[PATCH v7 3/6] powerpc/perf: consolidate read_user_stack_32

[Michael Ellerman]

Michal Suchánek <msuchanek at suse.de> writes:
> On Mon, 02 Sep 2019 14:01:17 +1000
> Michael Ellerman <mpe at ellerman.id.au> wrote:
>> Michael Ellerman <mpe at ellerman.id.au> writes:
>> > Michal Suchanek <msuchanek at suse.de> writes:  
>> ...
>> >> @@ -295,6 +279,12 @@ static inline int current_is_64bit(void)
>> >>  }
>> >>  
>> >>  #else  /* CONFIG_PPC64 */
>> >> +static int read_user_stack_slow(void __user *ptr, void *buf, int nb)
>> >> +{
>> >> +	return 0;
>> >> +}
>> >> +#endif /* CONFIG_PPC64 */  
>> >
>> > Ending the PPC64 else case here, and then restarting it below with an
>> > ifndef means we end up with two parts of the file that define 32-bit
>> > code, with a common chunk in the middle, which I dislike.
>> >
>> > I'd rather you add the empty read_user_stack_slow() in the existing
>> > #else section and then move read_user_stack_32() below the whole ifdef
>> > PPC64/else/endif section.
>> >
>> > Is there some reason that doesn't work?  
>> 
>> Gah, I missed that you split the whole file later in the series. Any
>> reason you did it in two steps rather than moving patch 6 earlier in the
>> series?
>
> To make this patch readable.

But it's not very readable :)

You also retained the comment about the 32-bit behaviour which is now a
bit confusing, because the function is used on both 32 & 64-bit.

I think moving it as I suggested in my first reply makes for a better
diff, something like eg:

diff --git a/arch/powerpc/perf/callchain.c b/arch/powerpc/perf/callchain.c
index c84bbd4298a0..82c0f81b89a5 100644
--- a/arch/powerpc/perf/callchain.c
+++ b/arch/powerpc/perf/callchain.c
@@ -165,22 +165,6 @@ static int read_user_stack_64(unsigned long __user *ptr, unsigned long *ret)
 	return read_user_stack_slow(ptr, ret, 8);
 }
 
-static int read_user_stack_32(unsigned int __user *ptr, unsigned int *ret)
-{
-	if ((unsigned long)ptr > TASK_SIZE - sizeof(unsigned int) ||
-	    ((unsigned long)ptr & 3))
-		return -EFAULT;
-
-	pagefault_disable();
-	if (!__get_user_inatomic(*ret, ptr)) {
-		pagefault_enable();
-		return 0;
-	}
-	pagefault_enable();
-
-	return read_user_stack_slow(ptr, ret, 4);
-}
-
 static inline int valid_user_sp(unsigned long sp, int is_64)
 {
 	if (!sp || (sp & 7) || sp > (is_64 ? TASK_SIZE : 0x100000000UL) - 32)
@@ -295,27 +279,6 @@ static inline int current_is_64bit(void)
 }
 
 #else  /* CONFIG_PPC64 */
-/*
- * On 32-bit we just access the address and let hash_page create a
- * HPTE if necessary, so there is no need to fall back to reading
- * the page tables.  Since this is called at interrupt level,
- * do_page_fault() won't treat a DSI as a page fault.
- */
-static int read_user_stack_32(unsigned int __user *ptr, unsigned int *ret)
-{
-	int rc;
-
-	if ((unsigned long)ptr > TASK_SIZE - sizeof(unsigned int) ||
-	    ((unsigned long)ptr & 3))
-		return -EFAULT;
-
-	pagefault_disable();
-	rc = __get_user_inatomic(*ret, ptr);
-	pagefault_enable();
-
-	return rc;
-}
-
 static inline void perf_callchain_user_64(struct perf_callchain_entry_ctx *entry,
 					  struct pt_regs *regs)
 {
@@ -333,6 +296,11 @@ static inline int valid_user_sp(unsigned long sp, int is_64)
 	return 1;
 }
 
+static int read_user_stack_slow(void __user *ptr, void *buf, int nb)
+{
+	return 0;
+}
+
 #define __SIGNAL_FRAMESIZE32	__SIGNAL_FRAMESIZE
 #define sigcontext32		sigcontext
 #define mcontext32		mcontext
@@ -341,6 +309,33 @@ static inline int valid_user_sp(unsigned long sp, int is_64)
 
 #endif /* CONFIG_PPC64 */
 
+static int read_user_stack_32(unsigned int __user *ptr, unsigned int *ret)
+{
+	int rc;
+
+	if ((unsigned long)ptr > TASK_SIZE - sizeof(unsigned int) ||
+	    ((unsigned long)ptr & 3))
+		return -EFAULT;
+
+	pagefault_disable();
+	rc = __get_user_inatomic(*ret, ptr);
+	pagefault_enable();
+
+	/*
+	 * On 32-bit we just access the address and let hash_page() create a
+	 * HPTE if necessary, so there is no need to fall back to reading the
+	 * page tables. Since this is called at interrupt level, do_page_fault()
+	 * won't treat a DSI as a page fault.
+	 *
+	 * On 64-bit if the access faults we fall back to
+	 * read_user_stack_slow(), see the comment there for more details.
+	 */
+	if (IS_ENABLED(CONFIG_PPC64) && rc)
+		return read_user_stack_slow(ptr, ret, 4);
+
+	return rc;
+}
+
 /*
  * Layout for non-RT signal frames
  */


cheers