ABI for MIPS

ABI for MIPS

[Scott A McConnell]

I found these MIPS ABI documents. Ralph mentions on his web page to look
at the System V MIPS ABI book which is out of print are these
equivalent? If I want to understand the ABI for GCC/Linux/MIPS are these
a good place to start?

http://ftp.stenstad.net/pub/mipslinux/docs/ABI/mipsabi.pdf
http://ftp.stenstad.net/pub/mipslinux/docs/ABI/psABI_mips3.0.pdf


-- 
Scott A. McConnell
Phone: (319) 364-0100

MIPS/Linux NonSGI

[Girish Gulawani]

hello all.
i'm trying to bringup linux 2.4.[2|9] on our board based on LSI mips r4k
core.
right now the kernel compiled with gcc-3.0, boots up & can only work with
statically linked commands. hence the root file system mounted from
ramdisk,nfs, & de-dream ide-disk can show some-prompt only if ash.static is
invoked.
this is seen with 2.4.3 & ditto with 2.4.9. the kernel is in un-cached
mode,since we have page-size problem with our core in cached, write
through/back
both, modes. so question is WHY THE COMMANDS WITH SHARED LIBRARY DONOT WORK.
FAILS TO LOAD SHARED LIBRARIES.

the problem no.2 is root on ide-disk. the disk is paritioned & formatted
using a linux pentium-pc. using a master disk the above said statically
linked commands are downloaded to slave disk. the board boots up. however
the bdflush/update process corrupts file-system. the UPDATE PROCESS CORRUPTS
SUPERBLOCK AND INODES WHILE FLUSHING THE DIRTY BUFERS.

PLEASE! PLEASE!! HELP ME ON THIS. THIS NEWSGROUP IS MY LAST HOPE.

many thanks in advance with regards,
girish.



_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

MIPS/Linux NonSGI

[Girish Gulawani]

hello all.
i'm trying to bringup linux 2.4.[2|9] on our board based on LSI mips r4k
core.
right now the kernel compiled with gcc-3.0, boots up & can only work with
statically linked commands. hence the root file system mounted from
ramdisk,nfs, & de-dream ide-disk can show some-prompt only if ash.static is
invoked.
this is seen with 2.4.3 & ditto with 2.4.9. the kernel is in un-cached
mode,since we have page-size problem with our core in cached, write
through/back
both, modes. so question is WHY THE COMMANDS WITH SHARED LIBRARY DONOT WORK.
FAILS TO LOAD SHARED LIBRARIES.

the problem no.2 is root on ide-disk. the disk is paritioned & formatted
using a linux pentium-pc. using a master disk the above said statically
linked commands are downloaded to slave disk. the board boots up. however
the bdflush/update process corrupts file-system. the UPDATE PROCESS CORRUPTS
SUPERBLOCK AND INODES WHILE FLUSHING THE DIRTY BUFERS.

PLEASE! PLEASE!! HELP ME ON THIS. THIS NEWSGROUP IS MY LAST HOPE.

many thanks in advance with regards,
girish.



_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

Re: MIPS/Linux NonSGI

[Ralf Baechle]

On Fri, Jan 25, 2002 at 08:41:10AM +0900, Girish Gulawani wrote:

> i'm trying to bringup linux 2.4.[2|9] on our board based on LSI mips r4k
> core.
> right now the kernel compiled with gcc-3.0, boots up & can only work with

There are known bugs in older kernels that will crash them when compiled
with gcc 3.0.  Even with current 2.4 kernels it's still playing with the
fire.

> statically linked commands. hence the root file system mounted from
> ramdisk,nfs, & de-dream ide-disk can show some-prompt only if ash.static is
> invoked.
> this is seen with 2.4.3 & ditto with 2.4.9. the kernel is in un-cached
> mode,since we have page-size problem with our core in cached, write
> through/back
> both, modes. so question is WHY THE COMMANDS WITH SHARED LIBRARY DONOT WORK.
> FAILS TO LOAD SHARED LIBRARIES.
> 
> the problem no.2 is root on ide-disk. the disk is paritioned & formatted
> using a linux pentium-pc. using a master disk the above said statically
> linked commands are downloaded to slave disk. the board boots up. however
> the bdflush/update process corrupts file-system. the UPDATE PROCESS CORRUPTS
> SUPERBLOCK AND INODES WHILE FLUSHING THE DIRTY BUFERS.
> 
> PLEASE! PLEASE!! HELP ME ON THIS. THIS NEWSGROUP IS MY LAST HOPE.

Seems pretty obvious that cacheflushing for your system is broken.
Verify that arch/mips/mm/c-r4k.c knows how to handle your system.

  Ralf

PS: Seems your caps lock key occasionally gets stuck ;-)

Re: MIPS/Linux NonSGI

[Kevin D. Kissell]

> On Fri, Jan 25, 2002 at 08:41:10AM +0900, Girish Gulawani wrote:
>
> > i'm trying to bringup linux 2.4.[2|9] on our board based on LSI mips r4k
> > core.

[snip]

> Seems pretty obvious that cacheflushing for your system is broken.

Sure sounds like it.

> Verify that arch/mips/mm/c-r4k.c knows how to handle your system.
>
>   Ralf

LSI has done a number of R3K and R4K-like designs under
their MIPS architecture license which have features
that differ from the main stream of MIPS CPUs where the OS
is concerned.  Cache manipulation is a case in point.
If it's not obvious to you from the cache management
code, compare the relevant sections of your CPU manual
with the MIPS32 spec (which I think is on the web somewhere
at www.mips.com) or a copy of the R4000 manual if you
can find one kicking around somewhere.

            Regards,

            Kevin K.

Re: MIPS/Linux NonSGI

[Kevin D. Kissell]

> On Fri, Jan 25, 2002 at 08:41:10AM +0900, Girish Gulawani wrote:
>
> > i'm trying to bringup linux 2.4.[2|9] on our board based on LSI mips r4k
> > core.

[snip]

> Seems pretty obvious that cacheflushing for your system is broken.

Sure sounds like it.

> Verify that arch/mips/mm/c-r4k.c knows how to handle your system.
>
>   Ralf

LSI has done a number of R3K and R4K-like designs under
their MIPS architecture license which have features
that differ from the main stream of MIPS CPUs where the OS
is concerned.  Cache manipulation is a case in point.
If it's not obvious to you from the cache management
code, compare the relevant sections of your CPU manual
with the MIPS32 spec (which I think is on the web somewhere
at www.mips.com) or a copy of the R4000 manual if you
can find one kicking around somewhere.

            Regards,

            Kevin K.

Re: MIPS/Linux NonSGI

[Girish Gulawani]

hi, again.
thanks a lot for quick reply. re-compiled the kernels, using egcc-2.91, but
no luck. is there still older version i should be using? & i guess you were
referring to file arch/mips/mm/r4xx.c. the board support is detected &
load_mmu_r4k() is called. so this looks okay.
yes, LSI CPUs differ from the main stream, right in the cache opcodes. but
this problem apart, as i have been using kernel in un-cached mode & anyway i
have build almost all the commands by static linking. hence shared library
issue i shall deal with later.
***right now my main concern is file system getting corrupted. the
bdflush/update processes flushes the inode buffers & corrupts the fs. to be
precise the "update" process wakes up & writes some buffers to the disk from
the function sync_old_buffers() in fs/buffer.c. the pci-ide bus mastering
controller is non standard & that the driver is also self written. however
since reads from the disk are working as i can see all the commands getting
loaded & executed, the writes also should be working properly, as they
differ only in a command to the controller.
so where could be the problem? please help me.
many thanks & best regards,
girish.

ps. ralf, liked your comment abt my caps lock getting stuck;-) it just shows
how desperate i'm in solving this problem.



_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

Re: MIPS/Linux NonSGI

[Girish Gulawani]

hi, again.
thanks a lot for quick reply. re-compiled the kernels, using egcc-2.91, but
no luck. is there still older version i should be using? & i guess you were
referring to file arch/mips/mm/r4xx.c. the board support is detected &
load_mmu_r4k() is called. so this looks okay.
yes, LSI CPUs differ from the main stream, right in the cache opcodes. but
this problem apart, as i have been using kernel in un-cached mode & anyway i
have build almost all the commands by static linking. hence shared library
issue i shall deal with later.
***right now my main concern is file system getting corrupted. the
bdflush/update processes flushes the inode buffers & corrupts the fs. to be
precise the "update" process wakes up & writes some buffers to the disk from
the function sync_old_buffers() in fs/buffer.c. the pci-ide bus mastering
controller is non standard & that the driver is also self written. however
since reads from the disk are working as i can see all the commands getting
loaded & executed, the writes also should be working properly, as they
differ only in a command to the controller.
so where could be the problem? please help me.
many thanks & best regards,
girish.

ps. ralf, liked your comment abt my caps lock getting stuck;-) it just shows
how desperate i'm in solving this problem.



_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

Re: MIPS/Linux NonSGI

[Girish Gulawani]

hello, ralf.

could solve the disk corruption problem, thanks to you. it was due to my
over-confident source code for bus mastering device. while re-targeting the
driver from vxworks to linux i had introduced a bug. its all working great
now. thanks again. while this problem is solved  i have one more query.

as i mentioned earlier, our core supports page size of 16K. hence the mmu
code is changed accordingly, for kernel 2.4.9. but any command load fails
giving page alignment error. seems some problem from binfmt_elf.c file. i
have build the command, ash, using egcc-2.91.66. to change the page
alignment the option given was "--n16384". this option is surely wrong,
hence could you tell me what is correct option? HOW TO CHANGE PAGE ALIGNMENT
TO 16K WHILE LINKING USER COMMANDS??

many thanks in advance & best regards,
girish.


----- Original Message -----
From: "Ralf Baechle" <ralf@oss.sgi.com>
To: "Girish Gulawani" <girishvg@yahoo.com>
Cc: "MIPS/Linux List (SGI)" <linux-mips@oss.sgi.com>
Sent: Friday, January 25, 2002 10:45 AM
Subject: Re: MIPS/Linux NonSGI


> On Fri, Jan 25, 2002 at 08:41:10AM +0900, Girish Gulawani wrote:
>
> > i'm trying to bringup linux 2.4.[2|9] on our board based on LSI mips r4k
> > core.
> > right now the kernel compiled with gcc-3.0, boots up & can only work
with
>
> There are known bugs in older kernels that will crash them when compiled
> with gcc 3.0.  Even with current 2.4 kernels it's still playing with the
> fire.
>
> > statically linked commands. hence the root file system mounted from
> > ramdisk,nfs, & de-dream ide-disk can show some-prompt only if ash.static
is
> > invoked.
> > this is seen with 2.4.3 & ditto with 2.4.9. the kernel is in un-cached
> > mode,since we have page-size problem with our core in cached, write
> > through/back
> > both, modes. so question is WHY THE COMMANDS WITH SHARED LIBRARY DONOT
WORK.
> > FAILS TO LOAD SHARED LIBRARIES.
> >
> > the problem no.2 is root on ide-disk. the disk is paritioned & formatted
> > using a linux pentium-pc. using a master disk the above said statically
> > linked commands are downloaded to slave disk. the board boots up.
however
> > the bdflush/update process corrupts file-system. the UPDATE PROCESS
CORRUPTS
> > SUPERBLOCK AND INODES WHILE FLUSHING THE DIRTY BUFERS.
> >
> > PLEASE! PLEASE!! HELP ME ON THIS. THIS NEWSGROUP IS MY LAST HOPE.
>
> Seems pretty obvious that cacheflushing for your system is broken.
> Verify that arch/mips/mm/c-r4k.c knows how to handle your system.
>
>   Ralf
>
> PS: Seems your caps lock key occasionally gets stuck ;-)


_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

Re: MIPS/Linux NonSGI

[Girish Gulawani]

hello, ralf.

could solve the disk corruption problem, thanks to you. it was due to my
over-confident source code for bus mastering device. while re-targeting the
driver from vxworks to linux i had introduced a bug. its all working great
now. thanks again. while this problem is solved  i have one more query.

as i mentioned earlier, our core supports page size of 16K. hence the mmu
code is changed accordingly, for kernel 2.4.9. but any command load fails
giving page alignment error. seems some problem from binfmt_elf.c file. i
have build the command, ash, using egcc-2.91.66. to change the page
alignment the option given was "--n16384". this option is surely wrong,
hence could you tell me what is correct option? HOW TO CHANGE PAGE ALIGNMENT
TO 16K WHILE LINKING USER COMMANDS??

many thanks in advance & best regards,
girish.


----- Original Message -----
From: "Ralf Baechle" <ralf@oss.sgi.com>
To: "Girish Gulawani" <girishvg@yahoo.com>
Cc: "MIPS/Linux List (SGI)" <linux-mips@oss.sgi.com>
Sent: Friday, January 25, 2002 10:45 AM
Subject: Re: MIPS/Linux NonSGI


> On Fri, Jan 25, 2002 at 08:41:10AM +0900, Girish Gulawani wrote:
>
> > i'm trying to bringup linux 2.4.[2|9] on our board based on LSI mips r4k
> > core.
> > right now the kernel compiled with gcc-3.0, boots up & can only work
with
>
> There are known bugs in older kernels that will crash them when compiled
> with gcc 3.0.  Even with current 2.4 kernels it's still playing with the
> fire.
>
> > statically linked commands. hence the root file system mounted from
> > ramdisk,nfs, & de-dream ide-disk can show some-prompt only if ash.static
is
> > invoked.
> > this is seen with 2.4.3 & ditto with 2.4.9. the kernel is in un-cached
> > mode,since we have page-size problem with our core in cached, write
> > through/back
> > both, modes. so question is WHY THE COMMANDS WITH SHARED LIBRARY DONOT
WORK.
> > FAILS TO LOAD SHARED LIBRARIES.
> >
> > the problem no.2 is root on ide-disk. the disk is paritioned & formatted
> > using a linux pentium-pc. using a master disk the above said statically
> > linked commands are downloaded to slave disk. the board boots up.
however
> > the bdflush/update process corrupts file-system. the UPDATE PROCESS
CORRUPTS
> > SUPERBLOCK AND INODES WHILE FLUSHING THE DIRTY BUFERS.
> >
> > PLEASE! PLEASE!! HELP ME ON THIS. THIS NEWSGROUP IS MY LAST HOPE.
>
> Seems pretty obvious that cacheflushing for your system is broken.
> Verify that arch/mips/mm/c-r4k.c knows how to handle your system.
>
>   Ralf
>
> PS: Seems your caps lock key occasionally gets stuck ;-)


_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

Re: MIPS/Linux NonSGI

[Ralf Baechle]

On Wed, Jan 30, 2002 at 07:39:41PM +0900, Girish Gulawani wrote:

> could solve the disk corruption problem, thanks to you. it was due to my
> over-confident source code for bus mastering device. while re-targeting the
> driver from vxworks to linux i had introduced a bug. its all working great
> now. thanks again. while this problem is solved  i have one more query.
> 
> as i mentioned earlier, our core supports page size of 16K. hence the mmu
> code is changed accordingly, for kernel 2.4.9. but any command load fails
> giving page alignment error. seems some problem from binfmt_elf.c file. i
> have build the command, ash, using egcc-2.91.66. to change the page
> alignment the option given was "--n16384". this option is surely wrong,
> hence could you tell me what is correct option? HOW TO CHANGE PAGE ALIGNMENT
> TO 16K WHILE LINKING USER COMMANDS??

I hate to say it but at this time we don't support any other pagesize
than 4kb on MIPS.  I haven't looked into details but I'd expect to hit
a number of subtle bugs when we go to a larger pagesize.  So I expect
your way to be stony.

  Ralf