Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

H. Peter Anvin wrote:
> Actually, as far as I can see, he has re-invented having a real-mode 
> code chunk which then gets run before the protected-mode kernel.  We 
> already have that!

  I did not claim to have invented anything there, this is just a quite
 simple C code to execute instead of the current real mode assembly:
 it is a rewrite with obvious advantages/disadvantages.
 New features are more that this real-mode function can return an error
 to the bootloader to tell something to the user, so the user can select
 another kernel with the right processor, another video mode... with
 clean error messages - not a crash dump because this assembly
 instruction is not for that processor.
  I am still saying that the bootloader knows the root filesystem to
 be used by the kernel it loads, and that ELF is a clean format to
 store different sections to be loaded into memory at predefined
 addresses. Also there isn't any more kernel size limit.

 Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[H. Peter Anvin]

Etienne Lorrain wrote:
> Also there isn't any more kernel size limit.

What kernel size limit?

	-hpa

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Eric W. Biederman]

Etienne Lorrain <etienne_lorrain@yahoo.fr> writes:

> H. Peter Anvin wrote:
>> Actually, as far as I can see, he has re-invented having a real-mode 
>> code chunk which then gets run before the protected-mode kernel.  We 
>> already have that!
>
>   I did not claim to have invented anything there, this is just a quite
>  simple C code to execute instead of the current real mode assembly:
>  it is a rewrite with obvious advantages/disadvantages.
>  New features are more that this real-mode function can return an error
>  to the bootloader to tell something to the user, so the user can select
>  another kernel with the right processor, another video mode... with
>  clean error messages - not a crash dump because this assembly
>  instruction is not for that processor.

Having an error handling compatibility that is backwards compatible sounds
interesting.

>   I am still saying that the bootloader knows the root filesystem to
>  be used by the kernel it loads, and that ELF is a clean format to
>  store different sections to be loaded into memory at predefined
>  addresses.

Yes.  Although when you think sections instead of segments I'm a little
worried.

>  Also there isn't any more kernel size limit.

I think as HPA points out we have gotten past that a long time
ago with the bzImage format.

With the right delicacy, and preserving backwards compatibility
with existing bootloaders I think we can achieve things.

The big issue is that sometimes bootloaders are a little bit brittle.

Eric

Re: RE : Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Eric W. Biederman]

Etienne Lorrain <etienne_lorrain@yahoo.fr> writes:

> Eric W. Biederman wrote:
>> Etienne Lorrain writes:
>> >   Well, a self relocating image cannot be an ELF file because the code
>> >  to relocate the ELF cannot be executed at the wrong place.
>> >  If relocation is needed, I would better like not to link vmlinux at a
>> >  fixed address first. In fact I wonder if we are talking of the same
>> >  kind of relocation: you seem to talk about "ld --pic-executable" while
>> >  I am thinking of "ld -r" to "locate" it at the bootloader loading time.
>> >   The main problem I see is that I do not have the code for that, and
>> >  I am going deeper/earlier into the generation of vmlinux, while comments
>> >  are "already you are too early, loading an ELF file is too complex for
>> >  a bootloader". The solution I have already is working.
>> 
>> Being very clear.  ld --pic-executable or ld -shared is essentially
>> what we are talking when we are discussing building a relocatable
>> kernel.  Something with the properties of an ELF ET_DYN executable
>> that does not use an interpreter.  ld.so is the only common executable
>> of this type in linux.
>
>  So we are talking of two different things.
>  You want to _compile_ the totality of the kernel as code and data
> relocatable - right now only modules are compiled that way.
>  On IA32 architecture, there is no problem having the code relocatable,
> because most/all calls and jumps are PC relative, so very few instructions
> need to be different.
>  The problem is having the data relocatable, then every access to memory
> has to be relative to a register, and so the register EBX is reserved
> for that by the compiler.
>  The IA32 processor does not have enough registers already, removing one
> of them makes the assembly seriously worse - mostly when dealing with
> 64 bits "long long" - you can only get two of them in register and a
> simple addition produces long and slow code.
>  Having the data relocatable could be done by using the base address of
> IA32 segments, but this does not seem to be the preferred method because
> segments use seems to slow down the processor a lot more.

There are two more options in addition to segments:
- Running at a fixed virtual address.
- Processing relocation entries to modify the initial code/data to
  live where we want.  We don't have the that shared libraries do of wanting
  to share text/data pages loaded at different addresses.

>  What I was talking of was "loading" a non relocatable kernel (compiled
> as usual without -fPIC/-shared) at any address, so not inducing slow
> down whatever the base address used. I said that it is possible, maybe
> quite easy to do so, but I do not have the code for it - and people
> say right now that an ELF bootloader is not what they want to.

The thing to pick up on is that we really don't want the gcc flag
-fPIC but if ld wasn't an idiot we would want the linker flag -shared.
Anyway we have working code merged that processes all of the
relocation entries and modifies the kernel to run properly where we
load it at.  This has all been done and merged at this point.  It's
only available to a bzImage loader but it is all there.  Currently the
only overhead is something like 5 or 10 percent overhead to hold the
relocation entries.

The only thing currently compiled with -fPIC is the kernel gzip
decompressor and it is carefully coded so it doesn't generate any
relocation entries that need processing.  It is just throw away code
so we don't much care about it's efficiency.

>  Now, if people wants to switch to a completely relocatable kernel,
> because they do not mind the slow down (that I am not really able
> to quantify) - and because anyway most of their kernel is modules,
> then Gujin can relocate that without _any_ problem; in fact it
> does already load in intermediate (HIMEM malloc'ed) memory under
> DOS to relocate it later (called 'late relocation' in Gujin code,
> done every times the code is not at the right address in memory,
> interrupt disabled and in protected mode).

Yes.  That is roughly what we are saying.  We want that at least
as a well defined option.

>  Loading and executing at whatever address is not a problem,
> and should be doable already in the real-mode function of the
> kernel called by Gujin, by modifying few elements of the
> structure passed in parameter of this real-mode function:
> it should be done after loading the E820 memory map for the
> BIOS parameter page.
> I am reading right now Gujin source and note that you may want
> to access array elf_reloc[] and variable nb_elf_reloc in this
> Linux function - will be added in next Gujin version.
>  You already can access type LOADER_str parameter "loader"
> where you can get runadr which can overwrite the ELF entry
> point address. The old ELF header is at loader.load_address
> if it is needed.
>
>  The main problem Gujin has, is to decide where to load a
> relocatable kernel, so this policy is left to the kernel and
> not to the user. For instance and as you made me think of, if you
> do not want for security reasons that the kernel memory could
> be modified by an attacker by using the legacy DMA, then you
> want to load it at an address which cannot be modified this way,
> i.e. over 16 Mbytes. Note that some chipset had an extension I/O
> address to extend this legacy DMA to 32 bits addresses, but it
> has not been documented and probably not implemented widely.
>
>  I do not know how many people are like me, but I am currently
> typing on a monolitic kernel because I find it a _lot_ simpler
> to do a ".config" once and make it slowly evolve with new kernel
> versions (answering No to most of the questions), than managing
> hundreds of module files each time I regenerate/remove a kernel.
>  The kernel I am working on is not relocatable because - well
> I do not have crash to debug...

Yes.   In this sense I'm a tool builder.  I build general capabilities
that can be used to solve specific problems.  A relocatable kernel
is a general capability and the only interesting problem to date
has been taking crash dumps.  To reduce the maintenance overhead I
try and remove the cases where you need an extra config option for
the new functionality.  If there is not overhead a config option just
makes things harder to test.

>  So I think that from the Gujin point of view, everything you need
> is done and working (but the access to elf_reloc)- please explain what
> you want more if not. If you want a solution to load now, you can
> use patch linux-2.6.20-rc5 in sourceforge, which uses an intermediate
> binary dump of the ELF file, and can be relocated the non ELF way.

This requires a more detailed examination.  The previous conversation
indicated that if we could replace setup.S, video.S and edd.S with
C versions it would be very interesting.  If we happen to change the
link so they are available in an ELF executable so much the better.

If we wind up adding two versions of the real mode code one written
in assembly and one written in C accessible by different means
by different bootloaders that is not interesting.

>  So, why do I want an ELF file loaded by a bootloader?

>  The main reason is to be able to use all the usual ELF tools to do
> all sort of things, like showing segments size and base address with
> "objdump" and "readelf", adding and extracting sections with "strip"
> and "objcopy", and dumping section content with either "objdump -m i8086"
> or "objdump" alone to display the assembly instructions after link,
> for real-mode or protected-mode, with symbols.
>  For all that, I need a 100% compatible ELF file structure - that is
> what I have already.

Yes.  That is nice. 

>  I know that to load this ELF kernel file the user will need a different
> loader like Gujin or your kexec. I know that I am asking you not to treat
> an impossible action (loading a section at address 0 over the interrupt
> table), and I say please.

It is my belief that we can solve this without hacks, and still have
a useable structure.

[snip]
>  So technically - and if I understand you correctly (feel free to correct
>  me) - you ask me to put the real mode code into the NOTE section, by
>  adding a few words header and concatening the code.
>
>  There is two way to do so, the first being simply to replace in
> vmlinux.lds.S (in the patch I sent) the lines:
>   .realmode 0 : AT (ADDR(.bss) + SIZEOF(.bss) - LOAD_OFFSET) {
>   ..... *(.init.text16) ...
>   } :text16 = 0x9090
> by something like:
>   .realmode 0 : AT (ADDR(.bss) + SIZEOF(.bss) - LOAD_OFFSET) {
>   LONG(8)
>   LONG(SIZEOF (.realmode))
>   LONG(1)
>   BYTE('R') BYTE('E') BYTE('A') BYTE('L')
>   BYTE('M') BYTE('O') BYTE('D') BYTE('E')
>   ..... *(.init.text16) ...
>   } :note = 0x9090

That would certainly get the job done.

>  The other way to do is to produce an independant link for the real-mode
> program (linked at base address 0), convert it to binary and include
> the resulting block into the NOTE segment. First that is a lot more
> complex (it can be done but is more complex), and two I loose all
> the debuging information I wanted in the ELF file.
>  I can still see the real-mode assembly by forcing objdump to interpret
> the NOTE section as real-mode code, but the offset calculated for
> assembly calls and jumps is wrong, I lost all symbols, in short it
> goes against what I wanted initially.

So I really don't care if we move whole real mode section into a note
or if we just put a pointer to it into a note.  What is important is
that we use a note to find it.

Which means that we could do something goofy in the linker script
like we do with the current vdso.  So we could give it a virtual
address of 0 and a physical address in the init code section.
Allowing us to load it normally and then just drop it later.  Not
perfect but something that would not need special treatment by
a bootloader unless it wanted to copy the data from there into
some location in low memory and execute it.

For me the objective is not so much reusing the existing tools
(although that is a plus) but more to be able to build a unified
binary that can be used for everything, and will give us the freedom
to do interesting things with the kernel in the future, and hopefully
something that is more or less usable by portable bootloaders.  Having
a different file format and different rules for different
architectures is a pain.

Eric

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

--- "Eric W. Biederman" wrote:
> So I really don't care if we move whole real mode section into a note
> or if we just put a pointer to it into a note.  What is important is
> that we use a note to find it.

 Well, we can advertise by a note the section number or the section
name which contains the real-mode code, but finding the section of
type SHT_PROGBITS having SHF_EXECINSTR flags linked at zero is not
that difficult to do: it is what Gujin does right now when it does
not find the program header linked at zero.

> Which means that we could do something goofy in the linker script
> like we do with the current vdso.  So we could give it a virtual
> address of 0 and a physical address in the init code section.

 Gujin loads at the physical address, i.e. kernel is loaded at
0x100000 and not 0xC0100000, is that wrong?
 I am not sure playing these games with addresses is cleaner than
not loading a section which is not in the program header.

> For me the objective is not so much reusing the existing tools
> (although that is a plus) but more to be able to build a unified
> binary that can be used for everything, and will give us the freedom
> to do interesting things with the kernel in the future, and hopefully
> something that is more or less usable by portable bootloaders.  Having
> a different file format and different rules for different
> architectures is a pain.

 For me the objective is to have a reasonable bootloader,
I will not have the time to port back and test with every other
bootloader some transfert of code from Gujin to an ELF looking
like bzImage file.

  Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Eric W. Biederman]

Etienne Lorrain <etienne_lorrain@yahoo.fr> writes:

> --- "Eric W. Biederman" wrote:
>> So I really don't care if we move whole real mode section into a note
>> or if we just put a pointer to it into a note.  What is important is
>> that we use a note to find it.
>
>  Well, we can advertise by a note the section number or the section
> name which contains the real-mode code, but finding the section of
> type SHT_PROGBITS having SHF_EXECINSTR flags linked at zero is not
> that difficult to do: it is what Gujin does right now when it does
> not find the program header linked at zero.

What I was thinking is that in the not we place the physical address
and length that we load the real mode code at.  My assumption being
that we have marked the real mode code __init or the equivalent,
so we always load and just ignore it later on.

>> Which means that we could do something goofy in the linker script
>> like we do with the current vdso.  So we could give it a virtual
>> address of 0 and a physical address in the init code section.
>
>  Gujin loads at the physical address, i.e. kernel is loaded at
> 0x100000 and not 0xC0100000, is that wrong?

No.  For practical purposes we can say virtual addresses are for
the code and the kernel.  Physical addresses are for the loader.

>  I am not sure playing these games with addresses is cleaner than
> not loading a section which is not in the program header.

It is some cleaner.  I won't call it perfectly clean, but it is
better.  It keeps everything but argument generation generic,
and there is already precedent in things like the vdso.

Playing the games with the addresses does allow the existing debugging
tools to work without problem because the end users of the code all
examine the virtual not the physical address.

>> For me the objective is not so much reusing the existing tools
>> (although that is a plus) but more to be able to build a unified
>> binary that can be used for everything, and will give us the freedom
>> to do interesting things with the kernel in the future, and hopefully
>> something that is more or less usable by portable bootloaders.  Having
>> a different file format and different rules for different
>> architectures is a pain.
>
>  For me the objective is to have a reasonable bootloader,
> I will not have the time to port back and test with every other
> bootloader some transfert of code from Gujin to an ELF looking
> like bzImage file.

I agree I want a reasonable bootloader as well.

With ELF we get a single file format that works on multiple
architectures and for multiple OS-s, with well understood rules.  This
allows for a broader audience who can review and maintain the code.
In addition to the real possibility of multi-architecture or multi-os
bootloaders.

What ELF cannot standardize is the how we pass the initial arguments.
ELF does standardize the initial parameters for unix executables but
the requirements of arguments to kernels, and the years of practice
are quite different.

ELF should make it much easier for people implementing simple
stand-alone executables for testing.  As you don't even need a linker
script or any fancy games if you don't take arguments.   Similarly
the switch from linux to multi-boot that almost standard is just
supporting a different argument passing format.

Eric

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

--- "Eric W. Biederman" <ebiederm@xmission.com> wrote:
> What I was thinking is that in the not we place the physical address
> and length that we load the real mode code at.  My assumption being
> that we have marked the real mode code __init or the equivalent,
> so we always load and just ignore it later on.

 But Gujin need the real-mode part linked at zero and there isn't
anything linked at this address in the current Linux kernel, no
section at all. If you ask to do a separate link then I loose all
the symbols and the ELF standard tools are a lot less useable.

> Playing the games with the addresses does allow the existing debugging
> tools to work without problem because the end users of the code all
> examine the virtual not the physical address.

 The linker also resolve relocation in virtual addresses so this part
 needs virtual address = 0.

> I agree I want a reasonable bootloader as well.
>  [snip]
> ELF should make it much easier for people implementing simple
> stand-alone executables for testing.  As you don't even need a linker
> script or any fancy games if you don't take arguments.   Similarly
> the switch from linux to multi-boot that almost standard is just
> supporting a different argument passing format.

 There is plenty of possible future, but right now I have a simple
(OK, not perfect, but code is shown) solution which works.
I can modify bits, but there is no point complexifying the system
for possible theoretical problems - by experience you always miss
important future problem while overdesigning for some problem which
never appear.
If someone can show me a real problem - then I'd like to hear from him,
but ELF compatibility will not be up to the point where the user will
be able to run the kernel from a Xterm command line.

  Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[H. Peter Anvin]

Eric W. Biederman wrote:
> 
> With ELF we get a single file format that works on multiple
> architectures and for multiple OS-s, with well understood rules.  This
> allows for a broader audience who can review and maintain the code.
> In addition to the real possibility of multi-architecture or multi-os
> bootloaders.
> 

Well understood MAYBE (although the "virtual" versus "physical" address 
stuff in ELF is clearly misspecified in the extreme), but they are 
clearly not sufficient for presenting a well-defined entry condition. 
So it's not really anywhere near as straightforward as you make it sound.

	-hpa

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Eric W. Biederman]

"H. Peter Anvin" <hpa@zytor.com> writes:

> Eric W. Biederman wrote:
>>
>> With ELF we get a single file format that works on multiple
>> architectures and for multiple OS-s, with well understood rules.  This
>> allows for a broader audience who can review and maintain the code.
>> In addition to the real possibility of multi-architecture or multi-os
>> bootloaders.
>>
>
> Well understood MAYBE (although the "virtual" versus "physical" address stuff in
> ELF is clearly misspecified in the extreme), but they are clearly not sufficient
> for presenting a well-defined entry condition. So it's not really anywhere near
> as straightforward as you make it sound.

About the entry condition I completely agree.

The fact that the generic tools almost do what you need and then fall short is
also problematic.

However it still appears a reasonable approach to consider.

Eric

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

> >  I assume you care about this ELF header because you are also a user of
> > the ELF file vmlinux, aren't you?
>
> Yes I am. I use kexec boot loader which has capability to load ELF kernel
> images (vmlinux). That's why I am concerned about linking real mode code
> in vmlinux as for kdump case I shall have to be aware that kernel vmlinux
> might contain a special PT_LOAD type program header which will contain
> real mode code and it does not have to be loaded. Then I will run into
> guessing business which one is that real mode PT_LOAD  program header and
> my assumption might very well break in next few kernel release.

 So is kexec able to relocate this vmlinux? If kexec and Gujin do approximately
the same thing they should do it the same way. If you cannot get a PT_LOAD
section, maybe we can put a simple system in NOTE, or just create a
PT_LOAD16 if the linker accepts other values.
 I do not really like to relocate after vmlinux has been linked at a fixed address,
because I am not sure you can guess each address to relocate or not:
you can define permanent address in the linker file by simply "symbol = address"
and those should never be relocated. For instance the very high addresses
on ia32 may point to registers or FPU instead of memory, so may or may not
have to be relocated. I also would better like you not to relocate the real-mode
addresses.
 
> >  And you do not want to write protect the kernel code (if the CPU write protection
> > is not working, the hardware is not working so debug will be difficult, and a simple
> > CRC32 can tell kernel memory failure) and use twice the same code memory
> > (with different data area or saving kernel data elsewhere before reload).
> >  Is that related to module loading or instruction set detection/patch or multiprocessor?
>
> If I understand it right, you seem to be suggesting that I don't have to 
> reload the kernel text and I can only reload the data for second kernel?
> 
> We run the whole of the kernel from a mutually execlusive location from
> first kernel to mitigate the concerns that first kernel's ongoing DMA might
> corrupt second kernel. That's why first kernel's text can't be reused.

  So in some exception handler, you detect something is wrong and then
 jump to a new kernel. Maybe this class of bug (DMA or hardware bit flipped)
 should be detected even without this double kernel environment, by running a
 CRC32 on the kernel text section in this exception handler, displayed that in
 the crash dump.

> Secondly, it gives flexibility to user that either he can choose to use 
> the production kernel as capture kernel or an entirely different custom
> kernel can be used as capture kernel.

  IHMO if it is just capturing the memory, I would have a kernel without
 any modules (usb-storage drivers linked-in) and save to USB key the
 core. Automatically loading modules after the crash may be a problem.

> If real mode code is linked with vmlinux, then kdump will be broken.

 I do not want to break anything for fun. I need either a reason or a bug.

> bzImage is relocatable. If a new kernel image format is introduced
> (compressed ELF), then I will prefer it to be a relocatable one
> (if possible).

 It may be possible to do some kind of "ld -r -o vmlinux", but delaying
the decision where to run the kernel is just delaying: someone has to
decide, and even if Gujin do not decide It will ask the user who most
of the time will have no clue... and on which basis do you want the
bootloader to decide... maybe this SDRAM is faulty and usually the
SDRAM are 512 Mbytes so run the kernel at 512 + 16 Mbytes, but
only if there is more than 1024 Mbytes of SDRAM?

  Thanks,
  Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Vivek Goyal]

On Thu, Feb 08, 2007 at 01:10:58PM +0000, Etienne Lorrain wrote:
> > >  I assume you care about this ELF header because you are also a user of
> > > the ELF file vmlinux, aren't you?
> >
> > Yes I am. I use kexec boot loader which has capability to load ELF kernel
> > images (vmlinux). That's why I am concerned about linking real mode code
> > in vmlinux as for kdump case I shall have to be aware that kernel vmlinux
> > might contain a special PT_LOAD type program header which will contain
> > real mode code and it does not have to be loaded. Then I will run into
> > guessing business which one is that real mode PT_LOAD  program header and
> > my assumption might very well break in next few kernel release.
> 
>  So is kexec able to relocate this vmlinux? If kexec and Gujin do approximately
> the same thing they should do it the same way.

No, as of today, kexec does not relocate vmlinux. At compilation time,
vmlinux is compiled for a fixed address and vmlinux is loaded at that
address. This compile address can be controlled with CONFIG_PHYSICAL_START,
as you already mentioned in your patchset.

So in the case of kdump, if one wishes to use vmlinux instead of a relocatable
bzImage, he will re-compile the kernel for a different address and then use
that vmlinux.

OTOH, boot loader can possibly relocate vmlinux because all the relocation
information is retained in vmlinux (Using ld option --emit-relocs). But
probably self relocating image should be a better option as all the
boot-loaders out there don't have to be modifed.

> If you cannot get a PT_LOAD
> section, maybe we can put a simple system in NOTE, or just create a
> PT_LOAD16 if the linker accepts other values.

My guess is that PT_LOAD16 is not an acceptable value. Putting information
in PT_NOTE seems interesting (As Eric already mentioned).

>  I do not really like to relocate after vmlinux has been linked at a fixed address,
> because I am not sure you can guess each address to relocate or not:
> you can define permanent address in the linker file by simply "symbol = address"
> and those should never be relocated. For instance the very high addresses
> on ia32 may point to registers or FPU instead of memory, so may or may not
> have to be relocated. I also would better like you not to relocate the real-mode
> addresses.
>  

I think there will have to be an upper limit on the address where you can
relocate your image to. Anyway, given i386 virtual memory address space
constraint, I think you can not load kernel image in high memory (896MB).
It has to run anywhere below that.

I think you are referring to defining absolute symbols. Already there are
some absolute symbols generated when kernel is compiled and these are
not relocated while image is being loaded at a non-compiled address. So if
a boot loader decides to relocate a vmlinux image, it will have to skip
relocations present w.r.t absolute symbols. (Unfortunately, ld does generate
relocation entries even for absolute symbols).
 
> > >  And you do not want to write protect the kernel code (if the CPU write protection
> > > is not working, the hardware is not working so debug will be difficult, and a simple
> > > CRC32 can tell kernel memory failure) and use twice the same code memory
> > > (with different data area or saving kernel data elsewhere before reload).
> > >  Is that related to module loading or instruction set detection/patch or multiprocessor?
> >
> > If I understand it right, you seem to be suggesting that I don't have to 
> > reload the kernel text and I can only reload the data for second kernel?
> > 
> > We run the whole of the kernel from a mutually execlusive location from
> > first kernel to mitigate the concerns that first kernel's ongoing DMA might
> > corrupt second kernel. That's why first kernel's text can't be reused.
> 
>   So in some exception handler, you detect something is wrong and then
>  jump to a new kernel.

In a nutshell, Yes.

>  Maybe this class of bug (DMA or hardware bit flipped)
>  should be detected even without this double kernel environment, by running a
>  CRC32 on the kernel text section in this exception handler, displayed that in
>  the crash dump.

We already do checksum verification to make sure newly loaded kernel has
not been corrupted before we jump to it. What's the point in doing the
verification existing kernel text. Because even that is intact, and you
reload your data section, that data section will have to be reloaded at
the same place where previous data section was. Now after you re-loaded 
the data section, some DMA might corrupt it now (Because we never stopped
DMAs). So loading a fresh copy of text and data section at a reserved 
memory location seems safer. 

> 
> > Secondly, it gives flexibility to user that either he can choose to use 
> > the production kernel as capture kernel or an entirely different custom
> > kernel can be used as capture kernel.
> 
>   IHMO if it is just capturing the memory, I would have a kernel without
>  any modules (usb-storage drivers linked-in) and save to USB key the
>  core. Automatically loading modules after the crash may be a problem.

As a user you can very well do it. Distros are also doing something 
similar. They just insert 1-2 modules from initrd (custom initrd created
based on choice of dump saving destination) and capture the dump either
to a local storage or over network and then system is booted back to
production kernel.

> 
> > If real mode code is linked with vmlinux, then kdump will be broken.
> 
>  I do not want to break anything for fun. I need either a reason or a bug.
> 
> > bzImage is relocatable. If a new kernel image format is introduced
> > (compressed ELF), then I will prefer it to be a relocatable one
> > (if possible).
> 
>  It may be possible to do some kind of "ld -r -o vmlinux", but delaying
> the decision where to run the kernel is just delaying: someone has to
> decide, and even if Gujin do not decide It will ask the user who most
> of the time will have no clue... and on which basis do you want the
> bootloader to decide... maybe this SDRAM is faulty and usually the
> SDRAM are 512 Mbytes so run the kernel at 512 + 16 Mbytes, but
> only if there is more than 1024 Mbytes of SDRAM?
> 

Interesting question, How does a boot loader/user decide where to load
the relocatable image? I think it depends on the new interesting usages
of the relocatable kernel. As of today, kexec knows where is reserved
memory region (Read from /proc/iomem) and it loads the image at the
start of that reserved region (Meeting alignment restrictions, if any). So
in this case boot loader takes the decision. May be a user option also
can be created, something like --load-address=0xXYZ and then people
can have fun loading same image at various addresses.

I think this is interesting usage if boot loader can somehow decide
that first few MB is faulty and load kernel at a memory address higher
than that. If first whole 1G is faulty then you can't do much on i386,
but on 64bit arches like x86_64, this should not be an issue. 

Thanks
Vivek

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

--- Vivek Goyal <vgoyal@in.ibm.com> wrote:
> No, as of today, kexec does not relocate vmlinux. At compilation time,
> vmlinux is compiled for a fixed address and vmlinux is loaded at that
> address. This compile address can be controlled with CONFIG_PHYSICAL_START,
> as you already mentioned in your patchset.
> 
> So in the case of kdump, if one wishes to use vmlinux instead of a relocatable
> bzImage, he will re-compile the kernel for a different address and then use
> that vmlinux.
> 
> OTOH, boot loader can possibly relocate vmlinux because all the relocation
> information is retained in vmlinux (Using ld option --emit-relocs). But
> probably self relocating image should be a better option as all the
> boot-loaders out there don't have to be modifed.

  Well, a self relocating image cannot be an ELF file because the code
 to relocate the ELF cannot be executed at the wrong place.
 If relocation is needed, I would better like not to link vmlinux at a
 fixed address first. In fact I wonder if we are talking of the same
 kind of relocation: you seem to talk about "ld --pic-executable" while
 I am thinking of "ld -r" to "locate" it at the bootloader loading time.
  The main problem I see is that I do not have the code for that, and
 I am going deeper/earlier into the generation of vmlinux, while comments
 are "already you are too early, loading an ELF file is too complex for
 a bootloader". The solution I have already is working.

> > If you cannot get a PT_LOAD
> > section, maybe we can put a simple system in NOTE, or just create a
> > PT_LOAD16 if the linker accepts other values.
> 
> My guess is that PT_LOAD16 is not an acceptable value. Putting information
> in PT_NOTE seems interesting (As Eric already mentioned).

 In fact, thinking more about that, I am going back to my implementation
 of it, because on ia32 the interrupt vectors are at address zero and it is
 obviouly an invalid address to load an ELF for this architecture.
 But for the linker, it is the right address to link it (being an offset
 into a non-null segment in real mode), and because the entry point has
 to be zero (I cannot use the ELF entry value) the program header base
 address has to be zero.
 Anyway, your loader in (probably) written in C, so a test against zero
 is a simple thing to do, and should be done anyway to check for an
 incorrect ELF program header. I wonder if this NOTE program header is
 not simply designed as an "end" marker, it does not seem to contain
 anything, so me defining the realmode after that program header may
 be a good idea.

> We already do checksum verification to make sure newly loaded kernel has
> not been corrupted before we jump to it. What's the point in doing the
> verification existing kernel text. Because even that is intact, and you
> reload your data section, that data section will have to be reloaded at
> the same place where previous data section was. Now after you re-loaded 
> the data section, some DMA might corrupt it now (Because we never stopped
> DMAs). So loading a fresh copy of text and data section at a reserved 
> memory location seems safer.

 If you really are tring to catch an erroneous DMA into the kernel,
 is it better to keep an exact copy of the kernel you are using somewhere
 else to do a bit-to-bit comparisson after the crash, and so no relocate.
 Anyway if the DMA crash has crashed the exception handling area the system
 is dead anyway.

> Interesting question, How does a boot loader/user decide where to load
> the relocatable image? I think it depends on the new interesting usages
> of the relocatable kernel. As of today, kexec knows where is reserved
> memory region (Read from /proc/iomem) and it loads the image at the
> start of that reserved region (Meeting alignment restrictions, if any). So
> in this case boot loader takes the decision. May be a user option also
> can be created, something like --load-address=0xXYZ and then people
> can have fun loading same image at various addresses.

 I think that you are asking too much for the bootloader user, and that
is a decision he has to take *before* the crash; even me, I would select
one address like 16 Mbytes and stick with it...
 If the running Linux kernel do not erase Gujin from memory, it could
also go back to real mode and do a "longjmp()" to return to the
Gujin interface - but most of the times the system had a reason to
crash (for instance a ventilator stopped working) and you can plan
whatever you want in software...

  Thanks,
  Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Eric W. Biederman]

Etienne Lorrain <etienne_lorrain@yahoo.fr> writes:

>   Well, a self relocating image cannot be an ELF file because the code
>  to relocate the ELF cannot be executed at the wrong place.
>  If relocation is needed, I would better like not to link vmlinux at a
>  fixed address first. In fact I wonder if we are talking of the same
>  kind of relocation: you seem to talk about "ld --pic-executable" while
>  I am thinking of "ld -r" to "locate" it at the bootloader loading time.
>   The main problem I see is that I do not have the code for that, and
>  I am going deeper/earlier into the generation of vmlinux, while comments
>  are "already you are too early, loading an ELF file is too complex for
>  a bootloader". The solution I have already is working.

Being very clear.  ld --pic-executable or ld -shared is essentially
what we are talking when we are discussing building a relocatable
kernel.  Something with the properties of an ELF ET_DYN executable
that does not use an interpreter.  ld.so is the only common executable
of this type in linux.

Loading an ELF executable is very much:
- Walk through the program headers and for each PT_LOAD segment load
  it at the address it requests.
- Jump to e_entry from the ELF header.

If you are working with a relocatable ELF object the rules become:
- Walk through the program header once to find the size and alignment
  of the chunk of memory that the linux kernel needs.
- Find a hole in the memory map that meets those requirements.
- Compute the offset 
- Walk through the program headers and for each PT_LOAD segment
  add offset to the addresses and load the segment like normal.
- Jump to offset + e_entry.

This is within the scope of what a bootloader can reasonably do, and
I have implemented it in etherboot as well as /sbin/kexec. 

>> > If you cannot get a PT_LOAD
>> > section, maybe we can put a simple system in NOTE, or just create a
>> > PT_LOAD16 if the linker accepts other values.
>> 
>> My guess is that PT_LOAD16 is not an acceptable value. Putting information
>> in PT_NOTE seems interesting (As Eric already mentioned).
>
>  In fact, thinking more about that, I am going back to my implementation
>  of it, because on ia32 the interrupt vectors are at address zero and it is
>  obviouly an invalid address to load an ELF for this architecture.

No special games no special rules with the well defined ELF components
either add a note that you can define all of the semantics yourself
or don't do it.  That is what the notes are there for.

>  But for the linker, it is the right address to link it (being an offset
>  into a non-null segment in real mode), and because the entry point has
>  to be zero (I cannot use the ELF entry value) the program header base
>  address has to be zero.

Agreed. When the object file is linked using offset 0, and letting the
real mode segments do have different bases to do your relocation is fine.

>  Anyway, your loader in (probably) written in C, so a test against zero
>  is a simple thing to do, and should be done anyway to check for an
>  incorrect ELF program header. I wonder if this NOTE program header is
>  not simply designed as an "end" marker, it does not seem to contain
>  anything, so me defining the realmode after that program header may
>  be a good idea.

We have been very sparse on the usage of ELF notes but yes they exist
and yes people do look at them.  Please dig up a copy of the ELF spec
and read up on them or look at etherboot for an example.

>  If you really are tring to catch an erroneous DMA into the kernel,
>  is it better to keep an exact copy of the kernel you are using somewhere
>  else to do a bit-to-bit comparisson after the crash, and so no relocate.
>  Anyway if the DMA crash has crashed the exception handling area the system
>  is dead anyway.

No.  We are not trying to catch a erroneous DMA in that sense.  We
do not shutdown any drivers when switching to the new kernel from
panic(), because we don't know what is broken.   Any single bit
of kernel code of could cause problems not just the exception table.
Running in an area that we have never used for DMA and is completely
reserved gives us freedom to not worry about it.  I.e. This is not
error detection but future error prevention and it works.

Before starting the new kernel we do a sha256 checksum test on the
new kernel and on our code that is running the checksum.  All of which
comes from /sbin/kexec.  Not compiled into the running kernel.  The
policy is in user space.

>> Interesting question, How does a boot loader/user decide where to load
>> the relocatable image? I think it depends on the new interesting usages
>> of the relocatable kernel. As of today, kexec knows where is reserved
>> memory region (Read from /proc/iomem) and it loads the image at the
>> start of that reserved region (Meeting alignment restrictions, if any). So
>> in this case boot loader takes the decision. May be a user option also
>> can be created, something like --load-address=0xXYZ and then people
>> can have fun loading same image at various addresses.
>
>  I think that you are asking too much for the bootloader user, and that
> is a decision he has to take *before* the crash; even me, I would select
> one address like 16 Mbytes and stick with it.

Yes.  Unfortunately there is no one value that works on all machines,
which is why we are moving to a relocatable kernel.

Currently we specify crashkernel=size@location on the kernel command
line to reserve the memory.  Hopefully we can reduce this to just
crashkernel=size and have the kernel find a reasonable whole in
the memory map to reserve.  Where that hole is, is exported to
userspace so /sbin/kexec just puts the kernel in that hole.

>  If the running Linux kernel do not erase Gujin from memory, it could
> also go back to real mode and do a "longjmp()" to return to the
> Gujin interface - but most of the times the system had a reason to
> crash (for instance a ventilator stopped working) and you can plan
> whatever you want in software...

True, and no solution is perfect.   The target is to catch the maximum
number of situations that can be caught.

Going back to real mode generally doesn't work because the BIOS get's
confused with the changes in hardware state from linux running.  Some
systems it does work on though.  Going back to real mode especially
don't work well when the kernel doesn't do a clean shutdown.

In a normal context there are two practical advantages to a bootloader
speaking ELF.
1) The load address is no longer fixed at 1MB.  So if (for example) we
   want to get the performance advantages of 4MB pages all we have to
   do is tweak the  alignment and the load address to be at 4MB.
2) Being able to choose somewhere else in the memory map that works.
   If we have a big (32MB uncompressed) kernel with all of the modules
   compiled in and there is a memory hole at 15MB-16MB.  The normal
   load address won't work so the bootloader can pick another address.
   Similarly problems appear when people place acpi tables at lower
   addresses.

For a 64bit kernel I have thought placing the kernel above 4GB has
several interesting advantages for making more memory available for
DMA accesses without needed an IOMMU.

Then we get into the cases like Xen, which have no real mode to go
through so need completely different bootloaders.

Eric

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

Vivek Goyal wrote:
> > Etienne Lorrain wrote:
> >  Yes, any PT_LOAD below 64 Kbytes can only be real mode, and real-mode
> > cannot be loaded higher, and cannot be bigger than 640 Kbytes, anything
> > different (like with virtual address at 0xC0000000) is Linux protected mode.
> > Considering the linker used it is always the 4th program header, before there
> > were only 3 program header,third one stay the NOTE one.
>
> Well this is all a lot of speculation. This is not standard way of retrieving
> information from ELF. Number of program headers finally created keep on
> changing. Previously it was left to the linker and now people have changed
> it to 3 by specifically using PHDR directive.

  Well, the standard way is:
1 - the code program header, to be write protected if possible
2 - the (initialised) data program header, to be able to restart the program by just reload/re-entry,
    and after that (memsiz - filesz) of BSS
 Anything after is non standard. The number of program header is written in the ELF header,
but anything over 2 probably need a special loader for the extra treatment.
 That isn't usual to have something near zero, because it is better not to map
any memory there to catch null pointer dereference (in virtual memory system).

 I assume you care about this ELF header because you are also a user of
the ELF file vmlinux, aren't you?
  
> >  Just I did not completely understand why you need relocation (and the announce
> > was when I was in holidays far away). I know a simple kernel is needed to do
> > some debugging when the main kernel has crashed, this kernel is better loaded
> > at for instance 16 Mbytes. You probably do not want the same kernel as the main
> > one because it may crash the same way, and you start a loop - and even then
> > if you have exactly the same kernel it is easier to use the same write-protected
> > block of memory with different data sections.
> >  But I probably do not understand the problem so do not know what to write.
> 
> As long as one can make sure that test kernel boots (commonly the case with
> distros), same kernel can be used as capture kernel too. So the idea here
> is to be able to use same kernel binary as production kernel and capture
> kernel hence distros don't have to ship an additional kernel binary compiled
> for a different physical addr just for dump capturing purposes.

 And you do not want to write protect the kernel code (if the CPU write protection
is not working, the hardware is not working so debug will be difficult, and a simple
CRC32 can tell kernel memory failure) and use twice the same code memory
(with different data area or saving kernel data elsewhere before reload).
 Is that related to module loading or instruction set detection/patch or multiprocessor?

> Currently relocation information is extracted from vmlinux and packed in
> final bzImage after some processing. After execution of real mode code
> and once the image is decompressed, all the relocations are performed and
> then control is transferred to kernel.

  So here you are not really using the initial ELF program header of vmlinux,
but more the section header and my PT_LOAD section bother you, I better
understand. You cannot really claim you are only doing standard/usual
ELF treatment.

 Thanks,
 Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Eric W. Biederman]

Etienne Lorrain <etienne_lorrain@yahoo.fr> writes:

>> Currently relocation information is extracted from vmlinux and packed in
>> final bzImage after some processing. After execution of real mode code
>> and once the image is decompressed, all the relocations are performed and
>> then control is transferred to kernel.
>
>   So here you are not really using the initial ELF program header of vmlinux,
> but more the section header and my PT_LOAD section bother you, I better
> understand. You cannot really claim you are only doing standard/usual
> ELF treatment.

The difference is that a boot loader is not doing the work.  The format is.
Compiling -fPIC almost gets us there except ld has bugs processing absolute
symbols.

Anything we expect a bootloader to do need to be as close as possible to
the usual ELF treatment.

Eric

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Vivek Goyal]

On Thu, Feb 08, 2007 at 09:48:31AM +0000, Etienne Lorrain wrote:
> Vivek Goyal wrote:
> > > Etienne Lorrain wrote:
> > >  Yes, any PT_LOAD below 64 Kbytes can only be real mode, and real-mode
> > > cannot be loaded higher, and cannot be bigger than 640 Kbytes, anything
> > > different (like with virtual address at 0xC0000000) is Linux protected mode.
> > > Considering the linker used it is always the 4th program header, before there
> > > were only 3 program header,third one stay the NOTE one.
> >
> > Well this is all a lot of speculation. This is not standard way of retrieving
> > information from ELF. Number of program headers finally created keep on
> > changing. Previously it was left to the linker and now people have changed
> > it to 3 by specifically using PHDR directive.
> 
>   Well, the standard way is:
> 1 - the code program header, to be write protected if possible
> 2 - the (initialised) data program header, to be able to restart the program by just reload/re-entry,
>     and after that (memsiz - filesz) of BSS
>  Anything after is non standard. The number of program header is written in the ELF header,
> but anything over 2 probably need a special loader for the extra treatment.
>  That isn't usual to have something near zero, because it is better not to map
> any memory there to catch null pointer dereference (in virtual memory system).
> 

May be that's the ideal scenario. But if you leave it to linker to decide the
number of progaram headers to linker then linker sometimes optimizes and
creates more program headers if it finds empty space between various sections.
Sometime one or two page space gets created between sections because of
alignment restrictions. So effectively, number of program headers were not
fixed, till recently.

>  I assume you care about this ELF header because you are also a user of
> the ELF file vmlinux, aren't you?
>   

Yes I am. I use kexec boot loader which has capability to load ELF kernel
images (vmlinux). That's why I am concerned about linking real mode code
in vmlinux as for kdump case I shall have to be aware that kernel vmlinux
might contain a special PT_LOAD type program header which will contain
real mode code and it does not have to be loaded. Then I will run into
guessing business which one is that real mode PT_LOAD  program header and
my assumption might very well break in next few kernel release.
 

> > >  Just I did not completely understand why you need relocation (and the announce
> > > was when I was in holidays far away). I know a simple kernel is needed to do
> > > some debugging when the main kernel has crashed, this kernel is better loaded
> > > at for instance 16 Mbytes. You probably do not want the same kernel as the main
> > > one because it may crash the same way, and you start a loop - and even then
> > > if you have exactly the same kernel it is easier to use the same write-protected
> > > block of memory with different data sections.
> > >  But I probably do not understand the problem so do not know what to write.
> > 
> > As long as one can make sure that test kernel boots (commonly the case with
> > distros), same kernel can be used as capture kernel too. So the idea here
> > is to be able to use same kernel binary as production kernel and capture
> > kernel hence distros don't have to ship an additional kernel binary compiled
> > for a different physical addr just for dump capturing purposes.
> 
>  And you do not want to write protect the kernel code (if the CPU write protection
> is not working, the hardware is not working so debug will be difficult, and a simple
> CRC32 can tell kernel memory failure) and use twice the same code memory
> (with different data area or saving kernel data elsewhere before reload).
>  Is that related to module loading or instruction set detection/patch or multiprocessor?
> 

If I understand it right, you seem to be suggesting that I don't have to 
reload the kernel text and I can only reload the data for second kernel?

We run the whole of the kernel from a mutually execlusive location from
first kernel to mitigate the concerns that first kernel's ongoing DMA might
corrupt second kernel. That's why first kernel's text can't be reused.
Secondly, it gives flexibility to user that either he can choose to use 
the production kernel as capture kernel or an entirely different custom
kernel can be used as capture kernel.
 
> > Currently relocation information is extracted from vmlinux and packed in
> > final bzImage after some processing. After execution of real mode code
> > and once the image is decompressed, all the relocations are performed and
> > then control is transferred to kernel.
> 
>   So here you are not really using the initial ELF program header of vmlinux,
> but more the section header and my PT_LOAD section bother you, I better
> understand. You cannot really claim you are only doing standard/usual
> ELF treatment.
> 

Kexec as a boot loader allows to load both ELF vmlinux file or bzImage.
Hence for kdump, a user got the flexibility to either use vmlinux or 
bzImage for dump captruing purposes. Hence I am concerned about both. 

If real mode code is linked with vmlinux, then kdump will be broken.

bzImage is relocatable. If a new kernel image format is introduced
(compressed ELF), then I will prefer it to be a relocatable one
(if possible).

Thanks
Vivek

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Eric W. Biederman]

Vivek Goyal <vgoyal@in.ibm.com> writes:

> Kexec as a boot loader allows to load both ELF vmlinux file or bzImage.
> Hence for kdump, a user got the flexibility to either use vmlinux or 
> bzImage for dump captruing purposes. Hence I am concerned about both. 
>
> If real mode code is linked with vmlinux, then kdump will be broken.
>
> bzImage is relocatable. If a new kernel image format is introduced
> (compressed ELF), then I will prefer it to be a relocatable one
> (if possible).

If we do make the real mode code findable from a set of ELF headers
instead of a weird program header it should be an ELF note in the note
segment that tells us where to find or possibly contains the real mode
code.

Eric

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

Eric W. Biederman wote:
> H. Peter Anvin wrote:
> > That's not a valid comparison, because you're using 32-bit registers to hold
> > 16-bit pointers, and then sticking 67 prefixes on completely unnecessarily.
>
> Regardless if the size is good enough we can use it :)
>
> For romcc I had a 3x code bloat over had crafted assembler, because I
> had no memory and had to inline every single function call because I
> didn't have someplace to put a return register.  So I have the
> scenario with the most code bloat.  I win!

 It is true that I try to put all my variables in big structures in Gujin, not only
because I think it is cleaner (when variables are related in their use) but also
because it reduces the problem of 32 bits addresses by using offsets to the
structure pointer, most of the time.
 In some cases, using 32 bits address prefix is better because offsets to ESP
are possible but in 16 bits you first have to copy SP to BP to reference stack
parameters. Also you can use any register as a pointer.
 Anyway, we are currently talking of 6-7 Kbytes of real-mode section for the ELF,
including root autodetection and command line generation - there is a lot of
source code to describe all those structures but no "code bloat".
 I did try to remove by "sed" some structure easily converted, like:
  mov 26(%ebx),%eax
 by:
  mov 26(%bx),%eax
for all registers esi,edi,ebx,ebp but it saved so few bytes that I did not keep it.

Eric W. Biederman wote:
> This code is currently completely Gujin specific.  If concieved as a
> replacement for setup.S it has a chance of passing review.

  If someone ever want to transfer some code from Gujin to Linux,
 without reducing too much the functionalities, I would be more than
 happy about it - I could even try to help - but considering that I
 even do not have enough time to timely answer E-mails (considering
 my paying job) and I have a lot of other parts of Gujin to improve,
 I can not lead at all on this project.

  I am not sure this task is even possible, I do not know where
 should be the limit in between bootloader and kernel - considering
 that there is also the initrd (why people didn't switch to initramfs
 already?), and being able to debug the real-mode code (like those
 DBG*.exe Gujin files), return error code, check CRC32 is a minimum.

> LILO is a lot saner then Grub, and it still supports more filesystems...
> Just because it memorizes it all before you shut down the system for
> simplicity doesn't mean lilo is worse.

  LILO is quite good, but I cannot talk too much of other bootloader
 because I simply do not use them enough (excluding ISOLINUX).
 At least LILO do not switch too early to protected mode, like Gujin.
 The thing I do not like about it is that there isn't much checks done,
 like a CRC32, so you may load the right sectors from the wrong disk.
 Also, "ioctl (file_desc, FIBMAP, &block)" seem to be less and less
 supported, either by filesystems themselves of by lower layers like
 LVM and the like. The standalone Linux executable "showmap" in Gujin
 show problems sometimes.
 
 Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[H. Peter Anvin]

Etienne Lorrain wrote:
>  I did try to remove by "sed" some structure easily converted, like:
>   mov 26(%ebx),%eax
>  by:
>   mov 26(%bx),%eax
> for all registers esi,edi,ebx,ebp but it saved so few bytes that I did not keep it.

Well, *of course*, since you still do all your pointer arithmetic in 
32-bit space, plus you can't convert all 32-bit modes this way.  This is 
a far cry from a proper port.

	-hpa

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

H. Peter Anvin, Vivek Goyal wrote:
>> Either way, though, putting Gujin-specific code in the main kernel 
>> output is a pretty dull thud.
>
> Agreed.

 May I ask if you are refering to the Gujin structures under the
ROOT_EXTENSIVE_SEARCH compilation switch, or the
C structures describing the BIOS which could even be used
in the kernel?

Eric W. Biederman wrote:
> From what little I skimmed part of what Gujin wanted to do was sane
> at first glance. Just boot a gziped vmlinux like the other
> architectures.  The problem was the 16bit code.

 So without ROOT_EXTENSIVE_SEARCH defined.
 It is so usefull when you have loads of distributions...
 But yes, maybe that part should be in another file, maybe
 not a complete directory for a single file.
 There is also a special problem with ia32, the number of processor
 involved and their partial compatibilities, that Gujin try to help with.

> So there may be some good ideas buried in there somewhere, but it
> likely to take some doing, and patches that I have to save before
> I read them are a real pain!

 Last time I used this web interface it did not encode text files...

> Eric

 Thanks,
  Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

> Building real mode code with kernel binary (vmlinux) has got another
> disadvantage that it breaks using vmlinux for kdump purposes. One compiles
> the kernel binary to execute from a different address but real mode code/data
> will continue to be at virtual/physical addr 0 and kexec can not load it
> as that physical memory is not available at all. Kdump skips the real mode
> code execution.

 But that is exactly what you want and need for kdump, isn't it?
 The ELF file did not change, the program header has the last index at
address 0 that you do not want to load because you do not want to
execute the real-mode code. Load the rest and provide the 4 Kbytes
parameter page - it should work.

> I don't know much about Gujin, but advantage here seems to be that it has
> capability to load elf files and that's why the attempt to turn kernel binary
> into a compressed elf image. Why don't we then simply add an ELF header to
> bzImage and Gujin and any ELF loader including Gujin, should be able to load
> it? (As Eric had done in one of the implementations in the past?) Why to 
>create the new infrastructure?

 Because I think when a program evolves it has to get simpler to generate,
run and maintain/debug - while doing more stuff. The number of assembler
lines has to reduce because they are difficult to maintain.
 Removing a ELF header to modify the binary and stick another ELF header
is not exactly what I think simpler - a bit like linking at a fix address and
then modifying the whole set.

HPA wrote:
> Well, Gujin wants additional code too.
> Putting an ELF header on bzImage broke some bootloaders (GRUB, I believe),
> so that's not going to happen again. See the relocatable bzImage thread...

  I also refuses to load a big file at a fixed address before asking the BIOS
 for information, I only crash once the memory when I am pretty sure everything
 seems alright, interruption disabled, just before jumping to the kernel
 entry point. Cannot do that with bzImage. Cannot easily debug without
 this feature.

> Thanks
> Vivek

  Thanks for you comment.
 Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com

Re: Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Vivek Goyal]

On Tue, Feb 06, 2007 at 01:12:05PM +0000, Etienne Lorrain wrote:
> > Building real mode code with kernel binary (vmlinux) has got another
> > disadvantage that it breaks using vmlinux for kdump purposes. One compiles
> > the kernel binary to execute from a different address but real mode code/data
> > will continue to be at virtual/physical addr 0 and kexec can not load it
> > as that physical memory is not available at all. Kdump skips the real mode
> > code execution.
> 
>  But that is exactly what you want and need for kdump, isn't it?
>  The ELF file did not change, the program header has the last index at
> address 0 that you do not want to load because you do not want to
> execute the real-mode code. Load the rest and provide the 4 Kbytes
> parameter page - it should work.

How do I know which program header is real mode code and the boot loader
is not supposed to load it? May be PT_LOAD header with physical addr 0?
What happens if changes happen and down the line we start compiling 
real mode code for non-zero address? Hence I think keeping real mode
code out of vmlinux might prove to be a good idea.

Secondly, if you compile real mode code with vmlinux, what would be the
entry point for this ELF file? Real mode entry? Then I have not way to
find out from ELF headers where is the protected mode entry point and
I can not do use this vmlinux with kexec/kdump. 

OTOH, now bzImage is relocatable. Is this image going to be relocatable?
How do we take care of that?

Thanks
Vivek

Re : [PATCH] Compressed ia32 ELF file generation for loading by Gujin 1/3

[Etienne Lorrain]

> First of all, if sending attached patches, *MAKE SURE* they're text/plain.

  Sorry, next time I'll do.

> The standard way to do this is to have a command line argument named 
> BOOT_IMAGE (as in BOOT_IMAGE=/foo/bar/baz).  There is no reason to do 
> this differently from every other bootloader.

  I was more thinking as a command line like the shell command line,
  I can probably modify that. It is only to guess the root filesystem,
  I do not know if BOOT_IMAGE= usually contains the fully qualified path.
  Gujin just adds this string from the "name" field of the GZIP file,
  so it can be manually adjusted.

> Why build the real-mode code as part of the kernel binary?  If you have 
> to reference kernel symbols, you can do that with the -R option to ld. 
> Bundling it into the kernel binary just to then extract it later is 
> silly at best.

  Gujin do not reference any kernel symbol; the complete file is loaded
  at once at a variable memory address. Then Gujin moves the real-mode
  section at a valid address and execute it, then it moves the code and data
  protected mode at a valid address and execute it.
  In this mode Gujin just act as an ELF program loader.
  The variable memory address is good when you are running DOS/HIMEM
  for instance to debug hardware/BIOS - it does not crash random address
  before requesting BIOS services like all other bootloaders.

>> +       @echo  "  /boot/linux-$(KERNELRELEASE).kgz - create a boot kernel for the Gujin bootloader"
> This doesn't exactly fit very well the standard pattern.  Something like 
>  make gujin TARGET=... would be more like it.

  For me, the make parameter is the file you want to get - up to "make" 
  to sort out how to generate it.

> Given how many of your files are highly Gujin-specific, and have generic 
> names, please put them in a subdirectory (e.g. arch/i386/boot/gujin/).

  I do agree that there is a functionality I am not sure of the interface, that
  is the automatic root detection. Nearly everybody put the kernel in its
  own distribution filesystem (Fedora kernel in Fedora root filesystem, Debian
  kernel in Debian root filesystem...) or only has a single distribution.
  The bootloader is loading this kernel file - so it knows exactly which
  disk/partition to propose as root to the kernel.
  So under the compilation switch "ROOT_EXTENSIVE_SEARCH", there
  is the exposed Gujin interface to help the real-mode function of the
  ELF file decide what to do.

  I do disagree that the BIOS interface like "union drive_info" of realmode.h
  is Gujin stuff, it is just that I a lot prefer a C interface with a clean
  structure than "#define OFFSET_OF_THIS_BYTE, #define MASK_OF_THIS_BIT"
  everywhere, I prefer programing in C than programing in CPP.
  Ok, most of the fields are not used - even by Gujin - it is just how the BIOS
  documentation describes them.

>    -hpa

  Thanks for your comments. Sorry I could not answer immediately but
  I had to sleep a bit. I try to reply to the other messages together.
  Etienne.


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com