Re: [PATCH 0/14] score - New architecure port to SunplusCT S+CORE

Re: [PATCH 0/14] score - New architecure port to SunplusCT S+CORE

[Arnd Bergmann]

On Sunday 31 May 2009 09:58:28 liqin.chen@sunplusct.com wrote:
> Hi Arnd,
> 
> > Have you had a chance to look at the asm-generic header files I posted
> > a few times?
> > 
> > ...
> > Please tell me if you find problems in the generic header files during
> > the conversion.
> > 
> 
> I can't clone the linux/kernel/git/arnd/asm-generic.git repository, 
> and have not test score arch code until now.

Hmm, what is the error message you get? It works fine here:

arnd@wuerfel:/tmp$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic.git
Initialized empty Git repository in /tmp/asm-generic/.git/
remote: Counting objects: 1180676, done.
remote: Compressing objects: 100% (189568/189568), done.
remote: Total 1180676 (delta 984963), reused 1179831 (delta 984290)
Receiving objects: 100% (1180676/1180676), 288.80 MiB | 1308 KiB/s, done.
Resolving deltas: 100% (984963/984963), done.
Checking out files: 100% (27918/27918), done.
arnd@wuerfel:/tmp$ cd asm-generic/
arnd@wuerfel:/tmp/asm-generic$ git checkout -b next origin/next
Branch next set up to track remote branch refs/remotes/origin/next.
Switched to a new branch "next"

> In addition, next time I send the patches mail should base on
> linux/kernel/git/arnd/asm-generic.git or
> linux/kernel/git/sfr/linux-next.git?

Feel free to base you code on top of asm-generic.git#next, but tell me
when you intend to do that so I can warn you of possibly rebases.

The easiest way for you to merge code would be if you can arrange to have
a git account on kernel.org (see http://www.kernel.org/faq/#account), on
your company web server, or on a free hosting service like github.com
or gitorious.org.

Your patches or git tree should merge fine with the linux-next tree so
that Stephen can pick them up as early as possibly, but in general they
should not be _based_ on linux-next because that would mean that
Linus needs to pull all of linux-next in order to merge your code.

	Arnd <><

Re: [PATCH 0/14] score - New architecure port to SunplusCT S+CORE

[liqin.chen]

Thanks Arnd, our MIS help me solve the download problem.

> Feel free to base you code on top of asm-generic.git#next, but tell me
> when you intend to do that so I can warn you of possibly rebases.

Last week, I had updated score code base on asm-generic.git#next.

and I will apply a account on kernel.org, But before I get the account,
can you let score code merge in asm-generic.git?

liqin

--

Re: [PATCH 0/14] score - New architecure port to SunplusCT S+CORE

[Arnd Bergmann]

On Monday 08 June 2009, liqin.chen@sunplusct.com wrote:
> Thanks Arnd, our MIS help me solve the download problem.
> 
> > Feel free to base you code on top of asm-generic.git#next, but tell me
> > when you intend to do that so I can warn you of possibly rebases.
> 
> Last week, I had updated score code base on asm-generic.git#next.
> 
> and I will apply a account on kernel.org, But before I get the account,
> can you let score code merge in asm-generic.git?

I can take your patches and apply them to a new branch in my tree
if that helps. I hope that Linus can pull my tree early in the
merge window so your architecture port and the microblaze changes
based on it can also get in soon.

I guess for the new architecture, it makes sense to combine all
your patches into a single changeset in git, since they don't
make sense until you have the complete set, and your changeset
comments are rather boring so far anyway.

Can you write a few paragraphs of text about your architecture
for the initial changeset comment? I think it would be good
to permanently store some facts in there, e.g. in which ways
the architecture is different from others, where to get a
toolchain and what kinds of hardware exist for it.

I'll go through the patches once more and see if I find anything
that should still be changed before the code goes in, but
from a brief look, the code looks pretty good now.

	Arnd <><

Re: [PATCH 0/14] score - New architecure port to SunplusCT S+CORE

[liqin.chen]

Thanks Arnd, tglx, Sam and all who pay attention to score code.

We will update score arch code to catch up the 
linux/kernel code level. The next patches will
send out soon.

Best Regards
liqin

--


Arnd Bergmann <arnd@arndb.de> 写于 2009-06-10 00:37:28:

> On Monday 08 June 2009, liqin.chen@sunplusct.com wrote:
> > Thanks Arnd, our MIS help me solve the download problem.
> > 
> > > Feel free to base you code on top of asm-generic.git#next, but tell 
me
> > > when you intend to do that so I can warn you of possibly rebases.
> > 
> > Last week, I had updated score code base on asm-generic.git#next.
> > 
> > and I will apply a account on kernel.org, But before I get the 
account,
> > can you let score code merge in asm-generic.git?
> 
> I can take your patches and apply them to a new branch in my tree
> if that helps. I hope that Linus can pull my tree early in the
> merge window so your architecture port and the microblaze changes
> based on it can also get in soon.
> 
> I guess for the new architecture, it makes sense to combine all
> your patches into a single changeset in git, since they don't
> make sense until you have the complete set, and your changeset
> comments are rather boring so far anyway.
> 
> Can you write a few paragraphs of text about your architecture
> for the initial changeset comment? I think it would be good
> to permanently store some facts in there, e.g. in which ways
> the architecture is different from others, where to get a
> toolchain and what kinds of hardware exist for it.
> 
> I'll go through the patches once more and see if I find anything
> that should still be changed before the code goes in, but
> from a brief look, the code looks pretty good now.
> 
>    Arnd <><

[PATCH 0/14] score - New architecure port to SunplusCT S+CORE

[liqin.chen]

Hi tglx, Arnd, Kyle, Sam and linus,

According to your comment last week, We had updated the linux/score code,
use arch blackfin code as reference.
 
because some reasons below, I have not fixed these problems you mentioned
 
1:      bdflush is not a nop, remove it?

2:      because module_arch_cleanup, makcine_halt, machine_power_off and 
some other funcs are global, 
        so i have not move them into .h and not declare them as static 
inline.

3: 
> +#define __NR_exit                      (__NR_Linux +   1)
> +#define __NR_fork                      (__NR_Linux +   2)
> +#define __NR_read                      (__NR_Linux +   3)
> +#define __NR_write                     (__NR_Linux +   4)
> +#define __NR_open                      (__NR_Linux +   5)
> +#define __NR_close                     (__NR_Linux +   6)
> +#define __NR_waitpid                   (__NR_Linux +   7)
> +#define __NR_creat                     (__NR_Linux +   8)
> +#define __NR_link                      (__NR_Linux +   9)
> +#define __NR_unlink                    (__NR_Linux +  10)
> +#define __NR_execve                    (__NR_Linux +  11)
> +#define __NR_chdir                     (__NR_Linux +  12)
> +#define __NR_time                      (__NR_Linux +  13)
> +#define __NR_mknod                     (__NR_Linux +  14)
> +#define __NR_chmod                     (__NR_Linux +  15)
> +#define __NR_lchown                    (__NR_Linux +  16)
> +#define __NR_break                     (__NR_Linux +  17)
I still leave these syscalls there, bacause i think it clear to everybody
even linux have the new syscall replace them.

4:      asm/atomic.h and asm/bitops.h generic version have not finished,
        so I dont updated them.

5: 
> +#define __ARCH_WANT_SYS_FADVISE64
> +#define __ARCH_OMIT_COMPAT_SYS_GETDENTS64
> +#define __ARCH_WANT_IPC_PARSE_VERSION
> +#define __ARCH_WANT_OLD_READDIR
> +#define __ARCH_WANT_SYS_ALARM
> +#define __ARCH_WANT_SYS_GETHOSTNAME
> +#define __ARCH_WANT_SYS_PAUSE
> +#define __ARCH_WANT_SYS_SGETMASK
> +#define __ARCH_WANT_SYS_UTIME
> +#define __ARCH_WANT_SYS_WAITPID
> +#define __ARCH_WANT_SYS_SOCKETCALL
> +#define __ARCH_WANT_SYS_GETPGRP
> +#define __ARCH_WANT_SYS_LLSEEK
> +#define __ARCH_WANT_SYS_NICE
> +#define __ARCH_WANT_SYS_OLD_GETRLIMIT
> +#define __ARCH_WANT_SYS_OLDUMOUNT
> +#define __ARCH_WANT_SYS_SIGPENDING
> +#define __ARCH_WANT_SYS_SIGPROCMASK
> +#define __ARCH_WANT_SYS_RT_SIGACTION
> +#define __ARCH_WANT_STAT64
> +#define __ARCH_WANT_SYS_TIME
Once the kernel code update, I will remove these code.

6:      sys_sigaltstack should rename to sys_rt_sigaltstack?

7:      use score-linux compile linux/score kernel still have some 
warning, 
        but we think it not the problems in score head files.
 
8:      score have big and little endian support, now we only use 
        little endian running linux.


        score-linux toolchain could be download from
 
http://www.sunplusct.com/images/linux-score-patch/linux-score-20090324.patch
 
        linux/score lastest patch place at 
 
http://www.sunplusct.com/images/linux-score-patch/linux-score-20090324.patch

-- 
Best regards
Liqin
liqin.chen@sunplusct.com

[PATCH 0/14] score - New architecure port to SunplusCT S+CORE

[liqin.chen]

Hi tglx, Arnd, Kyle, Sam and linus,

According to your comment last week, We had updated the linux/score code,
use arch blackfin code as reference.
 
because some reasons below, I have not fixed these problems you mentioned
 
1:      bdflush is not a nop, remove it?

2:      because module_arch_cleanup, makcine_halt, machine_power_off and 
some other funcs are global, 
        so i have not move them into .h and not declare them as static 
inline.

3: 
> +#define __NR_exit                      (__NR_Linux +   1)
> +#define __NR_fork                      (__NR_Linux +   2)
> +#define __NR_read                      (__NR_Linux +   3)
> +#define __NR_write                     (__NR_Linux +   4)
> +#define __NR_open                      (__NR_Linux +   5)
> +#define __NR_close                     (__NR_Linux +   6)
> +#define __NR_waitpid                   (__NR_Linux +   7)
> +#define __NR_creat                     (__NR_Linux +   8)
> +#define __NR_link                      (__NR_Linux +   9)
> +#define __NR_unlink                    (__NR_Linux +  10)
> +#define __NR_execve                    (__NR_Linux +  11)
> +#define __NR_chdir                     (__NR_Linux +  12)
> +#define __NR_time                      (__NR_Linux +  13)
> +#define __NR_mknod                     (__NR_Linux +  14)
> +#define __NR_chmod                     (__NR_Linux +  15)
> +#define __NR_lchown                    (__NR_Linux +  16)
> +#define __NR_break                     (__NR_Linux +  17)
I still leave these syscalls there, bacause i think it clear to everybody
even linux have the new syscall replace them.

4:      asm/atomic.h and asm/bitops.h generic version have not finished,
        so I dont updated them.

5: 
> +#define __ARCH_WANT_SYS_FADVISE64
> +#define __ARCH_OMIT_COMPAT_SYS_GETDENTS64
> +#define __ARCH_WANT_IPC_PARSE_VERSION
> +#define __ARCH_WANT_OLD_READDIR
> +#define __ARCH_WANT_SYS_ALARM
> +#define __ARCH_WANT_SYS_GETHOSTNAME
> +#define __ARCH_WANT_SYS_PAUSE
> +#define __ARCH_WANT_SYS_SGETMASK
> +#define __ARCH_WANT_SYS_UTIME
> +#define __ARCH_WANT_SYS_WAITPID
> +#define __ARCH_WANT_SYS_SOCKETCALL
> +#define __ARCH_WANT_SYS_GETPGRP
> +#define __ARCH_WANT_SYS_LLSEEK
> +#define __ARCH_WANT_SYS_NICE
> +#define __ARCH_WANT_SYS_OLD_GETRLIMIT
> +#define __ARCH_WANT_SYS_OLDUMOUNT
> +#define __ARCH_WANT_SYS_SIGPENDING
> +#define __ARCH_WANT_SYS_SIGPROCMASK
> +#define __ARCH_WANT_SYS_RT_SIGACTION
> +#define __ARCH_WANT_STAT64
> +#define __ARCH_WANT_SYS_TIME
Once the kernel code update, I will remove these code.

6:      sys_sigaltstack should rename to sys_rt_sigaltstack?

7:      use score-linux compile linux/score kernel still have some 
warning, 
        but we think it not the problems in score head files.
 
8:      score have big and little endian support, now we only use 
        little endian running linux.


        score-linux toolchain could be download from
 
http://www.sunplusct.com/images/linux-score-patch/linux-score-20090324.patch
 
        linux/score lastest patch place at 
 
http://www.sunplusct.com/images/linux-score-patch/linux-score-20090324.patch

-- 
Best regards
Liqin
liqin.chen@sunplusct.com

Re: [PATCH 0/14] score - New architecure port to SunplusCT S+CORE

[Andrew Morton]

On Wed, 8 Apr 2009 15:18:30 +0800 liqin.chen@sunplusct.com wrote:
>

This emailed patches are badly wordwrapped and

> http://www.sunplusct.com/images/linux-score-patch/linux-score-20090324.patch

has no changelogs, no attributions and no Signed-off-by:'s.

That's not a problem right now as we can still review the code, but
these things will need to be sorted out soon, please.

Re: [PATCH 0/14] score - New architecure port to SunplusCT S+CORE

[Arnd Bergmann]

Hi Liquin,

Have you had a chance to look at the asm-generic header files I posted
a few times?

I think it would be ideal if you can change your code to use the
asm-generic version whereever possible, by replacing your files
with something like

arch/score/include/asm/foo.h:
#include <asm-generic/foo.h>

In particular, all of the exported header files that get included
in user space should use only the asm-generic version, with the
exception of byteorder.h, elf.h and ptrace.h, which are by definition
architecture specific.

Please tell me if you find problems in the generic header files during
the conversion.

	Arnd <><

Re: [PATCH 0/14] score - New architecure port to SunplusCT S+CORE

[liqin.chen]

Hi Arnd,

> Have you had a chance to look at the asm-generic header files I posted
> a few times?
> 
> ...
> Please tell me if you find problems in the generic header files during
> the conversion.
> 

I can't clone the linux/kernel/git/arnd/asm-generic.git repository, 
and have not test score arch code until now.

In addition, next time I send the patches mail should base on
linux/kernel/git/arnd/asm-generic.git or
linux/kernel/git/sfr/linux-next.git?

Liqin