Jack Ren and Eric Miao tracked down the following long standing
problem in the NOHZ code:
scheduler switch to idle task
enable interrupts
Window starts here
----> interrupt happens (does not set NEED_RESCHED)
irq_exit() stops the tick
----> interrupt happens (does set NEED_RESCHED)
return from schedule()
cpu_idle(): preempt_disable();
Window ends here
The interrupts can happen at any point inside the race window. The
first interrupt stops the tick, the second one causes the scheduler to
rerun and switch away from idle again and we end up with the tick
disabled.
The fact that it needs two interrupts where the first one does not set
NEED_RESCHED and the second one does made the bug obscure and extremly
hard to reproduce and analyse. Kudos to Jack and Eric.
Solution: Limit the NOHZ functionality to the idle loop to make sure
that we can not run into such a situation ever again.
cpu_idle()
{
preempt_disable();
while(1) {
tick_nohz_stop_sched_tick(1); <- tell NOHZ code that we
are in the idle loop
while (!need_resched())
halt();
tick_nohz_restart_sched_tick(); <- disables NOHZ mode
preempt_enable_no_resched();
schedule();
preempt_disable();
}
}
In hindsight we should have done this forever, but ...
/me grabs a large brown paperbag.
Debugged-by: Jack Ren <jack.ren@marvell.com>,
Debugged-by: eric miao <eric.y.miao@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Remove proc_root export. Creation and removal works well if parent PDE is
supplied as NULL -- it worked always that way.
So, one useless export removed and consistency added, some drivers created
PDEs with &proc_root as parent but removed them as NULL and so on.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Avoids sparse warnings:
kernel/sched.c:2170:17: warning: symbol 'schedule_tail' was not declared. Should it be static?
Avoids the need for an external declaration in arch/um/process.c
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Suppress A.OUT library support if CONFIG_ARCH_SUPPORTS_AOUT is not set.
Not all architectures support the A.OUT binfmt, so the ELF binfmt should not
be permitted to go looking for A.OUT libraries to load in such a case. Not
only that, but under such conditions A.OUT core dumps are not produced either.
To make this work, this patch also does the following:
(1) Makes the existence of the contents of linux/a.out.h contingent on
CONFIG_ARCH_SUPPORTS_AOUT.
(2) Renames dump_thread() to aout_dump_thread() as it's only called by A.OUT
core dumping code.
(3) Moves aout_dump_thread() into asm/a.out-core.h and makes it inline. This
is then included only where needed. This means that this bit of arch
code will be stored in the appropriate A.OUT binfmt module rather than
the core kernel.
(4) Drops A.OUT support for Blackfin (according to Mike Frysinger it's not
needed) and FRV.
This patch depends on the previous patch to move STACK_TOP[_MAX] out of
asm/a.out.h and into asm/processor.h as they're required whether or not A.OUT
format is available.
[jdike@addtoit.com: uml: re-remove accidentally restored code]
Signed-off-by: David Howells <dhowells@redhat.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Joe Perches noticed some printks in smp.c that needed fixing.
While I was in there, I did the usual tidying in arch/um/kernel, which
should be fairly style-clean at this point:
copyright updates
emacs formatting comments removal
include tidying
style fixes
Cc: Joe Perches <joe@perches.com>
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix the repetition of the NET symbol. It was once in UML specific options and
once in networking. I removed the first occurrence, as it makes more sense to
me to keep it only in networking.
It also removes a mostly empty file which is not used anymore and some
unused variables.
Signed-off-by: Karol Swietlicki <magotari@gmail.com>
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Some register accessor cleanups -
userspace() was calling restore_registers and save_registers for no
reason, since userspace() is on the libc side of the house, and these
add no value over calling ptrace directly
init_thread_registers and get_safe_registers were the same thing,
so init_thread_registers is gone
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Tidy current-related stuff. There was a comment in current.h saying
that current_thread was obsolete, so this patch turns all instances of
current_thread into current_thread_info(). There's some simplifying
of the result in arch/um/sys-i386/signal.c.
current.h and thread_info also get style cleanups.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Untangle UML headers somewhat and add some includes where they were
needed explicitly, but gotten accidentally via some other header.
arch/um/include/um_uaccess.h loses asm/fixmap.h because it uses no
fixmap stuff and gains elf.h, because it needs FIXADDR_USER_*, and
archsetjmp.h, because it needs jmp_buf.
pmd_alloc_one is uninlined because it needs mm_struct, and that's
inconvenient to provide in asm-um/pgtable-3level.h.
elf_core_copy_fpregs is also uninlined from elf-i386.h and
elf-x86_64.h, which duplicated the code anyway, to
arch/um/kernel/process.c, so that the reference to current_thread
doesn't pull sched.h or anything related into asm/elf.h.
arch/um/sys-i386/ldt.c, arch/um/kernel/tlb.c and
arch/um/kernel/skas/uaccess.c got sched.h because they dereference
task_structs. Its includes of linux and asm headers got turned from
"" to <>.
arch/um/sys-i386/bug.c gets asm/errno.h because it needs errno
constants.
asm/elf-i386 gets asm/user.h because it needs user_regs_struct.
asm/fixmap.h gets page.h because it needs PAGE_SIZE and PAGE_MASK and
system.h for BUG_ON.
asm/pgtable doesn't need sched.h.
asm/processor-generic.h defined mm_segment_t, but didn't use it. So,
that definition is moved to uaccess.h, which defines a bunch of
mm_segment_t-related stuff. thread_info.h uses mm_segment_t, and
includes uaccess.h, which causes a recursion. So, the definition is
placed above the include of thread_info. in uaccess.h. thread_info.h
also gets page.h because it needs PAGE_SIZE.
ObCheckpatchViolationJustification - I'm not adding a typedef; I'm
moving mm_segment_t from one place to another.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patchset makes UML build and run with three-level page tables on
32-bit hosts. This is an uncommon use case, but the code here needed
fixing and cleaning up, so 32-bit three-level pages tables were tested
to make sure the changes are good.
Patch 1 - code movement
Patch 2 - header untangling
Patch 3 - style fixups in files affected so far
Patch 4 - clean up use of current.h
Patch 5 - fix sizes of types that are different between 2 and 3-level
page tables - three-level page table support should build at
this point
Patch 6 - tidy (i.e. eliminate much of) the code that figures out how
big the address space is
Patch 7 - change um_virt_to_phys into virt_to_pte, clean its
interface, and clean its (so far) one caller
Patch 8 - the stub pages are covered with a VMA, allowing some nasty
code to be thrown out - three-level page tables now work
This patch:
um_virt_to_phys only has one user, so it can be moved to the same file
and made static. Its declarations in pgtable.h and ksyms.c are also
gone.
current_cmd was another apparent user, but it itself isn't used, so it
is deleted.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch removes a variable which was not used in two functions. Yet
another code cleanup, nothing really significant.
Please note that I could not test this on x86_64. I don't have the
hardware for it.
[ jdike - Bits of tidying around the affected code. Also, it's fine on
x86_64 ]
Signed-off-by: Karol Swietlicki <magotari@gmail.com>
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch also does some improvements for uml code. Improvements include
dropping unnecessary cast, killing some unnecessary code and still some
constifying for pointers etc..
Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Implement get_wchan - the algorithm is similar to x86. It starts with the
stack pointer of the process in question and looks above that for addresses
that are kernel text. The second one which isn't in the scheduler is the one
that's returned. The first one is ignored because that will be UML's own
context switching routine.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now, the idle loop now longer needs SIGALRM firing - it can just sleep for the
requisite amount of time and fake a timer interrupt when it finishes.
Any use of ITIMER_REAL now goes away. disable_timer only turns off
ITIMER_VIRTUAL. switch_timers is no longer needed, so it, and all calls, goes
away.
disable_timer now returns the amount of time remaining on the timer.
default_idle uses this to tell idle_sleep how long to sleep. idle_sleep will
call alarm_handler if nanosleep returns 0, which is the case if it didn't
return early due to an interrupt. Otherwise, it just returns.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Enable tickless support.
CONFIG_TICK_ONESHOT and CONFIG_NO_HZ are enabled.
itimer_clockevent gets CLOCK_EVT_FEAT_ONESHOT and an implementation of
.set_next_event.
CONFIG_UML_REAL_TIME_CLOCK goes away because it only makes sense when there is
a clock ticking away all the time. timer_handler now just calls do_IRQ once
without trying to figure out how many ticks to emulate.
The idle loop now needs to turn ticking on and off.
Userspace ticks keep happening as usual. However, the userspace loop keep
track of when the next wakeup should happen and suppresses process ticks until
that happens.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix up the switching between virtual and real timers. The idle loop sleeps,
so the timer at that point must be real time. At all other times, the timer
must be virtual. Even when userspace is running, and the kernel is asleep,
the virtual timer is correct because the process timer will be running and the
process timer will be firing.
The timer switch used to be in the context switch and timer handler code.
This is moved to the idle loop and the signal handler, making it much more
clear why it is happening.
switch_timers now returns the old timer type so that it may be restored. The
signal handler uses this in order to restore the previous timer type when it
returns.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Before the removal of tt mode, access to a register on the skas-mode side of a
pt_regs struct looked like pt_regs.regs.skas.regs.regs[FOO]. This was bad
enough, but it became pt_regs.regs.regs.regs[FOO] with the removal of the
union from the middle. To get rid of the run of three "regs", the last field
is renamed to "gp".
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Formatting changes in the files which have been changed in the course
of folding foo_skas functions into their callers. These include:
copyright updates
header file trimming
style fixes
adding severity to printks
These changes should be entirely non-functional.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch makes a number of simplifications enabled by the removal of
CHOOSE_MODE. There were lots of functions that looked like
int foo(args){
foo_skas(args);
}
The bodies of foo_skas are now folded into foo, and their declarations (and
sometimes entire header files) are deleted.
In addition, the union uml_pt_regs, which was a union between the tt and skas
register formats, is now a struct, with the tt-mode arm of the union being
removed.
It turns out that usr2_handler was unused, so it is gone.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The next stage after removing code which depends on CONFIG_MODE_TT is removing
the CHOOSE_MODE abstraction, which provided both compile-time and run-time
branching to either tt-mode or skas-mode code.
This patch removes choose-mode.h and all inclusions of it, and replaces all
CHOOSE_MODE invocations with the skas branch. This leaves a number of trivial
functions which will be dealt with in a later patch.
There are some changes in the uaccess and tls support which go somewhat beyond
this and eliminate some of the now-redundant functions.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patchset throws out tt mode, which has been non-functional for a while.
This is done in phases, interspersed with code cleanups on the affected files.
The removal is done as follows:
remove all code, config options, and files which depend on
CONFIG_MODE_TT
get rid of the CHOOSE_MODE macro, which decided whether to
call tt-mode or skas-mode code, and replace invocations with their
skas portions
replace all now-trivial procedures with their skas equivalents
There are now a bunch of now-redundant pieces of data structures, including
mode-specific pieces of the thread structure, pt_regs, and mm_context. These
are all replaced with their skas-specific contents.
As part of the ongoing style compliance project, I made a style pass over all
files that were changed. There are three such patches, one for each phase,
covering the files affected by that phase but no later ones.
I noticed that we weren't freeing the LDT state associated with a process when
it exited, so that's fixed in one of the later patches.
The last patch is a tidying patch which I've had for a while, but which caused
inexplicable crashes under tt mode. Since that is no longer a problem, this
can now go in.
This patch:
Start getting rid of tt mode support.
This patch throws out CONFIG_MODE_TT and all config options, code, and files
which depend on it.
CONFIG_MODE_SKAS is gone and everything that depends on it is included
unconditionally.
The few changed lines are in re-written Kconfig help, lines which needed
something skas-related removed from them, and a few more which weren't
strictly deletions.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Map all of physical memory as executable to avoid having to change stack
protections during fork and exit.
unprotect_stack is now called only from MODE_TT code, so it is marked as such.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
UML had two wrapper procedures for kmalloc, um_kmalloc and um_kmalloc_atomic
because the flag constants weren't available in userspace code.
kern_constants.h had made kernel constants available for a long time, so there
is no need for these wrappers any more. Rather, userspace code calls kmalloc
directly with the userspace versions of the gfp flags.
kmalloc isn't a real procedure, so I had to essentially copy the inline
wrapper around __kmalloc.
vmalloc also had its own wrapper for no good reason. This is now gone.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kernel_thread() should just return an error value on do_fork failure, not
panic.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
userspace code used to have to call the kernelspace function page_size() in
order to determine the value of the kernel's PAGE_SIZE. Since this is now
available directly from kern_constants.h as UM_KERN_PAGE_SIZE, page_size() can
be deleted and calls changed to use the constant.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Clean up arch/um/kernel/process.c:
- lots of return(x); -> return x; conversions
- a number of the small functions are either unused, in which case they are
gone, along any declarations in a header, or could be made static.
- current_pid is ifdefed on CONFIG_MODE_TT and its declaration is ifdefed on
both CONFIG_MODE_TT and UML_CONFIG_MODE_TT because we don't know whether
it's being used in a userspace or kernel file.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
user_util.h isn't needed any more, so delete it and remove all includes of it.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch moves all the the symbols defined in um_arch.c, which are mostly
boundaries between different parts of the UML kernel address space, to a new
header, as-layout.h. There are also a few things here which aren't really
related to address space layout, but which don't really have a better place to
go.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
user.h is too generic a header name. I've split out allocation routines from
it.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Move some foo_kern.c files to foo.c now that the old foo.c files are out
of the way.
Also cleaned up some whitespace and an emacs formatting comment.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The serial UML OS-abstraction layer patch (um/kernel dir).
This moves all systemcalls from process.c file under os-Linux dir and join
process.c and process_kern.c files.
Signed-off-by: Gennady Sharapov <gennady.v.sharapov@intel.com>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
With Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Revert the following patch, because of miscompilation problems in different
environments leading to UML not working *at all* in TT mode; it was merged
lately in 2.6 development cycle, a little after being written, and has
caused problems to lots of people; I know it's a bit too long, but it
shouldn't have been merged in first place, so I still apply for inclusion
in the -stable tree. Anyone using this feature currently is either using
some older kernel (some reports even used 2.6.12-rc4-mm2) or using this
patch, as included in my -bs patchset.
For now there's not yet a fix for this patch, so for now the best thing is
to drop it (which was widely reported to give a working kernel, and as such
was even merged in -stable tree).
"Convert the boot-time host ptrace testing from clone to fork. They were
essentially doing fork anyway. This cleans up the code a bit, and makes
valgrind a bit happier about grinding it."
URL:
http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=98fdffccea6cc3fe9dba32c0fcc310bcb5d71529
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This adds the "skas0" parameter to force skas0 operation on SKAS3 host and
shows which operating mode has been selected.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
UML has had two modes of operation - an insecure, slow mode (tt mode) in
which the kernel is mapped into every process address space which requires
no host kernel modifications, and a secure, faster mode (skas mode) in
which the UML kernel is in a separate host address space, which requires a
patch to the host kernel.
This patch implements something very close to skas mode for hosts which
don't support skas - I'm calling this skas0. It provides the security of
the skas host patch, and some of the performance gains.
The two main things that are provided by the skas patch, /proc/mm and
PTRACE_FAULTINFO, are implemented in a way that require no host patch.
For the remote address space changing stuff (mmap, munmap, and mprotect),
we set aside two pages in the process above its stack, one of which
contains a little bit of code which can call mmap et al.
To update the address space, the system call information (system call
number and arguments) are written to the stub page above the code. The
%esp is set to the beginning of the data, the %eip is set the the start of
the stub, and it repeatedly pops the information into its registers and
makes the system call until it sees a system call number of zero. This is
to amortize the cost of the context switch across multiple address space
updates.
When the updates are done, it SIGSTOPs itself, and the kernel process
continues what it was doing.
For a PTRACE_FAULTINFO replacement, we set up a SIGSEGV handler in the
child, and let it handle segfaults rather than nullifying them. The
handler is in the same page as the mmap stub. The second page is used as
the stack. The handler reads cr2 and err from the sigcontext, sticks them
at the base of the stack in a faultinfo struct, and SIGSTOPs itself. The
kernel then reads the faultinfo and handles the fault.
A complication on x86_64 is that this involves resetting the registers to
the segfault values when the process is inside the kill system call. This
breaks on x86_64 because %rcx will contain %rip because you tell SYSRET
where to return to by putting the value in %rcx. So, this corrupts $rcx on
return from the segfault. To work around this, I added an
arch_finish_segv, which on x86 does nothing, but which on x86_64 ptraces
the child back through the sigreturn. This causes %rcx to be restored by
sigreturn and avoids the corruption. Ultimately, I think I will replace
this with the trick of having it send itself a blocked signal which will be
unblocked by the sigreturn. This will allow it to be stopped just after
the sigreturn, and PTRACE_SYSCALLed without all the back-and-forth of
PTRACE_SYSCALLing it through sigreturn.
This runs on a stock host, so theoretically (and hopefully), tt mode isn't
needed any more. We need to make sure that this is better in every way
than tt mode, though. I'm concerned about the speed of address space
updates and page fault handling, since they involve extra round-trips to
the child. We can amortize the round-trip cost for large address space
updates by writing all of the operations to the data page and having the
child execute them all at the same time. This will help fork and exec, but
not page faults, since they involve only one page.
I can't think of any way to help page faults, except to add something like
PTRACE_FAULTINFO to the host. There is PTRACE_SIGINFO, but UML doesn't use
siginfo for SIGSEGV (or anything else) because there isn't enough
information in the siginfo struct to handle page faults (the faulting
operation type is missing). Adding that would make PTRACE_SIGINFO a usable
equivalent to PTRACE_FAULTINFO.
As for the code itself:
- The system call stub is in arch/um/kernel/sys-$(SUBARCH)/stub.S. It is
put in its own section of the binary along with stub_segv_handler in
arch/um/kernel/skas/process.c. This is manipulated with run_syscall_stub
in arch/um/kernel/skas/mem_user.c. syscall_stub will execute any system
call at all, but it's only used for mmap, munmap, and mprotect.
- The x86_64 stub calls sigreturn by hand rather than allowing the normal
sigreturn to happen, because the normal sigreturn is a SA_RESTORER in
UML's address space provided by libc. Needless to say, this is not
available in the child's address space. Also, it does a couple of odd
pops before that which restore the stack to the state it was in at the
time the signal handler was called.
- There is a new field in the arch mmu_context, which is now a union.
This is the pid to be manipulated rather than the /proc/mm file
descriptor. Code which deals with this now checks proc_mm to see whether
it should use the usual skas code or the new code.
- userspace_tramp is now used to create a new host process for every UML
process, rather than one per UML processor. It checks proc_mm and
ptrace_faultinfo to decide whether to map in the pages above its stack.
- start_userspace now makes CLONE_VM conditional on proc_mm since we need
separate address spaces now.
- switch_mm_skas now just sets userspace_pid[0] to the new pid rather
than PTRACE_SWITCH_MM. There is an addition to userspace which updates
its idea of the pid being manipulated each time around the loop. This is
important on exec, when the pid will change underneath userspace().
- The stub page has a pte, but it can't be mapped in using tlb_flush
because it is part of tlb_flush. This is why it's required for it to be
mapped in by userspace_tramp.
Other random things:
- The stub section in uml.lds.S is page aligned. This page is written
out to the backing vm file in setup_physmem because it is mapped from
there into user processes.
- There's some confusion with TASK_SIZE now that there are a couple of
extra pages that the process can't use. TASK_SIZE is considered by the
elf code to be the usable process memory, which is reasonable, so it is
decreased by two pages. This confuses the definition of
USER_PGDS_IN_LAST_PML4, making it too small because of the rounding down
of the uneven division. So we round it to the nearest PGDIR_SIZE rather
than the lower one.
- I added a missing PT_SYSCALL_ARG6_OFFSET macro.
- um_mmu.h was made into a userspace-usable file.
- proc_mm and ptrace_faultinfo are globals which say whether the host
supports these features.
- There is a bad interaction between the mm.nr_ptes check at the end of
exit_mmap, stack randomization, and skas0. exit_mmap will stop freeing
pages at the PGDIR_SIZE boundary after the last vma. If the stack isn't
on the last page table page, the last pte page won't be freed, as it
should be since the stub ptes are there, and exit_mmap will BUG because
there is an unfreed page. To get around this, TASK_SIZE is set to the
next lowest PGDIR_SIZE boundary and mm->nr_ptes is decremented after the
calls to init_stub_pte. This ensures that we know the process stack (and
all other process mappings) will be below the top page table page, and
thus we know that mm->nr_ptes will be one too many, and can be
decremented.
Things that need fixing:
- We may need better assurrences that the stub code is PIC.
- The stub pte is set up in init_new_context_skas.
- alloc_pgdir is probably the right place.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Convert the boot-time host ptrace testing from clone to fork. They were
essentially doing fork anyway. This cleans up the code a bit, and makes
valgrind a bit happier about grinding it.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
A few files include the same header twice.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This makes SIGWINCH work again, and fixes a couple of SIGWINCH-associated
crashes. First, the sigio thread disables SIGWINCH because all hell breaks
loose if it ever gets one and tries to call the signal handling code. Second,
there was a problem with deferencing tty structs after they were freed. The
SIGWINCH support for a tty wasn't being turned off or freed after the tty went
away.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
