| Message ID | 20190430090044.16345-1-tranmanphong@gmail.com |
|---|---|
| State | Superseded, archived |
| Headers | show |
| Series | of: replace be32_to_cpu to be32_to_cpup | expand |
| Context | Check | Description |
|---|---|---|
| robh/checkpatch | success |
+ Nick and the list On Tue, Apr 30, 2019 at 04:00:44PM +0700, Phong Tran wrote: > The cell is a pointer to __be32. > with the be32_to_cpu a lot of clang warning show that: > > ./include/linux/of.h:238:37: warning: multiple unsequenced modifications > to 'cell' [-Wunsequenced] > r = (r << 32) | be32_to_cpu(*(cell++)); > ^~ > ./include/linux/byteorder/generic.h:95:21: note: expanded from macro > 'be32_to_cpu' > ^ > ./include/uapi/linux/byteorder/little_endian.h:40:59: note: expanded > from macro '__be32_to_cpu' > ^ > ./include/uapi/linux/swab.h:118:21: note: expanded from macro '__swab32' > ___constant_swab32(x) : \ > ^ > ./include/uapi/linux/swab.h:18:12: note: expanded from macro > '___constant_swab32' > (((__u32)(x) & (__u32)0x000000ffUL) << 24) | \ > ^ > > Signed-off-by: Phong Tran <tranmanphong@gmail.com> > --- > include/linux/of.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/include/linux/of.h b/include/linux/of.h > index e240992e5cb6..1c35fc8f19b0 100644 > --- a/include/linux/of.h > +++ b/include/linux/of.h > @@ -235,7 +235,7 @@ static inline u64 of_read_number(const __be32 *cell, int size) > { > u64 r = 0; > while (size--) > - r = (r << 32) | be32_to_cpu(*(cell++)); > + r = (r << 32) | be32_to_cpup(cell++); > return r; > } > > -- > 2.21.0 > While the patch does remove the warning, I am not convinced that this isn't a clang bug based on my brief analysis here: https://github.com/ClangBuiltLinux/linux/issues/460#issuecomment-487808008 However, I'm waiting for people smarter than I am to comment on whether that sounds correct or not. I am not familiar with the various different big/little endian functions enough to review this but thank you for the patch! Nathan
From: Phong Tran > Sent: 30 April 2019 10:01 > The cell is a pointer to __be32. > with the be32_to_cpu a lot of clang warning show that: > > ./include/linux/of.h:238:37: warning: multiple unsequenced modifications > to 'cell' [-Wunsequenced] > r = (r << 32) | be32_to_cpu(*(cell++)); > ^~ > ./include/linux/byteorder/generic.h:95:21: note: expanded from macro > 'be32_to_cpu' > ^ > ./include/uapi/linux/byteorder/little_endian.h:40:59: note: expanded > from macro '__be32_to_cpu' > ^ > ./include/uapi/linux/swab.h:118:21: note: expanded from macro '__swab32' > ___constant_swab32(x) : \ > ^ > ./include/uapi/linux/swab.h:18:12: note: expanded from macro > '___constant_swab32' > (((__u32)(x) & (__u32)0x000000ffUL) << 24) | \ > ^ > > Signed-off-by: Phong Tran <tranmanphong@gmail.com> > --- > include/linux/of.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/include/linux/of.h b/include/linux/of.h > index e240992e5cb6..1c35fc8f19b0 100644 > --- a/include/linux/of.h > +++ b/include/linux/of.h > @@ -235,7 +235,7 @@ static inline u64 of_read_number(const __be32 *cell, int size) > { > u64 r = 0; > while (size--) > - r = (r << 32) | be32_to_cpu(*(cell++)); > + r = (r << 32) | be32_to_cpup(cell++); > return r; That is a very strange loop. It is probably equivalent to: r = be32_to_cpu(*cell); if (size) r = r << 32 | be32_to_cpu(cell[1]); return r; In any case replacing the while with (say): for (; size--; cell++) r = (r << 32) | be32_to_cpu(*cell); would remove the ambiguity. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
On Tue, Apr 30, 2019 at 04:00:44PM +0700, Phong Tran wrote: > > diff --git a/include/linux/of.h b/include/linux/of.h > index e240992e5cb6..1c35fc8f19b0 100644 > --- a/include/linux/of.h > +++ b/include/linux/of.h > @@ -235,7 +235,7 @@ static inline u64 of_read_number(const __be32 *cell, int size) > { > u64 r = 0; > while (size--) > - r = (r << 32) | be32_to_cpu(*(cell++)); > + r = (r << 32) | be32_to_cpup(cell++); > return r; This whole function looks odd. It could simply be replaced with calls to get_unaligned_be64 / get_unaligned_be32. Given that we have a lot of callers we can't easily do that, but at least we could try something like static inline u64 of_read_number(const __be32 *cell, int size) { WARN_ON_ONCE(size < 1); WARN_ON_ONCE(size > 2); if (size == 1) return get_unaligned_be32(cell); return get_unaligned_be64(cell); }
On Tue, Apr 30, 2019 at 8:32 PM Christoph Hellwig <hch@infradead.org> wrote: > > On Tue, Apr 30, 2019 at 04:00:44PM +0700, Phong Tran wrote: > > > > diff --git a/include/linux/of.h b/include/linux/of.h > > index e240992e5cb6..1c35fc8f19b0 100644 > > --- a/include/linux/of.h > > +++ b/include/linux/of.h > > @@ -235,7 +235,7 @@ static inline u64 of_read_number(const __be32 *cell, int size) > > { > > u64 r = 0; > > while (size--) > > - r = (r << 32) | be32_to_cpu(*(cell++)); > > + r = (r << 32) | be32_to_cpup(cell++); > > return r; > > This whole function looks odd. It could simply be replaced with > calls to get_unaligned_be64 / get_unaligned_be32. Given that we have a > lot of callers we can't easily do that, but at least we could try > something like > It's risky. there are many callers of of_read_number(). There is suggestion from David (https://lore.kernel.org/lkml/46b3e8edf27e4c8f98697f9e7f2117d6@AcuMS.aculab.com/) only changing the loop. > static inline u64 of_read_number(const __be32 *cell, int size) > { > WARN_ON_ONCE(size < 1); > WARN_ON_ONCE(size > 2); > > if (size == 1) > return get_unaligned_be32(cell); > return get_unaligned_be64(cell); > } Thank you for your support. Phong.
diff --git a/include/linux/of.h b/include/linux/of.h index e240992e5cb6..1c35fc8f19b0 100644 --- a/include/linux/of.h +++ b/include/linux/of.h @@ -235,7 +235,7 @@ static inline u64 of_read_number(const __be32 *cell, int size) { u64 r = 0; while (size--) - r = (r << 32) | be32_to_cpu(*(cell++)); + r = (r << 32) | be32_to_cpup(cell++); return r; }
The cell is a pointer to __be32. with the be32_to_cpu a lot of clang warning show that: ./include/linux/of.h:238:37: warning: multiple unsequenced modifications to 'cell' [-Wunsequenced] r = (r << 32) | be32_to_cpu(*(cell++)); ^~ ./include/linux/byteorder/generic.h:95:21: note: expanded from macro 'be32_to_cpu' ^ ./include/uapi/linux/byteorder/little_endian.h:40:59: note: expanded from macro '__be32_to_cpu' ^ ./include/uapi/linux/swab.h:118:21: note: expanded from macro '__swab32' ___constant_swab32(x) : \ ^ ./include/uapi/linux/swab.h:18:12: note: expanded from macro '___constant_swab32' (((__u32)(x) & (__u32)0x000000ffUL) << 24) | \ ^ Signed-off-by: Phong Tran <tranmanphong@gmail.com> --- include/linux/of.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)