_REG_FS ) {rdmsrl (MSR_FS_BASE , base); -else if (seg_reg_idx ==INAT_SEG_REG_GS ) + }else if (seg_reg_idx ==INAT_SEG_REG_GS ) {/* * swapgs was called at the kernel entry point. Thus, * MSR_KERNEL_GS_BASE will have the user-space GS base. */ -rdmsrl (MSR_KERNEL_GS_BASE , base); -else +if (user_mode (regs)) +rdmsrl (MSR_KERNEL_GS_BASE , base); +else +rdmsrl (MSR_GS_BASE , base); + }else { base =0 ; + }return base; }
@@ -703,7 +707,7 @@ static unsigned long get_seg_limit(struct pt_regs *regs, int seg_reg_idx)if (sel <0 )return 0 ; -if (user_64bit_mode (regs) ||v8086_mode (regs)) +if (any_64bit_mode (regs) ||v8086_mode (regs))return -1L ;if (!sel)
@@ -948,7 +952,7 @@ static int get_eff_addr_modrm(struct insn *insn, struct pt_regs *regs,* following instruction. */ if (*regoff == -EDOM ) { -if (user_64bit_mode (regs)) +if (any_64bit_mode (regs)) tmp = regs->ip + insn->length;else tmp =0 ;
@@ -1250,7 +1254,7 @@ static void __user *get_addr_ref_32(struct insn *insn, struct pt_regs *regs)* After computed, the effective address is treated as an unsigned * quantity. */ -if (!user_64bit_mode (regs) && ((unsigned int )eff_addr > seg_limit)) +if (!any_64bit_mode (regs) && ((unsigned int )eff_addr > seg_limit)) goto out;/*
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
PATCH v5 2/4 x86/traps: Print address on #GP
A frequent cause of #GP exceptions are memory accesses to non-canonical addresses. Unlike #PF, #GP doesn't come with a fault address in CR2, so the kernel doesn't currently print the fault address for #GP. Luckily, we already have the necessary infrastructure for decoding X86 instructions and computing the memory address that is being accessed; hook it up to the #GP handler so that we can figure out the address operand of the faulting instruction and print it.
Distinguish two cases:
a) (Part of) the memory range being accessed lies in the non-canonical address range; in this case, is is likely that the address we decoded is actually the one that caused the #GP.
b) The entire memory range of the operand we decoded lies in canonical address space; the #GP may or may not be related in some way to the address we computed. We'll still print it, but with hedging language in the message.
While it is already possible to compute the faulting address manually by disassembling the opcode dump and evaluating the instruction against the register dump, this should make it slightly easier to identify crashes at a glance.
Note that the operand length, which we get from the instruction decoder and use to determine whether the access straddles into non-canonical address space, is currently somewhat unreliable; but it should be good enough, considering that Linux on x86-64 never maps the page directly before the start of the non-canonical range anyway, and therefore the case where a memory range begins in that page and potentially straddles into the non-canonical range should be fairly uncommon.
And if we do get this wrong, it only influences whether the error message claims that the access is canonical.
---
Notes:
v2:
- print different message for segment-related GP (Borislav)
- rewrite check for non-canonical address (Sean)
- make it clear we don't know for sure why the GP happened (Andy)
v3:
- change message format to one line (Borislav)
v4:
- rename insn_bytes to insn_buf (Ingo)
- add space after GPFSTR (Ingo)
- make sizeof(desc) clearer (Ingo, Borislav)
- also print the address (with a different message) if it's canonical (Ingo)
v5:
- reword comment on get_kernel_gp_address() (Sean)
- make get_kernel_gp_address() also work on 32-bit (Sean)
- minor nits (Sean)
- more hedging for canonical GP (Sean)
- let get_kernel_gp_address() return an enum (Sean)
- rewrite commit message
I have already sent a patch to syzkaller that relaxes their parsing of GPF messages (https://github.com/google/syzkaller/commit/432c7650) such that changes like the one in this patch don't break it.
That patch has already made its way into syzbot's syzkaller instances according to <https://syzkaller.appspot.com/upstream>.
arch/x86/kernel/traps.c | 70 +++++++++++++++++++++++++++++++++++++++-- 1 file changed, 67 insertions(+), 3 deletions(-)
arch/x86/kernel/traps.c +67 -3
@@ -56,6 +56,8 @@#include <asm/mpx.h> #include <asm/vm86.h> #include <asm/umip.h> +#include <asm/insn.h> +#include <asm/insn-eval.h> #ifdef CONFIG_X86_64 #include <asm/x86_init.h>
@@ -518,11 +520,56 @@ dotraplinkage void do_bounds(struct pt_regs *regs, long error_code)do_trap (X86_TRAP_BR ,SIGSEGV ,"bounds" , regs, error_code,0 ,NULL ); } +enum kernel_gp_hint { +GP_NO_HINT , +GP_NON_CANONICAL , +GP_CANONICAL +}; + +/* +* When an uncaught #GP occurs, try to determine a memory address accessed by +* the instruction and return that address to the caller. +* Also try to figure out whether any part of the access to that address was +* non-canonical. +*/ +static enum kernel_gp_hint get_kernel_gp_address (struct pt_regs *regs, +unsigned long *addr) +{ +u8 insn_buf[MAX_INSN_SIZE ]; +struct insn insn; + +if (probe_kernel_read (insn_buf, (void *)regs->ip,MAX_INSN_SIZE )) +return GP_NO_HINT ; + +kernel_insn_init (&insn, insn_buf,MAX_INSN_SIZE ); +insn_get_modrm (&insn); +insn_get_sib (&insn); + *addr = (unsigned long )insn_get_addr_ref (&insn, regs); + +if (*addr ==-1UL ) +return GP_NO_HINT ; + +#ifdef CONFIG_X86_64 +/* +* Check that: +* - the operand is not in the kernel half +* - the last byte of the operand is not in the user canonical half +*/ +if (*addr < ~__VIRTUAL_MASK && + *addr + insn.opnd_bytes -1 > __VIRTUAL_MASK) +return GP_NON_CANONICAL ; +#endif + +return GP_CANONICAL ; +} + +#define GPFSTR "general protection fault" + dotraplinkage voiddo_general_protection (struct pt_regs *regs,long error_code) { -const char *desc ="general protection fault" ;struct task_struct *tsk; +char desc[sizeof (GPFSTR ) +50 +2 *sizeof (unsigned long ) +1 ] =GPFSTR ;RCU_LOCKDEP_WARN (!rcu_is_watching (),"entry code didn't wake RCU" );cond_local_irq_enable (regs);
@@ -540,6 +587,9 @@ do_general_protection(struct pt_regs *regs, long error_code) tsk = current;if (!user_mode (regs)) { +enum kernel_gp_hint hint =GP_NO_HINT ; +unsigned long gp_addr; +if (fixup_exception (regs,X86_TRAP_GP , error_code,0 ))return ;
@@ -556,8 +606,22 @@ do_general_protection(struct pt_regs *regs, long error_code)return ;if (notify_die (DIE_GPF , desc, regs, error_code, -X86_TRAP_GP ,SIGSEGV ) !=NOTIFY_STOP ) -die (desc, regs, error_code); +X86_TRAP_GP ,SIGSEGV ) ==NOTIFY_STOP ) +return ; + +if (error_code) +snprintf (desc,sizeof (desc),"segment-related " GPFSTR ); +else + hint =get_kernel_gp_address (regs, &gp_addr); + +if (hint !=GP_NO_HINT ) +snprintf (desc,sizeof (desc),GPFSTR " %s 0x%lx" , + (hint ==GP_NON_CANONICAL ) ? +"probably for non-canonical address" : +"maybe for address" , + gp_addr); + +die (desc, regs, error_code);return ; }
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
Re: [PATCH v5 2/4] x86/traps: Print address on #GP
On Thu, Nov 28, 2019 at 12:49:14AM +0100, Jann Horn wrote:
With a few nits below,
arch/x86/kernel/traps.c dotraplinkage void do_bounds(struct pt_regs *regs, long error_code) · line 572
Quoted 8 lines
arch/x86/kernel/traps.c +3
-1+#define GPFSTR "general protection fault" + dotraplinkage voiddo_general_protection (struct pt_regs *regs,long error_code) { -const char *desc ="general protection fault" ;struct task_struct *tsk; +char desc[sizeof (GPFSTR ) +50 +2 *sizeof (unsigned long ) +1 ] =GPFSTR ;
Nit, x86 maintainers prefer inverse fir tree for variable declarations.
Quoted 19 lines
arch/x86/kernel/traps.c +3
-2RCU_LOCKDEP_WARN (!rcu_is_watching (),"entry code didn't wake RCU" );cond_local_irq_enable (regs);@@ -540,6 +587,9 @@ do_general_protection(struct pt_regs *regs, long error_code) tsk = current;if (!user_mode (regs)) { +enum kernel_gp_hint hint =GP_NO_HINT ; +unsigned long gp_addr; +if (fixup_exception (regs,X86_TRAP_GP , error_code,0 ))return ;@@ -556,8 +606,22 @@ do_general_protection(struct pt_regs *regs, long error_code)return ; if (notify_die(DIE_GPF, desc, regs, error_code, -X86_TRAP_GP ,SIGSEGV ) !=NOTIFY_STOP ) -die (desc, regs, error_code);
arch/x86/kernel/traps.c do_general_protection(struct pt_regs *regs, long error_code) · line 618
Quoted 10 lines
arch/x86/kernel/traps.c +10
-0+ X86_TRAP_GP, SIGSEGV) == NOTIFY_STOP) +return ; + +if (error_code) +snprintf (desc,sizeof (desc),"segment-related " GPFSTR ); +else + hint =get_kernel_gp_address (regs, &gp_addr); + +if (hint !=GP_NO_HINT ) + snprintf(desc, sizeof(desc), GPFSTR " %s 0x%lx",
Nit, probably should have a comma before the hint, i.e. GPFSTR ", %s...".
general protection fault maybe for address 0xffffc9000017cf58: 0000 [#1] SMP
general protection fault probably for non-canonical address 0xdead000000000000: 0000 [#1] SMPvs.
general protection fault, maybe for address 0xffffc9000017cf58: 0000 [#1] SMP
general protection fault, probably for non-canonical address 0xdead000000000000: 0000 [#1] SMPQuoted 12 lines
arch/x86/kernel/traps.c +6
-0+ (hint ==GP_NON_CANONICAL ) ? +"probably for non-canonical address" : +"maybe for address" , + gp_addr); + +die (desc, regs, error_code);return ; }Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
PATCH v5 3/4 x86/dumpstack: Split out header line printing from __die()
In reply to Jann HornSplit __die() into __die_header() and __die_body(). This allows callers to insert extra information below the header line that initiates the bug report.
This can e.g. be used by __die() callers to allow KASAN to print additional information below the header line of the bug report.
---
Notes:
I think that it's nicer to have KASAN's notes about the bug below the first oops line from the kernel. This also means that tools that work with kernel oops reports can just trigger on the "general protection fault" line with the die counter and so on, and just include the text from on there, and the KASAN message will automatically be included.
But if you think that the code looks too ugly, I'd be happy to change that back and drop this patch from the series.
v3:
new patch
v4-v5:
no changes
arch/x86/include/asm/kdebug.h | 3 +++ arch/x86/kernel/dumpstack.c | 13 ++++++++++++- 2 files changed, 15 insertions(+), 1 deletion(-)
arch/x86/include/asm/kdebug.h +3 -0
@@ -33,6 +33,9 @@ enum show_regs_mode { };extern void die (const char *,struct pt_regs *,long ); +extern void __die_header (const char *str,struct pt_regs *regs,long err); +extern int __must_check __die_body (const char *str,struct pt_regs *regs, +long err);extern int __must_check __die (const char *,struct pt_regs *,long );extern void show_stack_regs (struct pt_regs *regs);extern void __show_regs (struct pt_regs *regs,enum show_regs_mode );
arch/x86/kernel/dumpstack.c +12 -1
@@ -365,7 +365,7 @@ void oops_end(unsigned long flags, struct pt_regs *regs, int signr) }NOKPROBE_SYMBOL (oops_end); -int __die (const char *str,struct pt_regs *regs,long err) +void __die_header (const char *str,struct pt_regs *regs,long err) {const char *pr ="" ;
@@ -384,7 +384,11 @@ int __die(const char *str, struct pt_regs *regs, long err)IS_ENABLED (CONFIG_KASAN ) ?" KASAN" :"" ,IS_ENABLED (CONFIG_PAGE_TABLE_ISOLATION ) ? (boot_cpu_has(X86_FEATURE_PTI) ? " PTI" : " NOPTI") : ""); +} +NOKPROBE_SYMBOL (__die_header); +int __die_body (const char *str,struct pt_regs *regs,long err) +{show_regs (regs);print_modules ();
@@ -394,6 +398,13 @@ int __die(const char *str, struct pt_regs *regs, long err)return 0 ; } +NOKPROBE_SYMBOL (__die_body); + +int __die (const char *str,struct pt_regs *regs,long err) +{ +__die_header (str, regs, err); +return __die_body (str, regs, err); +}NOKPROBE_SYMBOL (__die);/*
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
PATCH v5 4/4 x86/kasan: Print original address on #GP
In reply to Jann HornMake #GP exceptions caused by out-of-bounds KASAN shadow accesses easier to understand by computing the address of the original access and printing that. More details are in the comments in the patch.
This turns an error like this:
kasan: CONFIG_KASAN_INLINE enabled
kasan: GPF could be caused by NULL-ptr deref or user memory access general protection fault probably for non-canonical address
0xe017577ddf75b7dd: 0000 [#1] PREEMPT SMP KASAN PTI
into this:
general protection fault probably for non-canonical address
0xe017577ddf75b7dd: 0000 [#1] PREEMPT SMP KASAN PTI
KASAN: maybe wild-memory-access in range
[0x00badbeefbadbee8-0x00badbeefbadbeef]The hook is placed in architecture-independent code, but is currently only wired up to the X86 exception handler because I'm not sufficiently familiar with the address space layout and exception handling mechanisms on other architectures.
---
Notes:
v2:
- move to mm/kasan/report.c (Dmitry)
- change hook name to be more generic
- use TASK_SIZE instead of TASK_SIZE_MAX for compiling on non-x86
- don't open-code KASAN_SHADOW_MASK (Dmitry)
- add "KASAN: " prefix, but not "BUG: " (Andrey, Dmitry)
- use same naming scheme as get_wild_bug_type (Andrey)
- this version was "Reviewed-by: Dmitry Vyukov <[email protected]>"
v3:
- adjusted example output in commit message based on changes in preceding patch
- ensure that KASAN output happens after bust_spinlocks(1)
- moved hook in arch/x86/kernel/traps.c such that output appears after the first line of KASAN-independent error report
v4:
- adjust patch to changes in x86/traps patch
v5:
- adjust patch to changes in x86/traps patch
- fix bug introduced in v3: remove die() call after oops_end()
arch/x86/kernel/traps.c | 12 ++++++++++- arch/x86/mm/kasan_init_64.c | 21 ------------------- include/linux/kasan.h | 6 ++++++ mm/kasan/report.c | 40 +++++++++++++++++++++++++++++++++++++ 4 files changed, 57 insertions(+), 22 deletions(-)
arch/x86/kernel/traps.c +11 -1
@@ -37,6 +37,7 @@#include <linux/mm.h> #include <linux/smp.h> #include <linux/io.h> +#include <linux/kasan.h> #include <asm/stacktrace.h> #include <asm/processor.h> #include <asm/debugreg.h>
@@ -589,6 +590,8 @@ do_general_protection(struct pt_regs *regs, long error_code)if (!user_mode (regs)) {enum kernel_gp_hint hint =GP_NO_HINT ;unsigned long gp_addr; +unsigned long flags; +int sig;if (fixup_exception (regs,X86_TRAP_GP , error_code,0 ))return ;
@@ -621,7 +624,14 @@ do_general_protection(struct pt_regs *regs, long error_code)"maybe for address" , gp_addr); -die (desc, regs, error_code); + flags =oops_begin (); + sig =SIGSEGV ; +__die_header (desc, regs, error_code); +if (hint ==GP_NON_CANONICAL ) +kasan_non_canonical_hook (gp_addr); +if (__die_body (desc, regs, error_code)) + sig =0 ; +oops_end (flags, regs, sig);return ; }
arch/x86/mm/kasan_init_64.c +0 -21
@@ -245,23 +245,6 @@ static void __init kasan_map_early_shadow(pgd_t *pgd) }while (pgd++, addr = next, addr != end); } -#ifdef CONFIG_KASAN_INLINE -static int kasan_die_handler (struct notifier_block *self, -unsigned long val, -void *data) -{ -if (val ==DIE_GPF ) { -pr_emerg ("CONFIG_KASAN_INLINE enabled\n" ); -pr_emerg ("GPF could be caused by NULL-ptr deref or user memory access\n" ); - } -return NOTIFY_OK ; -} - -static struct notifier_block kasan_die_notifier = { - .notifier_call = kasan_die_handler, -}; -#endif -void __init kasan_early_init (void ) {int i;
@@ -298,10 +281,6 @@ void __init kasan_init(void)int i;void *shadow_cpu_entry_begin, *shadow_cpu_entry_end; -#ifdef CONFIG_KASAN_INLINE -register_die_notifier (&kasan_die_notifier); -#endif -memcpy (early_top_pgt, init_top_pgt,sizeof (early_top_pgt));/*
include/linux/kasan.h +6 -0
@@ -194,4 +194,10 @@ static inline void *kasan_reset_tag(const void *addr)#endif /* CONFIG_KASAN_SW_TAGS */ +#ifdef CONFIG_KASAN_INLINE +void kasan_non_canonical_hook (unsigned long addr); +#else /* CONFIG_KASAN_INLINE */ +static inline void kasan_non_canonical_hook (unsigned long addr) { } +#endif /* CONFIG_KASAN_INLINE */ +#endif /* LINUX_KASAN_H */
mm/kasan/report.c +40 -0
@@ -512,3 +512,43 @@ void __kasan_report(unsigned long addr, size_t size, bool is_write, unsigned lonend_report (&flags); } + +#ifdef CONFIG_KASAN_INLINE +/* +* With CONFIG_KASAN_INLINE, accesses to bogus pointers (outside the high +* canonical half of the address space) cause out-of-bounds shadow memory reads +* before the actual access. For addresses in the low canonical half of the +* address space, as well as most non-canonical addresses, that out-of-bounds +* shadow memory access lands in the non-canonical part of the address space. +* Help the user figure out what the original bogus pointer was. +*/ +void kasan_non_canonical_hook (unsigned long addr) +{ +unsigned long orig_addr; +const char *bug_type; + +if (addr <KASAN_SHADOW_OFFSET ) +return ; + + orig_addr = (addr -KASAN_SHADOW_OFFSET ) <<KASAN_SHADOW_SCALE_SHIFT ; +/* +* For faults near the shadow address for NULL, we can be fairly certain +* that this is a KASAN shadow memory access. +* For faults that correspond to shadow for low canonical addresses, we +* can still be pretty sure - that shadow region is a fairly narrow +* chunk of the non-canonical address space. +* But faults that look like shadow for non-canonical addresses are a +* really large chunk of the address space. In that case, we still +* print the decoded address, but make it clear that this is not +* necessarily what's actually going on. +*/ +if (orig_addr <PAGE_SIZE ) + bug_type ="null-ptr-deref" ; +else if (orig_addr <TASK_SIZE ) + bug_type ="probably user-memory-access" ; +else + bug_type ="maybe wild-memory-access" ; +pr_alert ("KASAN: %s in range [0x%016lx-0x%016lx]\n" , bug_type, + orig_addr, orig_addr +KASAN_SHADOW_MASK ); +} +#endif
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog