Commit ee3d9bd4de ("uml: simplify SIGSEGV
handling"), while greatly simplifying the kernel SIGSEGV handler that
runs in the process address space, introduced a bug which corrupts FP
state in the process.
Previously, the SIGSEGV handler called the sigreturn system call by hand - it
couldn't return through the restorer provided to it because that could try to
call the libc restorer which likely wouldn't exist in the process address
space. So, it blocked off some signals, including SIGUSR1, on entry to the
SIGSEGV handler, queued a SIGUSR1 to itself, and invoked sigreturn. The
SIGUSR1 was delivered, and was visible to the UML kernel after sigreturn
finished.
The commit eliminated the signal masking and the call to sigreturn. The
handler simply hits itself with a SIGTRAP to let the UML kernel know that it
is finished. UML then restores the process registers, which effectively
longjmps the process out of the signal handler, skipping sigreturn's restoring
of register state and the signal mask.
The bug is that the host apparently sets used_fp to 0 when it saves the
process FP state in the sigcontext on the process signal stack. Thus, when
the process is longjmped out of the handler, its FP state is corrupt because
it wasn't saved on the context switch to the UML kernel.
This manifested itself as sleep hanging. For some reason, sleep uses floating
point in order to calculate the sleep interval. When a page fault corrupts
its FP state, it is faked into essentially sleeping forever.
This patch saves the FP state before entering the SIGSEGV handler and restores
it afterwards.
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 was panicing in the case of failures of libc calls which shouldn't happen.
This is an overreaction since a failure from libc doesn't normally mean that
kernel data structures are in an unknown state. Instead, the current process
should just be killed if there is no way to recover.
The case that prompted this was a failure of PTRACE_SETREGS restoring the same
state that was read by PTRACE_GETREGS. It appears that when a process tries
to load a bogus value into a segment register, it segfaults (as expected) and
the value is actually loaded and is seen by PTRACE_GETREGS (not expected).
This case is fixed by forcing a fatal SIGSEGV on the process so that it
immediately dies. fatal_sigsegv was added for this purpose. It was declared
as noreturn, so in order to pursuade gcc that it actually does not return, I
added a call to os_dump_core (and declared it noreturn) so that I get a core
file if somehow the process survives.
All other calls in arch/um/os-Linux/skas/process.c got the same treatment,
with failures causing the process to die instead of a kernel panic, with some
exceptions.
userspace_tramp exits with status 1 if anything goes wrong there. That will
cause start_userspace to return an error. copy_context_skas0 and
map_stub_pages also now return errors instead of panicing. Callers of thes
functions were changed to check for errors and do something appropriate.
Usually that's to return an error to their callers.
check_skas3_ptrace_faultinfo just exits since that's too early to do anything
else.
save_registers, restore_registers, and init_registers now return status
instead of panicing on failure, with their callers doing something
appropriate.
There were also duplicate declarations of save_registers and restore_registers
in os.h - these are gone.
I noticed and fixed up some whitespace damage.
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>
Handle floating point state in across signals correctly. UML/i386 needs to
know whether the host does PTRACE_[GS]ETFPXREGS, so an arch_init_registers
hook is added, which on x86_64 does nothing.
UML doesn't save and restore floating point registers on kernel entry and
exit, so they need to be copied between the host process and the sigcontext.
save_fpx_registers and restore_fpx_registers are added for this purpose.
save_fp_registers and restore_fp_registers already exist.
There was a bunch of floating point state conversion code in
arch/um/sys-i386/ptrace.c which isn't needed there, but is needed in signal.c,
so it is moved over.
The i386 code now distinguishes between fp and fpx state and handles them
correctly. The x86_64 code just needs to copy state as-is between the host
process and the stack. There are also some fixes there to pass the correct
address of the floating point state around.
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>
Throw out a lot of code dealing with saving and restoring floating-point
state. In skas mode, where processes run in a restoring floating-point state
on kernel entry and exit is pointless.
This eliminates most of arch/um/os-Linux/sys-{i386,x86_64}/registers.c. Most
of what remained is now arch-indpendent, and can be moved up to
arch/um/os-Linux/registers.c. Both arches need the jmp_buf accessor
get_thread_reg, and i386 needs {save,restore}_fp_regs because it cheats during
sigreturn by getting the fp state using ptrace rather than copying it out of
the process sigcontext.
After this, it turns out that arch/um/include/skas/mode-skas.h is almost
completely unneeded. The declarations in it are variables which either don't
exist or which don't have global scope. The one exception is
kill_off_processes_skas. If that's removed, this header can be deleted.
This uncovered a bug in user.h, which wasn't correctly making sure that a
size_t definition was available to both userspace and kernelspace files.
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 KSTK_* macros used an inordinate amount of stack. In order to overcome
an impedance mismatch between their interface, which just returns a single
register value, and the interface of get_thread_regs, which took a full
pt_regs, the implementation created an on-stack pt_regs, filled it in, and
returned one field. do_task_stat calls KSTK_* twice, resulting in two
local pt_regs, blowing out the stack.
This patch changes the interface (and name) of get_thread_regs to just
return a single register from a jmp_buf.
The include of archsetjmp.h" in registers.h to get the definition of
jmp_buf exposed a bogus include of <setjmp.h> in start_up.c. <setjmp.h>
shouldn't be used anywhere any more since UML uses the klibc
setjmp/longjmp.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
We weren't making sure that we initialized the FP registers of new processes
to sane values.
This patch also moves some defines in the affected area closer to where they
are used.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The old code had the IP and SP coming from the registers in the thread
struct, which are completely wrong since those are the userspace
registers. This fixes that by pulling the correct values from the
jmp_buf in which the kernel state of each thread is stored.
Signed-off-by: Allan Graves <allan.graves@oracle.com>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
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>
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!
Jeff Dike
Allan Graves
Linus Torvalds