Re: page size change on MIPS
On 02/03/2011 08:16 PM, naveen yadav wrote: > Hi, I am adding readelf info also . > > mips-linux-gnu-readelf -S squashfs-root/bin/busybox > There are 35 section headers, starting at offset 0x174290: > Section headers are not relevant to any of this. You need to look at the program headers. Try 'readelf -l' David Daney [...] >> >> >> On Mon, Jan 31, 2011 at 6:38 PM, Ralf Baechle wrote: >>> On Sun, Jan 30, 2011 at 08:32:43PM +0530, Himanshu Aggarwal wrote: >>> >>>> Why should the application or the toolchains depend on pagesize? I am >>>> not very clear on this. Can someone explain it? >>> >>> To allow loading directly with mmap the executable file's layout must >>> be such that it's it's segments are on offsets that are a multiple of >>> the page size so in turn the linker must know that alignment. >>> >>> Ralf >>> >> > ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: page size change on MIPS
On 01/28/2011 01:18 AM, naveen yadav wrote:
> Hi all
>
> I try changing the kernel_execve() function as suggested, but it fails,
> I am using code sourcery toolchain 4.4.1 but it still fails. I am
> using 2.6.30.9 kernel and this function is part of it.
>
>
Quit trying to change kernel_execve(). It is perfect as it is.
The problem is elsewhere.
>
> Kind regards
>
Thank you.
>
> On Thu, Jan 27, 2011 at 11:14 PM, David Daney
> wrote:
>> On 01/27/2011 07:18 AM, adnan iqbal wrote:
>>>
>>> Please try this. One line of code is added ( move%1, $7).
>>>
>>>
>>> int kernel_execve(const char *filename, char *const argv[], char *const
>>> envp[])
>>> {
>>> register unsigned long __a0 asm("$4") = (unsigned long) filename;
>>> register unsigned long __a1 asm("$5") = (unsigned long) argv;
>>> register unsigned long __a2 asm("$6") = (unsigned long) envp;
>>> register unsigned long __a3 asm("$7");
>>> unsigned long __v0;
>>> __asm__ volatile (" \n"
>>> " .setnoreorder \n"
>>> " li $2, %5 # __NR_execve \n"
>>> " syscall \n"
>>> " move%0, $2 \n"
>>> " move%1, $7\n"
>>> " .setreorder \n"
>>> : "=&r" (__v0), "=r" (__a3)
>>> : "r" (__a0), "r" (__a1), "r" (__a2), "i" (__NR_execve)
>>> : "$2", "$8", "$9", "$10", "$11", "$12", "$13", "$14", "$15",
>>> "$24",
>>> "memory");
>>>
>>>
>>> if (__a3 == 0)
>>> return __v0;
>>> return -__v0;
>>> }
>>>
>>
>> I don't know where you got that code. But really you should do what glibc
>> does. glibc gets it correct.
>>
>> At a minimum you are missing "hi" and "lo" clobbers.
Ignore that ^ advice.
>>
>> If the code works with 16K pages, and not 64K pages, then this snippet is
>> not the problem. Likely your problem is the layout of the PHDRs in the
>> executable is not compatible with the page size.
>>
Instead look at this issue.
>> David Daney
>>
>>
>>>
>>> On Thu, Jan 27, 2011 at 7:55 PM, naveen yadav
>>> wrote:
>>>
>>>> Hi David,
>>>>
>>>> thanks for your response.
>>>>
>>>> I check and found that kernel is booting with 16KB page size with
>>>> ramdisk booting. But when I change to 64KB it give me
>>>>
>>>> : applet not found
>>>> Kernel panic - not syncing: Attempted to kill init!
>>>> so I check and found that it is not able to execute well the system
>>>> call in kernel_execve function.
>>>> I am using codesourcercy toolchain(4.3.1). So is there a way to debug
>>>> this problem or how to debug below function.
>>>>
>>>> int kernel_execve(const char *filename, char *const argv[], char *const
>>>> envp[])
>>>> {
>>>> register unsigned long __a0 asm("$4") = (unsigned long) filename;
>>>> register unsigned long __a1 asm("$5") = (unsigned long) argv;
>>>> register unsigned long __a2 asm("$6") = (unsigned long) envp;
>>>> register unsigned long __a3 asm("$7");
>>>> unsigned long __v0;
>>>> __asm__ volatile (" \n"
>>>> " .setnoreorder \n"
>>>> " li $2, %5 # __NR_execve \n"
>>>> " syscall \n"
>>>> " move%0, $2 \n"
>>>> " .setreorder \n"
>>>> : "=&r" (__v0), "=r" (__a3)
>>>> : "r" (__a0
Re: page size change on MIPS
On 01/27/2011 07:18 AM, adnan iqbal wrote:
> Please try this. One line of code is added ( move%1, $7).
>
>
> int kernel_execve(const char *filename, char *const argv[], char *const
> envp[])
> {
> register unsigned long __a0 asm("$4") = (unsigned long) filename;
> register unsigned long __a1 asm("$5") = (unsigned long) argv;
> register unsigned long __a2 asm("$6") = (unsigned long) envp;
> register unsigned long __a3 asm("$7");
> unsigned long __v0;
> __asm__ volatile (" \n"
> " .setnoreorder \n"
> " li $2, %5 # __NR_execve \n"
> " syscall \n"
> " move%0, $2 \n"
> " move%1, $7\n"
> " .setreorder \n"
> : "=&r" (__v0), "=r" (__a3)
> : "r" (__a0), "r" (__a1), "r" (__a2), "i" (__NR_execve)
> : "$2", "$8", "$9", "$10", "$11", "$12", "$13", "$14", "$15", "$24",
> "memory");
>
>
> if (__a3 == 0)
> return __v0;
> return -__v0;
> }
>
I don't know where you got that code. But really you should do what
glibc does. glibc gets it correct.
At a minimum you are missing "hi" and "lo" clobbers.
If the code works with 16K pages, and not 64K pages, then this snippet
is not the problem. Likely your problem is the layout of the PHDRs in
the executable is not compatible with the page size.
David Daney
>
> On Thu, Jan 27, 2011 at 7:55 PM, naveen yadav wrote:
>
>> Hi David,
>>
>> thanks for your response.
>>
>> I check and found that kernel is booting with 16KB page size with
>> ramdisk booting. But when I change to 64KB it give me
>>
>> : applet not found
>> Kernel panic - not syncing: Attempted to kill init!
>> so I check and found that it is not able to execute well the system
>> call in kernel_execve function.
>> I am using codesourcercy toolchain(4.3.1). So is there a way to debug
>> this problem or how to debug below function.
>>
>> int kernel_execve(const char *filename, char *const argv[], char *const
>> envp[])
>> {
>> register unsigned long __a0 asm("$4") = (unsigned long) filename;
>> register unsigned long __a1 asm("$5") = (unsigned long) argv;
>> register unsigned long __a2 asm("$6") = (unsigned long) envp;
>> register unsigned long __a3 asm("$7");
>> unsigned long __v0;
>> __asm__ volatile (" \n"
>> " .setnoreorder \n"
>> " li $2, %5 # __NR_execve \n"
>> " syscall \n"
>> " move%0, $2 \n"
>> " .setreorder \n"
>> : "=&r" (__v0), "=r" (__a3)
>> : "r" (__a0), "r" (__a1), "r" (__a2), "i" (__NR_execve)
>> : "$2", "$8", "$9", "$10", "$11", "$12", "$13", "$14", "$15", "$24",
>> "memory");
>>
>>
>> if (__a3 == 0)
>> return __v0;
>> return -__v0;
>> }
>>
>>
>> On Tue, Jan 25, 2011 at 12:26 AM, David Daney
>> wrote:
>>> On 01/24/2011 07:02 AM, naveen yadav wrote:
>>>>
>>>> Hi All,
>>>>
>>>>
>>>> we are using mips32r2 so I want to know which all pages size it can
>>>> support?
>>>> When I modify arch/mips/Kconfig. it boot sucessfully on 16KB page
>>>> size. but hang/not boot crash when change page size to 8KB,32KB and 64
>>>> KB.
>>>
>>> I don't think 8KB and 32KB work on most mips32r2 processors. You would
>> have
>>> to check the processor manual to be sure.
>>>
>>>
>>>>
>&g
Re: page size change on MIPS
On 01/24/2011 07:02 AM, naveen yadav wrote: > Hi All, > > > we are using mips32r2 so I want to know which all pages size it can support? > When I modify arch/mips/Kconfig. it boot sucessfully on 16KB page > size. but hang/not boot crash when change page size to 8KB,32KB and 64 > KB. I don't think 8KB and 32KB work on most mips32r2 processors. You would have to check the processor manual to be sure. > > We are using 2.6.30 kernel. > > At Page Size 8KB and 32KB it hang in unpack_to_rootfs() function of > init/initramfs.c > > 64KB it hangs when execute init Kernel panic - not syncing: Attempted > to kill init! I regularly run 4K, 16K, and 64K page sizes with a Debian rootfs. If you run with a broken uClibc toolchain that doesn't support larger pages, it will of course fail. In this case the problem is with your toolchain, not the kernel. David Daney > > config PAGE_SIZE_4KB > bool "4kB" > help > This option select the standard 4kB Linux page size. On some > R3000-family processors this is the only available page size. Using > 4kB page size will minimize memory consumption and is therefore > recommended for low memory systems. > > config PAGE_SIZE_8KB > bool "8kB" > depends on (EXPERIMENTAL&& CPU_R8000) || CPU_CAVIUM_OCTEON > help >Using 8kB page size will result in higher performance kernel at >the price of higher memory consumption. This option is available >only on R8000 and cnMIPS processors. Note that you will need a >suitable Linux distribution to support this. > > config PAGE_SIZE_16KB > bool "16kB" > depends on !CPU_R3000&& !CPU_TX39XX > help >Using 16kB page size will result in higher performance kernel at >the price of higher memory consumption. This option is available > on >all non-R3000 family processors. Note that you will need a > suitable >Linux distribution to support this. > > config PAGE_SIZE_32KB > bool "32kB" > help >Using 32kB page size will result in higher performance kernel at >the price of higher memory consumption. This option is available >only on cnMIPS cores. Note that you will need a suitable Linux >distribution to support this. > > config PAGE_SIZE_64KB > bool "64kB" > depends on EXPERIMENTAL&& !CPU_R3000&& !CPU_TX39XX > help >Using 64kB page size will result in higher performance kernel at >the price of higher memory consumption. This option is available > on >all non-R3000 family processor. Not that at the time of this >writing this option is still high experimental. > > ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
