Typo in linux-2.6.38/arch/x86/kernel/entry_32.S

Hi,

Trying to compile 2.6.38, I found the follwoing bug. Compiler barks...

diff linux-2.6.38/arch/x86/kernel/entry_32.S linux-2.6.38/arch/x86/kernel/entry_32.S~ 1416c1416
< END(async_page_fault)
---

--eric

New compiler? (new binutils more likely)

Ingo and Thomas,

IIRC, you have patches to fix these, right?

Thanks,

-- Steve

Maybe. I use debian unstable + experimental and upgrade packages almost everyday...

But obviously the begin end is not symmetric as in other function.

You probably have new binutils.

What does "as --version" give you. Binutils 2.21... ?

I think it was 2.21 that started giving problems.

Right, there was a discussion before about this causing issues with new binutils, and I think Ingo or Thomas said that they had patches to fix these for x86. Which this looks like one.

The old binutils never complained.

-- Steve

Well I found the problem at work and I'm home now but since I use the same distrib and just updated, its probably the same.

as --version
GNU assembler (GNU Binutils for Debian) 2.21.0.20110302

So probably most debian user will soon or late have the problem.

Sound plausible.

Anyway this show once again that the world like symetry and you better remember it ;-) (lock/unlock, malloc free, ...)

Probably a stable candidate for 2.6.38.1

--eric

Right, but it should be a warning not an error. If it worked before, why not work now and just warn about it. This "BUG" existed forever in the kernel, and its in more than one place. Having the compiler crash on the first instance means you need to fix it and run the compiler again to find the next instance and so on.

binutils made a fatal error in making this a bug and not an warning.

Perhaps, but what about all older kernels. Now this version of binutils does not work for bisecting.

-- Steve

Yes like any other name change in a .h file with incompatible function signature...

The exact error message is a complaint about the .size directive not being a "constant"...

I guess its binutils git of the 2011/03/02

As far as I'm concerned I only found one occurence on one machine although I compiled tree 2.6.38 kernels for different machines. because its under the CONFIG_KVM_GUEST flag

--eric

Unofortunately there's also a CONFIG_XEN=y build breakage caused by this new binutils behavior, going back from v2.6.38 to v2.6.27-rc1 and breaking bisection builds for over 130,000 commits covering an almost 3 years window of kernel history.

GAS allowed such debug symbol mismatches for over a decade and never reported anything. It's nice that binutils now checks for this condition (we need more automation in this area) and the most helpful would be for the assembler to report the mismatch via a new *warning* - and by also ignore the symbol in question.

Instead this harmless condition was escallated into an unnecessary, retroactive, show-stopping build failure, which is an exceedingly lame and harmful. Tools should be permissive, and they should help not hinder development.

The kernel side fix is upstream already and i've attached it below. That obviously does not help bisecting testers who will trigger the build failure after the first bisection step ...

Thanks,

Ingo

----------------->

From: Alexander van Heukelum <[email protected]>
Date: Fri, 11 Mar 2011 21:59:38 +0100
Subject: [PATCH] x86, binutils, xen: Fix another wrong size directive

The latest binutils (2.21.0.20110302/Ubuntu) breaks the build yet another time, under CONFIG_XEN=y due to a .size directive that refers to a slightly differently named (hence, to the now very strict and unforgiving assembler, non-existent) symbol.

[ mingo:

This unnecessary build breakage caused by new binutils version 2.21 gets escallated back several kernel releases spanning several years of Linux history, affecting over 130,000 upstream kernel commits (!), on CONFIG_XEN=y 64-bit kernels (i.e. essentially affecting all major Linux distro kernel configs).

Git annotate tells us that this slight debug symbol code mismatch bug has been introduced in 2008 in commit 3d75e1b8:

     3d75e1b8        (Jeremy Fitzhardinge    2008-07-08 15:06:49 -0700 1231) ENTRY(xen_do_hypervisor_callback)   # do_hypervisor_callback(struct *pt_regs)

The 'bug' is just a slight assymetry in ENTRY()/END() debug-symbols sequences, with lots of assembly code between the ENTRY() and the END():

ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs)
...
END(do_hypervisor_callback)

Human reviewers almost never catch such small mismatches, and binutils never even warned about it either.

This new binutils version thus breaks the Xen build on all upstream kernels since v2.6.27, out of the blue.

This makes a straightforward Git bisection of all 64-bit Xen-enabled kernels impossible on such binutils, for a bisection window of over hundred thousand historic commits. (!)

This is a major fail on the side of binutils and binutils needs to turn this show-stopper build failure into a warning ASAP. ]

LKML-Reference: <[email protected]>

---
 arch/x86/kernel/entry_64.S |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

index aed1ffb..bbd5c80 100644
@@ -1248,7 +1248,7 @@ ENTRY(xen_do_hypervisor_callback)   # do_hypervisor_callback(struct *pt_regs)
 	decl PER_CPU_VAR(irq_count)
 	jmp  error_exit
 	CFI_ENDPROC
-END(do_hypervisor_callback)
+END(xen_do_hypervisor_callback)
 
 /*
  * Hypervisor uses this for application faults while it executes.