[3/8] signal/sparc: Document a conflict with SI_USER with SIGFPE

Submitted by Eric W. Biederman on June 30, 2017, 12:39 p.m.

Details

Message ID 20170630123906.8865-3-ebiederm@xmission.com
State New
Series "Series without cover letter"
Headers show

Commit Message

Eric W. Biederman June 30, 2017, 12:39 p.m.
Setting si_code to __SI_FAULT results in a userspace seeing
an si_code of 0.  This is the same si_code as SI_USER.  Posix
and common sense requires that SI_USER not be a signal specific
si_code.  As such this use of 0 for the si_code is a pretty
horribly broken ABI.

This was introduced in 2.3.41 so this mess has had a long time for
people to be able to start depending on it.

As this bug has existed for 17 years already I don't know if it is
worth fixing.  It is definitely worth documenting what is going
on so that no one decides to copy this bad decision.

Cc: "David S. Miller" <davem@davemloft.net>
Cc: sparclinux@vger.kernel.org
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 arch/sparc/include/uapi/asm/siginfo.h | 5 +++++
 arch/sparc/kernel/traps_32.c          | 2 +-
 arch/sparc/kernel/traps_64.c          | 2 +-
 3 files changed, 7 insertions(+), 2 deletions(-)

Patch hide | download patch | download mbox

diff --git a/arch/sparc/include/uapi/asm/siginfo.h b/arch/sparc/include/uapi/asm/siginfo.h
index 2d9b79ccaa50..6bc5c677e92f 100644
--- a/arch/sparc/include/uapi/asm/siginfo.h
+++ b/arch/sparc/include/uapi/asm/siginfo.h
@@ -17,6 +17,11 @@ 
 #define SI_NOINFO	32767		/* no information in siginfo_t */
 
 /*
+ * SIGFPE si_codes
+ */
+#define FPE_FIXME	(__SI_FAULT|0)	/* Broken dup of SI_USER */
+
+/*
  * SIGEMT si_codes
  */
 #define EMT_TAGOVF	(__SI_FAULT|1)	/* tag overflow */
diff --git a/arch/sparc/kernel/traps_32.c b/arch/sparc/kernel/traps_32.c
index 466d4aed06c7..581cf35ee7e3 100644
--- a/arch/sparc/kernel/traps_32.c
+++ b/arch/sparc/kernel/traps_32.c
@@ -306,7 +306,7 @@  void do_fpe_trap(struct pt_regs *regs, unsigned long pc, unsigned long npc,
 	info.si_errno = 0;
 	info.si_addr = (void __user *)pc;
 	info.si_trapno = 0;
-	info.si_code = __SI_FAULT;
+	info.si_code = FPE_FIXME;
 	if ((fsr & 0x1c000) == (1 << 14)) {
 		if (fsr & 0x10)
 			info.si_code = FPE_FLTINV;
diff --git a/arch/sparc/kernel/traps_64.c b/arch/sparc/kernel/traps_64.c
index 196ee5eb4d48..e882e128faa3 100644
--- a/arch/sparc/kernel/traps_64.c
+++ b/arch/sparc/kernel/traps_64.c
@@ -2258,7 +2258,7 @@  static void do_fpe_common(struct pt_regs *regs)
 		info.si_errno = 0;
 		info.si_addr = (void __user *)regs->tpc;
 		info.si_trapno = 0;
-		info.si_code = __SI_FAULT;
+		info.si_code = FPE_FIXME;
 		if ((fsr & 0x1c000) == (1 << 14)) {
 			if (fsr & 0x10)
 				info.si_code = FPE_FLTINV;

Comments

David Miller June 30, 2017, 4:45 p.m.
From: "Eric W. Biederman" <ebiederm@xmission.com>
Date: Fri, 30 Jun 2017 07:39:01 -0500

> diff --git a/arch/sparc/include/uapi/asm/siginfo.h b/arch/sparc/include/uapi/asm/siginfo.h
> index 2d9b79ccaa50..6bc5c677e92f 100644
> --- a/arch/sparc/include/uapi/asm/siginfo.h
> +++ b/arch/sparc/include/uapi/asm/siginfo.h
> @@ -17,6 +17,11 @@
>  #define SI_NOINFO	32767		/* no information in siginfo_t */
>  
>  /*
> + * SIGFPE si_codes
> + */
> +#define FPE_FIXME	(__SI_FAULT|0)	/* Broken dup of SI_USER */
> +
> +/*
>   * SIGEMT si_codes
>   */
>  #define EMT_TAGOVF	(__SI_FAULT|1)	/* tag overflow */

It's one thing to say FIXME in a comment in a kernel local header or
C file.

It's quite another to put this into the name of a macro which has
visibility in the global user compilation namespace.

I don't think you should really do that.
Eric W. Biederman June 30, 2017, 6:13 p.m.
David Miller <davem@davemloft.net> writes:

> From: "Eric W. Biederman" <ebiederm@xmission.com>
> Date: Fri, 30 Jun 2017 07:39:01 -0500
>
>> diff --git a/arch/sparc/include/uapi/asm/siginfo.h b/arch/sparc/include/uapi/asm/siginfo.h
>> index 2d9b79ccaa50..6bc5c677e92f 100644
>> --- a/arch/sparc/include/uapi/asm/siginfo.h
>> +++ b/arch/sparc/include/uapi/asm/siginfo.h
>> @@ -17,6 +17,11 @@
>>  #define SI_NOINFO	32767		/* no information in siginfo_t */
>>  
>>  /*
>> + * SIGFPE si_codes
>> + */
>> +#define FPE_FIXME	(__SI_FAULT|0)	/* Broken dup of SI_USER */
>> +
>> +/*
>>   * SIGEMT si_codes
>>   */
>>  #define EMT_TAGOVF	(__SI_FAULT|1)	/* tag overflow */
>
> It's one thing to say FIXME in a comment in a kernel local header or
> C file.
>
> It's quite another to put this into the name of a macro which has
> visibility in the global user compilation namespace.
>
> I don't think you should really do that.

Good point.

Sigh.   It almost fits because we did do something off in the uapi
exported to userspace and we don't have a header file definition for
that case.

Still.  At this point arch/sparc/include/asm/siginfo.h is
a better fit for that definition.

I will respin and fix that.

I wish I knew what would make a better default floating point si_code on
sparc.  Using 0 aka SI_USER is doesn't fit at all.  Sigh.

Unfortunately I don't know the architecture well enough to even guess
what is going on in do_fpe_common when when no bits in fsr are set.

Any suggests for a better fix than just documenting that linux does
something weird and ill advised here?

Eric
Eric W. Biederman July 18, 2017, 1:43 p.m.
ebiederm@xmission.com (Eric W. Biederman) writes:

> David Miller <davem@davemloft.net> writes:
>
>> From: "Eric W. Biederman" <ebiederm@xmission.com>
>> Date: Fri, 30 Jun 2017 07:39:01 -0500
>>
>>> diff --git a/arch/sparc/include/uapi/asm/siginfo.h b/arch/sparc/include/uapi/asm/siginfo.h
>>> index 2d9b79ccaa50..6bc5c677e92f 100644
>>> --- a/arch/sparc/include/uapi/asm/siginfo.h
>>> +++ b/arch/sparc/include/uapi/asm/siginfo.h
>>> @@ -17,6 +17,11 @@
>>>  #define SI_NOINFO	32767		/* no information in siginfo_t */
>>>  
>>>  /*
>>> + * SIGFPE si_codes
>>> + */
>>> +#define FPE_FIXME	(__SI_FAULT|0)	/* Broken dup of SI_USER */
>>> +
>>> +/*
>>>   * SIGEMT si_codes
>>>   */
>>>  #define EMT_TAGOVF	(__SI_FAULT|1)	/* tag overflow */
>>
>> It's one thing to say FIXME in a comment in a kernel local header or
>> C file.
>>
>> It's quite another to put this into the name of a macro which has
>> visibility in the global user compilation namespace.
>>
>> I don't think you should really do that.
>
> Good point.
>
> Sigh.   It almost fits because we did do something off in the uapi
> exported to userspace and we don't have a header file definition for
> that case.
>
> Still.  At this point arch/sparc/include/asm/siginfo.h is
> a better fit for that definition.
>
> I will respin and fix that.

Except arch/<arch>/include/asm/siginfo.h has now been eliminated.
So these definitions have to be
#ifdef __KERNEL__
#define ..._FIXME
#endif

In the uapi but not visible in the headers outside of the kernel.
Which seems appropriate for these values.

Eric