Skip to contentSkip to conversation

Blackfin updates for 3.11

GIT PULL Steven Miao 5 participants 8 messages Jul 9, 2013 – Jul 16, 2013

lore
Share & tools
lore
#0 lore raw

Hi Linus and Stephen,

I've signed up for an kernel.org account and moved the blackfin tree to kernel.org for convenience as some developers' suggestion. Pls update the url to:
http://git.kernel.org/pub/scm/linux/kernel/git/realmz6/blackfin-linux.git

Hi Linus,

please pull blackfin updates for Linux 3.11, some minor changes for performance and bug fixes.

The following changes since commit 8bb495e3f02401ee6f76d1b1d77f3ac9f079e376:

Linux 3.10 (2013-06-30 15:13:29 -0700)

are available in the git repository at:

http://git.kernel.org/pub/scm/linux/kernel/git/realmz6/blackfin-linux.git tags/blackfin-for-linus

for you to fetch changes up to 150382a53d11256e5666c86525c8bf8d23684532:

smp: refine bf561 smpboot code (2013-07-09 15:50:38 +0800)

----------------------------------------------------------------
blackfin updates for Linux 3.11

----------------------------------------------------------------
Scott Jiang (1):
bf609: rename bfin6xx_spi to bfin_spi3

Sonic Zhang (2):
kgdb: blackfin: include irq_regs.h in kgdb.c
bf609: add cpu revision 0.1

Steven Miao (2):
bf609: stmmac: fix build after stmmac_mdio_bus_data changed
smp: refine bf561 smpboot code

 arch/blackfin/Kconfig                              |    2 +-
 .../include/asm/{bfin6xx_spi.h => bfin_spi3.h}     |    4 ++--
 arch/blackfin/kernel/kgdb.c                        |    1 +
 arch/blackfin/mach-bf561/smp.c                     |    7 +++----
 arch/blackfin/mach-bf609/boards/ezkit.c            |   19 +++++++++----------
 arch/blackfin/mach-common/smp.c                    |   18 ++++++++----------
 6 files changed, 24 insertions(+), 27 deletions(-)
 rename arch/blackfin/include/asm/{bfin6xx_spi.h => bfin_spi3.h} (99%)
#1 lore raw

Hi Steven,

Quoted 3 lines · On Tue, 9 Jul 2013 17:15:30 +0800 Steven Miao <[email protected]> wrote:

I've signed up for an kernel.org account and moved the blackfin tree to kernel.org for convenience as some developers' suggestion. Pls update the url to:
http://git.kernel.org/pub/scm/linux/kernel/git/realmz6/blackfin-linux.git

That tree only has for-linus and master branches. linux-next uses the blackfin-linus branch of your github tree ... so what should I use now?

Signature · 2 lines
Cheers,
Stephen Rothwell                    [email protected]
#2 lore raw

Hi Stephen,

Quoted 9 lines · Stephen Rothwell

Hi Steven,

Quoted 3 lines · On Tue, 9 Jul 2013 17:15:30 +0800 Steven Miao <[email protected]> wrote:

I've signed up for an kernel.org account and moved the blackfin tree to kernel.org for convenience as some developers' suggestion. Pls update the url to:
http://git.kernel.org/pub/scm/linux/kernel/git/realmz6/blackfin-linux.git

That tree only has for-linus and master branches. linux-next uses the blackfin-linus branch of your github tree ... so what should I use now?

Pls use for-linus branch on
http://git.kernel.org/pub/scm/linux/kernel/git/realmz6/blackfin-linux.git
Thanks.

Quoted 4 lines

--
Cheers,
Stephen Rothwell [email protected]

#4 lore raw
In reply to Steven Miao

Hi Steven,

Quoted 2 lines · Steven Miao

Steven Miao (2):
smp: refine bf561 smpboot code

Unfortunately this landed too late in -next to notice:

arch/blackfin/mach-common/smp.c:150:2: error: conversion to non-scalar type requested
arch/blackfin/mach-common/smp.c:185:23: error: incompatible types when assigning to type 'atomic_t' from type 'int' arch/blackfin/mach-common/smp.c:186:24: error: incompatible types when assigning to type 'atomic_t' from type 'int'
make[2]: *** [arch/blackfin/mach-common/smp.o] Error 1

http://kisskb.ellerman.id.au/kisskb/buildresult/9115952/

It's a bit embarrassing to break your own core code (with a patch that was submitted for public review in August 2012?).

The last two can be trivially-fixed with the following (whitespace-damaged gmail)
patch:

arch/blackfin/mach-common/smp.c +2 -2
@@ -182,8 +182,8 @@ static void bfin_ipi_init(void)
  struct ipi_data *bfin_ipi_data;
  for_each_possible_cpu(cpu) {
  bfin_ipi_data = &per_cpu(bfin_ipi, cpu);
- bfin_ipi_data->bits = 0;
- bfin_ipi_data->count = 0;
+ atomic_set(&bfin_ipi_data->bits, 0);
+ atomic_set(&bfin_ipi_data->count, 0);
  }
 }
 

but the first one is non-trivial: using xchg() on atomic_t is a bit gross...

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

#5 lore raw
Quoted 2 lines · Geert Uytterhoeven

but the first one is non-trivial: using xchg() on atomic_t is a bit gross...

It's also broken. There's no guarantee that an "atomic_t" is just a value. Now, the old sparc32 stuff (which hid lock bits in atomic_t) may be gone, but it's still the case that atomic_t may not actually work with xchg.

(In *practice* it works on normal architectures, so I'm not saying that we don't have it, but it's a bug if we do).

There are "atomic_xchg()" and "atomic_xchg64()" functions that are supported.

Linus

#6 lore raw
Quoted 10 lines · Linus Torvalds
Quoted 2 lines · Geert Uytterhoeven

but the first one is non-trivial: using xchg() on atomic_t is a bit gross...

It's also broken. There's no guarantee that an "atomic_t" is just a value. Now, the old sparc32 stuff (which hid lock bits in atomic_t) may be gone, but it's still the case that atomic_t may not actually work with xchg.

Quoted 4 lines

(In *practice* it works on normal architectures, so I'm not saying that we don't have it, but it's a bug if we do).

There are "atomic_xchg()" and "atomic_xchg64()" functions that are supported.

I'll send a patch to fix it.

Quoted 2 lines

Linus

#7 lore raw
In reply to Geert Uytterhoeven

Hi Geert,

Quoted 28 lines · Geert Uytterhoeven

Hi Steven,

Quoted 2 lines · Steven Miao

Steven Miao (2):
smp: refine bf561 smpboot code

Unfortunately this landed too late in -next to notice:

arch/blackfin/mach-common/smp.c:150:2: error: conversion to non-scalar type requested
arch/blackfin/mach-common/smp.c:185:23: error: incompatible types when assigning to type 'atomic_t' from type 'int' arch/blackfin/mach-common/smp.c:186:24: error: incompatible types when assigning to type 'atomic_t' from type 'int'
make[2]: *** [arch/blackfin/mach-common/smp.o] Error 1

http://kisskb.ellerman.id.au/kisskb/buildresult/9115952/

It's a bit embarrassing to break your own core code (with a patch that was submitted for public review in August 2012?).

The last two can be trivially-fixed with the following (whitespace-damaged gmail)
patch:

arch/blackfin/mach-common/smp.c static void bfin_ipi_init(void) · line 188

Quoted 9 lines
arch/blackfin/mach-common/smp.c +2 -2
@@ -182,8 +182,8 @@ static void bfin_ipi_init(void)
  struct ipi_data *bfin_ipi_data;
  for_each_possible_cpu(cpu) {
  bfin_ipi_data = &per_cpu(bfin_ipi, cpu);
- bfin_ipi_data->bits = 0;
- bfin_ipi_data->count = 0;
+ atomic_set(&bfin_ipi_data->bits, 0);
+ atomic_set(&bfin_ipi_data->count, 0);
  }
 }

Sorry, I should use atomic_ops to hanle atomic_t type.

Quoted 13 lines

but the first one is non-trivial: using xchg() on atomic_t is a bit gross...

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds