Re: [PATCH v2 2/3] x86/traps: Print non-canonical address on #GP

From: Borislav Petkov
Date: Mon Nov 18 2019 - 09:21:51 EST


On Fri, Nov 15, 2019 at 08:17:27PM +0100, Jann Horn wrote:
> dotraplinkage void
> do_general_protection(struct pt_regs *regs, long error_code)
> {
> @@ -547,8 +581,15 @@ 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)
> + pr_alert("GPF is segment-related (see error code)\n");
> + else
> + print_kernel_gp_address(regs);
> +
> + die(desc, regs, error_code);

Right, this way, those messages appear before the main "general
protection ..." message:

[ 2.434372] traps: probably dereferencing non-canonical address 0xdfff000000000001
[ 2.442492] general protection fault: 0000 [#1] PREEMPT SMP

Can we glue/merge them together? Or is this going to confuse tools too much:

[ 2.542218] general protection fault while derefing a non-canonical address 0xdfff000000000001: 0000 [#1] PREEMPT SMP

(and that sentence could be shorter too:

"general protection fault for non-canonical address 0xdfff000000000001"

looks ok to me too.)

Here's a dirty diff together with a reproducer ontop of yours:

---
diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
index bf796f8c9998..dab702ba28a6 100644
--- a/arch/x86/kernel/traps.c
+++ b/arch/x86/kernel/traps.c
@@ -515,7 +515,7 @@ dotraplinkage void do_bounds(struct pt_regs *regs, long error_code)
* On 64-bit, if an uncaught #GP occurs while dereferencing a non-canonical
* address, print that address.
*/
-static void print_kernel_gp_address(struct pt_regs *regs)
+static unsigned long get_kernel_gp_address(struct pt_regs *regs)
{
#ifdef CONFIG_X86_64
u8 insn_bytes[MAX_INSN_SIZE];
@@ -523,7 +523,7 @@ static void print_kernel_gp_address(struct pt_regs *regs)
unsigned long addr_ref;

if (probe_kernel_read(insn_bytes, (void *)regs->ip, MAX_INSN_SIZE))
- return;
+ return 0;

kernel_insn_init(&insn, insn_bytes, MAX_INSN_SIZE);
insn_get_modrm(&insn);
@@ -532,22 +532,22 @@ static void print_kernel_gp_address(struct pt_regs *regs)

/* Bail out if insn_get_addr_ref() failed or we got a kernel address. */
if (addr_ref >= ~__VIRTUAL_MASK)
- return;
+ return 0;

/* Bail out if the entire operand is in the canonical user half. */
if (addr_ref + insn.opnd_bytes - 1 <= __VIRTUAL_MASK)
- return;
+ return 0;

- pr_alert("probably dereferencing non-canonical address 0x%016lx\n",
- addr_ref);
+ return addr_ref;
#endif
}

+#define GPFSTR "general protection fault"
dotraplinkage void
do_general_protection(struct pt_regs *regs, long error_code)
{
- const char *desc = "general protection fault";
struct task_struct *tsk;
+ char desc[90];

RCU_LOCKDEP_WARN(!rcu_is_watching(), "entry code didn't wake RCU");
cond_local_irq_enable(regs);
@@ -584,12 +584,18 @@ do_general_protection(struct pt_regs *regs, long error_code)
X86_TRAP_GP, SIGSEGV) == NOTIFY_STOP)
return;

- if (error_code)
- pr_alert("GPF is segment-related (see error code)\n");
- else
- print_kernel_gp_address(regs);
+ if (error_code) {
+ snprintf(desc, 90, "segment-related " GPFSTR);
+ } else {
+ unsigned long addr_ref = get_kernel_gp_address(regs);
+
+ if (addr_ref)
+ snprintf(desc, 90, GPFSTR " while derefing a non-canonical address 0x%lx", addr_ref);
+ else
+ snprintf(desc, 90, GPFSTR);
+ }

- die(desc, regs, error_code);
+ die((const char *)desc, regs, error_code);
return;
}

diff --git a/init/main.c b/init/main.c
index 91f6ebb30ef0..7acc7e660be9 100644
--- a/init/main.c
+++ b/init/main.c
@@ -1124,6 +1124,9 @@ static int __ref kernel_init(void *unused)

rcu_end_inkernel_boot();

+ asm volatile("mov $0xdfff000000000001, %rax\n\t"
+ "jmpq *%rax\n\t");
+
if (ramdisk_execute_command) {
ret = run_init_process(ramdisk_execute_command);
if (!ret)

--
Regards/Gruss,
Boris.

https://people.kernel.org/tglx/notes-about-netiquette