i spent a fair amount of time chasing a 64-bit bootup crash that manifested
itself as bootup segfaults:
S10network[1825]: segfault at 7f3e2b5d16b8 ip 00000031108748c9 sp 00007fffb9c14c70 error 4 in libc-2.7.so[3110800000+14d000]
eventually causing init to die and panic the system:
Kernel panic - not syncing: Attempted to kill init!
Pid: 1, comm: init Not tainted 2.6.26-rc9-tip #13878
after a maratonic bisection session, the bad commit turned out to be:
| b7675791859075418199c7af86a116ea34eaf5bd is first bad commit
| commit b7675791859075418199c7af86a116ea34eaf5bd
| Author: Jeremy Fitzhardinge <jeremy@goop.org>
| Date: Wed Jun 25 00:19:00 2008 -0400
|
| x86: remove open-coded save/load segment operations
|
| This removes a pile of buggy open-coded implementations of savesegment
| and loadsegment.
after some more bisection of this patch itself, it turns out that what
makes the difference are the savesegment() changes to __switch_to().
Taking a look at this portion of arch/x86/kernel/process_64.o revealed
this crutial difference:
| good: 99c: 8c e0 mov %fs,%eax
| 99e: 89 45 cc mov %eax,-0x34(%rbp)
|
| bad: 99c: 8c 65 cc mov %fs,-0x34(%rbp)
which is due to:
| unsigned fsindex;
| - asm volatile("movl %%fs,%0" : "=r" (fsindex));
| + savesegment(fs, fsindex);
savesegment() is implemented as:
#define savesegment(seg, value) \
asm("mov %%" #seg ",%0":"=rm" (value) : : "memory")
note the "m" modifier - it allows GCC to generate the segment move
into a memory operand as well.
But regarding segment operands there's a subtle detail in the x86
instruction set: the above 16-bit moves are zero-extend, but only
if it goes to a register.
If it goes to a memory operand, -0x34(%rbp) in the above case, there's
no zero-extend to 32-bit and the instruction will only save 16 bits
instead of the intended 32-bit.
The other 16 bits is random data - which can cause problems when that
value is used later on.
The solution is to only allow segment operands to go to registers.
This fix allows my test-system to boot up without crashing.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Add "memory" clobbers to savesegment and loadsegment, since they can
affect memory accesses and we never want the compiler to reorder them
with respect to memory references.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Cc: xen-devel <xen-devel@lists.xensource.com>
Cc: Stephen Tweedie <sct@redhat.com>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Liu Pingfan noticed that switch_to() clobbers more registers than its
asm constraints specify.
We get away with this due to luck mostly - schedule()
by its nature only has 'local' state which gets reloaded
automatically. Fix it nevertheless, we could hit this anytime.
it turns out that with the extra constraints gcc manages to make
schedule() even more compact:
text data bss dec hex filename
28626 684 2640 31950 7cce sched.o.before
28613 684 2640 31937 7cc1 sched.o.after
Reported-by: Liu Pingfan <kernelfans@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Make the code more readable and more hackable:
- use symbolic asm parameters
- use readable indentation
- add comments that explains the details
No code changed:
kernel/sched.o:
text data bss dec hex filename
28626 684 2640 31950 7cce sched.o.before
28626 684 2640 31950 7cce sched.o.after
md5:
2823d406c18b781975cdb2e7cfea0059 sched.o.before.asm
2823d406c18b781975cdb2e7cfea0059 sched.o.after.asm
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Comment says wmb is a nop, but it is implemented as lock addl
below... Should it be compiled to nop if we know we are running on
"good" Intel cpu?
At least remove confusing comment for now.
Signed-off-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
A few snuck back in to x86.
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The volatile keyword was removed from the clflush() prototype
in commit e34907ae180f4fe6c28bb4516c679c2f81b0c9ed; the comment there
states:
x86: remove volatile keyword from clflush.
the p parameter is an explicit memory reference, and is
enough to prevent gcc to being nasty here. The volatile
seems completely not needed.
This reflects incorrect understanding of the function of the volatile
keyword there. The purpose of the volatile keyword is informing gcc
that it is safe to pass a volatile pointer to this function.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Use the _ASM_EXTABLE macro from <asm/asm.h>, instead of open-coding
__ex_table entires in include/asm-x86/system.h.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Since the cr8 manipulation functions ended up staying in the tree,
they can't be defined just when PARAVIRT is off: In this patch,
those functions are defined for the PARAVIRT case too.
[ mingo@elte.hu: fixes ]
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
rdtsc_barrier() is a new barrier primitive that stops RDTSC speculation
to avoid races with timer interrupts on other CPUs.
It expands either to LFENCE (for Intel CPUs) or MFENCE (for
AMD CPUs) which stops RDTSC on all currently known microarchitectures
that implement SSE. On CPUs without SSE there is generally no RDTSC
speculation.
[ mingo@elte.hu: renamed it to rdtsc_barrier() and made it x86-only ]
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The patch introducing this left out x86-64, despite it also having
extra entries.
[ mingo@elte.hu: re-merged this to after the unification patches. ]
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch finishes the unification of system.h file.
i386 needs a constant to be defined, and it is defined inside an ifdef.
Other than that, pretty much nothing but includes are left in the arch
specific headers, and they are deleted.
[ mingo@elte.hu: 64-bit needs the cr8 access inlines. ]
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch moves the switch_to() macro to system.h
As those macros are fundamentally different between i386 and x86_64,
they are enclosed around an ifdef.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The memory barrier parts of system.h are not very different between
i386 and x86_64, the main difference being the availability of
instructions, which we handle with the use of ifdefs.
They are consolidated in system.h file, and then removed from
the arch-specific headers.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch moves the i386 control registers manipulation functions,
wbinvd, and clts functions to system.h. They are essentially the same
as in x86_64.
With this, system.h paravirt comes for free in x86_64.
[ mingo@elte.hu: reintroduced the cr8 bits - needed for resume images ]
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch unifies the load_segment() macro, making them equal in both
x86_64 and i386 architectures. The common version goes to system.h,
and the old are deleted.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch puts together pieces of system_{32,64}.h that
looks like the same. It's the first step towards integration
of this file.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Move the headers to include/asm-x86 and fixup the
header install make rules
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Ingo Molnar