Re: [PATCH] bpf: whitelist syscalls for error injection

[Howard McLauchlan <hmclauchlan@fb.com>]

On 03/13/2018 04:49 PM, Yonghong Song wrote:
> 
> 
> On 3/13/18 4:45 PM, Omar Sandoval wrote:
>> On Tue, Mar 13, 2018 at 04:16:27PM -0700, Howard McLauchlan wrote:
>>> Error injection is a useful mechanism to fail arbitrary kernel
>>> functions. However, it is often hard to guarantee an error propagates
>>> appropriately to user space programs. By injecting into syscalls, we can
>>> return arbitrary values to user space directly; this increases
>>> flexibility and robustness in testing, allowing us to test user space
>>> error paths effectively.
>>>
>>> The following script, for example, fails calls to sys_open() from a
>>> given pid:
>>>
>>> from bcc import BPF
>>> from sys import argv
>>>
>>> pid = argv[1]
>>>
>>> prog = r"""
>>>
>>> int kprobe__SyS_open(struct pt_regs *ctx, const char *pathname, int flags)
>>> {
>>>      u32 pid = bpf_get_current_pid_tgid();
>>>      if (pid == %s)
>>>          bpf_override_return(ctx, -ENOENT);
>>>      return 0;
>>> }
>>> """ % pid
>>>
>>> b = BPF(text = prog)
>>> while 1:
>>>      b.perf_buffer_poll()
>>>
>>> This patch whitelists all syscalls defined with SYSCALL_DEFINE for error
>>> injection.
>>>
>>> Signed-off-by: Howard McLauchlan <hmclauchlan@fb.com>
>>> ---
>>> based on 4.16-rc5
>>>   include/linux/syscalls.h | 3 +++
>>>   1 file changed, 3 insertions(+)
>>>
>>> diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
>>> index a78186d826d7..e8c6d63ace78 100644
>>> --- a/include/linux/syscalls.h
>>> +++ b/include/linux/syscalls.h
>>> @@ -191,6 +191,8 @@ static inline int is_syscall_trace_event(struct trace_event_call *tp_event)
>>>     #define SYSCALL_DEFINE0(sname)                    \
>>>       SYSCALL_METADATA(_##sname, 0);                \
>>> +    asmlinkage long sys_##sname(void);            \
>>> +    ALLOW_ERROR_INJECTION(sys_##sname, ERRNO);        \
>>>       asmlinkage long sys_##sname(void)
> 
> duplication of asmlinkage in the above?
> 
The pre-declaration is necessary to ensure ALLOW_ERROR_INJECTION works appropriately. There can be syscalls
that are not pre-declared elsewhere which will fail compilation if not declared in this block.
>>>     #define SYSCALL_DEFINE1(name, ...) SYSCALL_DEFINEx(1, _##name, __VA_ARGS__)
>>> @@ -210,6 +212,7 @@ static inline int is_syscall_trace_event(struct trace_event_call *tp_event)
>>>   #define __SYSCALL_DEFINEx(x, name, ...)                    \
>>>       asmlinkage long sys##name(__MAP(x,__SC_DECL,__VA_ARGS__))    \
>>>           __attribute__((alias(__stringify(SyS##name))));        \
>>> +    ALLOW_ERROR_INJECTION(sys##name, ERRNO);            \
>>>       static inline long SYSC##name(__MAP(x,__SC_DECL,__VA_ARGS__));    \
>>>       asmlinkage long SyS##name(__MAP(x,__SC_LONG,__VA_ARGS__));    \
>>>       asmlinkage long SyS##name(__MAP(x,__SC_LONG,__VA_ARGS__))    \
>>> -- 
>>> 2.14.1
>>>
>>
>> Adding a few more people to Cc
>>