From: Al Viro <viro@zeniv.linux.org.uk>
uml __user annotations
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
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>
*) Rather than duplicate in various buggy ways the application of
CFLAGS_NO_HARDENING and UNPROFILE (which apply to the same files),
centralize it in Makefile.rules. UNPROFILE_OBJS mustn't be listed in
USER_OBJS but are compiled as such.
I've also verified that unprofile didn't work in the current form, because we
set _c_flags directly (using CFLAGS and not USER_CFLAGS, which is wrong),
which is normally used by c_flags, but we also override c_flags for all
USER_OBJS, and there we don't call unprofile.
Instead it only worked for unmap.o, the only one which wasn't a USER_OBJ.
We need to set c_flags (which is not a public Kbuild API) to clear a lot of
compilation flags like -nostdinc which Kbuild forces on everything.
*) Rather than $(CFLAGS_$(notdir $@)), which expands to CFLAGS_anObj.s when
building "anObj.s", use $(CFLAGS_$(*F).o) which always accesses
CFLAGS_anObj.o, like done by Kbuild.
*) Make c_flags apply to all targets having the same basename, rather than
listing .s, .i, .lst and .o, with the use (which I tested) of
$(USER_OBJS:.o=.%): c_flags = ...
and of
- $(obj)/unmap.c: _c_flags = ...
+ $(obj)/unmap.%: _c_flags = ...
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
To make some half-assembly stubs compile, disable various "hardened" GCC
features:
*) we can't make it build PIC code as we need %ebx to do syscalls and GCC
wants it free for PIC
*) we can't leave stack protection as the stub is moved (not relocated!) in
memory so the RIP-relative access to the canary tries reading from an
unmapped address and causes a segfault, since we move the stub of various
megabytes (the exact amount will be decided at runtime) away from the
link-time address.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Move the build of user-offsets to arch/um/sys-$(SUBARCH), where it's located.
So we can also build it via Kbuild with its dependency tracking rather than by
hand. While hacking here, fix also a lot of little cosmetic things.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
We were missing __volatile__ on some bits of asm in the segfault handlers.
On x86_64, this was messing up the move from %rdx to uc because that was
moved to after the GET_FAULTINFO_FROM_SC, which changed %rdx.
Also changed the other bit of asm and the one in the i386 handler to
prevent any similar occurrences.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
bits of uml __user annotations lost in merge
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Noticed this for a compilation-time warning, so I'm fixing it even for TT mode
- this is not put_user, but copy_to_user, so we need a pointer to sp, not sp
itself (we're trying to write the word pointed to by the "sp" var.).
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>
Some functions are exported twice in current code - remove the excess export.
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>
Now that GCC warns about format errors, fix them. Nothing able to cause a
crash, however.
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>
Two small TLS fixes -
arch/um/os-Linux/sys-i386/tls.c uses errno and -E* so it should include
errno.h
__setup_host_supports_tls returns 1, but as an initcall, it should return 0
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
If running on a host not supporting TLS (for instance 2.4) we should report
that cleanly to the user, instead of printing not comprehensible "error 5" for
that.
Additionally, i386 and x86_64 support different ranges for
user_desc->entry_number, and we must account for that; we couldn't pass
ourselves -1 because we need to override previously existing TLS descriptors
which glibc has possibly set, so test at startup the range to use.
x86 and x86_64 existing ranges are hardcoded.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Newly forked threads have no arch_switch_to_skas() called before their first
run, because when schedule() switches to them they're resumed in the body of
thread_wait() inside fork_handler() rather than in switch_threads() in
switch_to_skas(). Compensate this missing call.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Implement sys_[gs]et_thread_area and the corresponding ptrace operations for
UML. This is the main chunk, additional parts follow. This implementation is
now well tested and has run reliably for some time, and we've understood all
the previously existing problems.
Their implementation saves the new GDT content and then forwards the call to
the host when appropriate, i.e. immediately when the target process is
running or on context switch otherwise (i.e. on fork and on ptrace() calls).
In SKAS mode, we must switch registers on each context switch (because SKAS
does not switches tls_array together with current->mm).
Also, added get_cpu() locking; this has been done for SKAS mode, since TT does
not need it (it does not use smp_processor_id()).
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Call arch_switch also in switch_to_skas, even if it's, for now, a no-op for
that case (and mark this in the comment); this will change soon.
Also, arch_switch for TT mode is actually useless when the PT proxy (a
complicate debugging instrumentation for TT mode) is not enabled. In fact, it
only calls update_debugregs, which checks debugregs_seq against seq (to check
if the registers are up-to-date - seq here means a "version number" of the
registers).
If the ptrace proxy is not enabled, debugregs_seq always stays 0 and
update_debugregs will be a no-op. So, optimize this out (the compiler can't
do it).
Also, I've been disappointed by the fact that it would make a lot of sense if,
after calling a successful
update_debugregs(current->thread.arch.debugregs_seq),
current->thread.arch.debugregs_seq were updated with the new debugregs_seq.
But this is not done. Is this a bug or a feature? For all purposes, it seems
a bug (otherwise the whole mechanism does not make sense, which is also a
possibility to check), which causes some performance only problems (not
correctness), since we write_debugregs when not needed.
Also, as suggested by Jeff, remove a redundant enabling of SIGVTALRM,
comprised in the subsequent local_irq_enable(). I'm just a bit dubious if
ordering matters there...
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
misc sparse annotations
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
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>
kills unmap magic
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
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>
kills symlinks in arch/um/sys-*
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
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>
This fixes a process segfault where a signal was being delivered such that a
new stack page needed to be allocated to hold the signal frame. This was
tripping some logic in the page fault handler which wouldn't allocate the page
if the faulting address was more that 32 bytes lower than the current stack
pointer. Since a signal frame is greater than 32 bytes, this exercised that
case.
It's fixed by updating the SP in the pt_regs before starting to copy the
signal frame. Since those are the registers that will be copied on to the
stack, we have to be careful to put the original SP, not the new one which
points to the signal frame, on the stack.
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>
For security reasons, UML in is_syscall() needs to have access to code in
vsyscall-page. The current implementation grants this access by explicitly
allowing access to vsyscall in access_ok_skas(). With this change,
copy_from_user() may be used to read the code. Ptrace access to vsyscall-page
for debugging already was implemented in get_user_pages() by mainline. In
i386, copy_from_user can't access vsyscall-page, but returns EFAULT.
To make UML behave as i386 does, I changed is_syscall to use
access_process_vm(current) to read the code from vsyscall-page. This doesn't
hurt security, but simplifies the code and prepares implementation of
stub-vmas.
Signed-off-by: Bodo Stroesser <bstroesser@fujitsu-siemens.com>
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 serial UML OS-abstraction layer patch (um/kernel dir).
This moves all systemcalls from irq_user.c file under os-Linux dir
Signed-off-by: Gennady Sharapov <Gennady.V.Sharapov@intel.com>
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>
To avoid conflicts, in kernel files errno is expanded to kernel_errno, to
distinguish it from glibc errno. In this case, the code wants to use the libc
errno but the kernel one is used; in the other usage, we return errno in place
of -errno in case of an error.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
CC arch/um/sys-i386/ldt.o
arch/um/sys-i386/ldt.c:19:21: proc_mm.h: No such file or directory
make[1]: *** [arch/um/sys-i386/ldt.o] Error 1
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
s390 doesn't have a LDT. So MM_COPY_SEGMENTS will not be supported on s390.
The only user of MM_COPY_SEGMENTS is new_mm(), but that's no longer useful, as
arch/sys-i386/ldt.c defines init_new_ldt(), which is called immediately after
new_mm(). So we should copy host's LDT in init_new_ldt(), if /proc/mm is
available, to have this subarch specific call in subarch code.
Signed-off-by: Bodo Stroesser <bstroesser@fujitsu-siemens.com>
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>
2.6.15-mm1 caused kernel-offsets.c to stop compiling with a syntax error in a
header. The problem was with KBUILD_BASENAME, which didn't get a definition
with the by-hand compilation in the main UML Makefile.
This was OK before since the expansion was syntactically the same as the
KBUILD_BASENAME token. With -mm1, the expansion is now a quote-delimited
string, so there needs to be a definition of it.
Since kernel-offsets.c is basically the same as other arches' asm-offsets.c,
and those seem to build OK, this patch turns kernel-offsets.c into
asm-offsets.c. kernel-offsets.c is in arch/um/sys-$(SUBARCH), i.e. sys-i386
and sys-x86_64, while kbuild expects it to be in arch/um/kernel.
kernel-offsets.c is moved to
arch/um/include/sysdep-$(SUBARCH)/kernel-offsets.h, which is included by
arch/um/kernel/asm-offsets.c. With that, include/asm-um/asm-offsets.h is
generated automatically. kernel-offsets.h continues to exist because it needs
to be accessible to userspace UML code, and include/asm-um isn't. So, a
symlink is made from arch/um/include/kernel-offsets.h to
include/asm-um/asm-offsets.h.
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 signal_user.c file under os-Linux dir
Signed-off-by: Gennady Sharapov <Gennady.V.Sharapov@intel.com>
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>
Fix UML compilation when SKAS mode is disabled. Indeed, we were compiling
SKAS-only object files, which failed due to some SKAS-only headers being
excluded from the search path.
Thanks to the bug report from Pekka J Enberg.
Acked-by: Pekka J Enberg <penberg (at) cs ! helsinki ! fi>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch makes stub_segv use the stub_syscall macros. This was needed
anyway, but the bug that prompted this was the discovery that gcc was storing
stuff in RCX, which is trashed across a system call. This is exactly the sort
of problem that the new macros fix.
There is a stub_syscall0 for getpid. stub_segv was changed to be a libc file,
and that caused some include changes.
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 gives a name to the anonymous union introduced in skas-hold-own-ldt,
allowing to build on a wider range of gccs.
It also removes ldt.h, which somehow became real, and replaces it with a
symlink, and creates ldt-x86_64.h as a copy of ldt-i386.h for now.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Patch imlements full LDT handling in SKAS:
* UML holds it's own LDT table, used to deliver data on
modify_ldt(READ)
* UML disables the default_ldt, inherited from the host (SKAS3)
or resets LDT entries, set by host's clib and inherited in
SKAS0
* A new global variable skas_needs_stub is inserted, that
can be used to decide, whether stub-pages must be supported
or not.
* Uses the syscall-stub to replace missing PTRACE_LDT (therefore,
write_ldt_entry needs to be modified)
Signed-off-by: Bodo Stroesser <bstroesser@fujitsu-siemens.com>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 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>
Al's build tidying missed one bit from me - without this UML doesn't boot.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Acked-by: Al Viro <viro@ftp.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
UML makefiles sanitized:
- number of generated headers reduced to 2 (from user-offsets.c and
kernel-offsets.c resp.). The rest is made constant and simply
includes those two.
- mk_... helpers are gone now that we don't need to generate these
headers
- arch/um/include2 removed since everything under arch/um/include/sysdep
is constant now and symlink can point straight to source tree.
- dependencies seriously simplified.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I am a lamer :-(. Luckily, Luo Xin performed LTP testing and found this failure.
Btw, the fact that the patch in which I introduced this was merged shows that:
a) I'm really trusted by people
b) sometimes they're wrong about point a).
c) lack of time for reviewers.
CC: Luo Xin <luothing@sina.com>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
elf_aux is userland code; it uses symbol (ELF_CLASS) that doesn't exist in
userland headers; pulled into kernel-offsets.h, switched elf_aux to using it.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Change syscall-stub's data to include a "expected retval".
Stub now checks syscalls retval and aborts execution of syscall list, if
retval != expected retval.
run_syscall_stub prints the data of the failed syscall, using the data pointer
and retval written by the stub to the beginning of the stack.
one_syscall_stub is removed, to simplify code, because only some instructions
are saved by one_syscall_stub, no host-syscall.
Using the stub with additional data (modify_ldt via stub)
is prepared also.
Signed-off-by: Bodo Stroesser <bstroesser@fujitsu-siemens.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>
Added missing include list to uml AFLAGS
Killed magic for stubs. [So] - it was needed only because of messed AFLAGS
Switched segv_stubs.c to kernel CFLAGS sans profile, instead of user ones
Killed STUBS_CFLAGS - it's not needed and the only remaining use had been
gratitious - it only polluted CFLAGS
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
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 adds VM op batching to skas0. Rather than having a context switch to and
from the userspace stub for each address space change, we write a number of
operations to the stub data page and invoke a different stub which loops over
them and executes them all in one go.
The operations are stored as [ system call number, arg1, arg2, ... ] tuples.
The set is terminated by a system call number of 0. Single operations, i.e.
page faults, are handled in the old way, since that is slightly more
efficient.
For a kernel build, a minority (~1/4) of the operations are part of a set.
These sets averaged ~100 in length, so for this quarter, the context switching
overhead is greatly reduced.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Build cleanups
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The copy_user stuff in the signal frame code was broke.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
It's wrong to pop a fixed number of words from stack before calling sigreturn,
as the number depends on what code is generated by the compiler for the start
of stub_segv_handler(). What we need is esp containing the address of
sigcontext. So we explicitly load that pointer into esp.
Signed-off-by: Bodo Stroesser <bstroesser@fujitsu-siemens.com>
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>
*) Reorganize the two cases of sys_modify_ldt to share all the reasonably
common code.
*) Avoid memory allocation when unneeded (i.e. when we are writing and the
passed buffer size is known), thus not returning ENOMEM (which isn't
allowed for this syscall, even if there is no strict "specification").
*) Add copy_{from,to}_user to modify_ldt for TT mode.
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>
A big bug has been diagnosed on hosts running the SKAS patch and built with
CONFIG_REGPARM, due to some missing prevent_tail_call().
On these hosts, this workaround is needed to avoid triggering that bug,
because "to" is kept by GCC only in EBX, which is corrupted at the return of
mmap2().
Since to trigger this bug int 0x80 must be used when doing the call, it rarely
manifests itself, so I'd prefer to get this merged to workaround that host
bug, since it should cause no functional change. Still, you might prefer to
drop it, I'll leave this to you.
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>
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>
Fix the do_fork calling convention: normal arch pass the regs and the new sp
value to do_fork instead of NULL.
Currently the arch-independent code ignores these values, while the UML code
(actually it's copy_thread) gets the right values by itself.
With this patch, things are fixed up.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Size of pointer doesn't seem right, but maybe my solution isn't either
(sig_size maybe?).
Signed-off-by: Domen Puncer <domen@coderock.org>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
With Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
To make sure switcheroo() can execute when we remap all the executable
image, we used a trick to make it use a local copy of errno... this trick
does not work with NPTL glibc, only with LinuxThreads, so use another
(simpler) one to make it work anyway.
Hopefully, a lot improved thanks to merging with the version of Al Viro
(which had his part of problems, though, i.e. removing a fix to another
bug and not fixing the problem on i386).
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>
Copy (and adapt) to UML the stack code dumper used in i386 when
CONFIG_FRAME_POINTER is enabled.
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 patch cleans up the delay implementations a bit, makes the loops
unoptimizable, and exports __udelay and __const_udelay.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Some changes that I sent in didn't make 2.6.12-rc4 for some reason. This
adds them back. We have
an x86_64 definition of TOP_ADDR
a reimplementation of the x86_64 csum_partial_copy_from_user
some syntax fixes in arch/um/kernel/ptrace.c
removal of a CFLAGS definition in the x86_64 Makefile
some include changes in the x86_64 ptrace.c and user-offsets.h
a syntax fix in elf-x86_64.h
Also moved an include in the i386 and x86_64 Makefiles to make the symlinks
work, and some small fixes from Al Viro.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
s390 needs to change some parts of arch/um/kernel/ptrace.c. Thus, the code
regarding PEEKUSER and POKEUSER are shifted to arch/um/sys-<subarch>/ptrace.c.
Also s390 debug registers need to be updated, when singlestepping is switched
on / off. Thus, setting/resetting of singlestepping is centralized in the new
function set_singlestep(), which also inserts the macro
SUBARCH_SET_SINGLESTEP(mode), if defined.
Finally, s390 has the "ieee_instruction_pointer" in its
registers, which also is allowed to be read via
ptrace( PTRACE_PEEKUSER, getpid(), PT_IEEE_IP, 0);
To implement this feature, sys_ptrace inserts the macro
SUBARCH_PTRACE_SPECIAL, if defined.
Signed-off-by: Bodo Stroesser <bstroesser@fujitsu-siemens.com>
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>
Al Viro