Skip to content

Commit

Permalink
vsprintf: Do not break early boot with probing addresses
Browse files Browse the repository at this point in the history
The commit 3e5903e ("vsprintf: Prevent crash when dereferencing
invalid pointers") broke boot on several architectures. The common
pattern is that probe_kernel_read() is not working during early
boot because userspace access framework is not ready.

It is a generic problem. We have to avoid any complex external
functions in vsprintf() code, especially in the common path.
They might break printk() easily and are hard to debug.

Replace probe_kernel_read() with some simple checks for obvious
problems.

Details:

1. Report on Power:

Kernel crashes very early during boot with with CONFIG_PPC_KUAP and
CONFIG_JUMP_LABEL_FEATURE_CHECK_DEBUG

The problem is the combination of some new code called via printk(),
check_pointer() which calls probe_kernel_read(). That then calls
allow_user_access() (PPC_KUAP) and that uses mmu_has_feature() too early
(before we've patched features). With the JUMP_LABEL debug enabled that
causes us to call printk() & dump_stack() and we end up recursing and
overflowing the stack.

Because it happens so early you don't get any output, just an apparently
dead system.

The stack trace (which you don't see) is something like:

  ...
  dump_stack+0xdc
  probe_kernel_read+0x1a4
  check_pointer+0x58
  string+0x3c
  vsnprintf+0x1bc
  vscnprintf+0x20
  printk_safe_log_store+0x7c
  printk+0x40
  dump_stack_print_info+0xbc
  dump_stack+0x8
  probe_kernel_read+0x1a4
  probe_kernel_read+0x19c
  check_pointer+0x58
  string+0x3c
  vsnprintf+0x1bc
  vscnprintf+0x20
  vprintk_store+0x6c
  vprintk_emit+0xec
  vprintk_func+0xd4
  printk+0x40
  cpufeatures_process_feature+0xc8
  scan_cpufeatures_subnodes+0x380
  of_scan_flat_dt_subnodes+0xb4
  dt_cpu_ftrs_scan_callback+0x158
  of_scan_flat_dt+0xf0
  dt_cpu_ftrs_scan+0x3c
  early_init_devtree+0x360
  early_setup+0x9c

2. Report on s390:

vsnprintf invocations, are broken on s390. For example, the early boot
output now looks like this where the first (efault) should be
the linux_banner:

[    0.099985] (efault)
[    0.099985] setup: Linux is running as a z/VM guest operating system in 64-bit mode
[    0.100066] setup: The maximum memory size is 8192MB
[    0.100070] cma: Reserved 4 MiB at (efault)
[    0.100100] numa: NUMA mode: (efault)

The reason for this, is that the code assumes that
probe_kernel_address() works very early. This however is not true on
at least s390. Uaccess on KERNEL_DS works only after page tables have
been setup on s390, which happens with setup_arch()->paging_init().

Any probe_kernel_address() invocation before that will return -EFAULT.

Fixes: 3e5903e ("vsprintf: Prevent crash when dereferencing invalid pointers")
Link: http://lkml.kernel.org/r/[email protected]
Cc: Andy Shevchenko <[email protected]>
Cc: Rasmus Villemoes <[email protected]>
Cc: "Tobin C . Harding" <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Sergey Senozhatsky <[email protected]>
Cc: Steven Rostedt <[email protected]>
Cc: [email protected]
Cc: Michael Ellerman <[email protected]>
Cc: [email protected]
Cc: Russell Currey <[email protected]>
Cc: Christophe Leroy <[email protected]>
Cc: Stephen Rothwell <[email protected]>
Cc: Heiko Carstens <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: Martin Schwidefsky <[email protected]>
Cc: Petr Mladek <[email protected]>
Reviewed-by: Sergey Senozhatsky <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
  • Loading branch information
pmladek committed May 10, 2019
1 parent 0968621 commit 2ac5a3b
Showing 1 changed file with 4 additions and 7 deletions.
11 changes: 4 additions & 7 deletions lib/vsprintf.c
Original file line number Diff line number Diff line change
Expand Up @@ -628,19 +628,16 @@ static char *error_string(char *buf, char *end, const char *s,
}

/*
* This is not a fool-proof test. 99% of the time that this will fault is
* due to a bad pointer, not one that crosses into bad memory. Just test
* the address to make sure it doesn't fault due to a poorly added printk
* during debugging.
* Do not call any complex external code here. Nested printk()/vsprintf()
* might cause infinite loops. Failures might break printk() and would
* be hard to debug.
*/
static const char *check_pointer_msg(const void *ptr)
{
char byte;

if (!ptr)
return "(null)";

if (probe_kernel_address(ptr, byte))
if ((unsigned long)ptr < PAGE_SIZE || IS_ERR_VALUE(ptr))
return "(efault)";

return NULL;
Expand Down

0 comments on commit 2ac5a3b

Please sign in to comment.