Task size depends from VM_SPLIT_* kernel option and cannot be hard coded.
This patch based on c0c0546c31 from
Christopher Covington.
Signed-off-by: Artem Kuzmitskiy <artem.kuzmitskiy@lge.com>
Reviewed-by: Christopher Covington <cov@codeaurora.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
If we want one CRIU binary to work across all AArch64 kernel
configurations, a single task size value cannot be hard coded.
Signed-off-by: Christopher Covington <cov@codeaurora.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
In AArch64, pages may be 4K or 64K depending on kernel configuration.
The GNU C Library documentation suggests [1], "the correct interface
to query about the page size is sysconf". Introduce one new
architecture-specific function-like macro, page_size(), that on x86
and AArch32 remains a constant so as to minimally affect performance,
but on AArch64 is sysconf(_SC_PAGESIZE) for correctness.
1. https://www.gnu.org/software/libc/manual/html_node/Query-Memory-Parameters.html
To minimize churn, the PAGE_SIZE macro is left as a build-time
estimation of what the run-time page size might be.
This fixes the following errors for CRIU on AArch64 kernels with
CONFIG_ARM64_64K_PAGES=y, allowing dump of
`setsid sleep < /dev/null &> /dev/null` to succeed.
Error (kerndat.c:48): Can't stat self map_files: No such file or directory
Error (util.c:668): Can't read pme for pid 90: No such file or directory
Error (parasite-syscall.c:1135): Can't open 89/map_files/0x3ffb7da0000-0x3ffb7dac000 on procfs: No such file or directory
Signed-off-by: Christopher Covington <cov@codeaurora.org>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Cc: Ruslan Kuprieiev <kupruser@gmail.com>
Signed-off-by: Andrew Vagin <avagin@openvz.org>
Tested-by: Ruslan Kuprieiev <kupruser@gmail.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Unfortunately the kernel doesn't flush hw breakpoints on
detaching ptrace. If a breakpoint is triggered without ptrace, it
will be killed by SIGTRAP.
Reported-by: Mr Jenkins
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Currently CRIU traces syscalls to catch a moment, when sigreturn() is
called. Now we trace recv(cmd), close(logfd), close(cmdfd), sigreturn().
We can reduce a number of steps by using hw breakpoints. A breakpoint is
set before sigreturn, so we will need to trace only it.
v2: In the first version a breakpoint is set after sigreturn. In this
case we have a problem with signals. If a process has pending signals,
it will start to precess them after exiting from sigreturn(), but before
returning to userspace. So the breakpoint will not be triggered.
And at the end Here are a few numbers how we catch sigreturn.
Before this patch criu executes 36 syscalls and gets 12 signals.
With this patch criu executes 18 syscalls and gets 5 signals.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Currently we build vDSO handling code for all archs provided
in the source code having some "common" parts inside pie/vdso.c,
pie/vdso-stub.c, vdso-stub.c and vdso.c. This were more or
less well but in new linux kernels (starting from 3.16 presumably)
the vDSO has been significantly reworked so every architecture
must have own vDSO handling engine (just like the kernel does).
So in this patch we move vDSO code to arch specific and because
aarch64 actually doesn't implement proxification yet due to
kernel restrictions -- we drops it out. When there will be
kernel support we bring it back in proper arch/aarch64
implementation.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This patch moves the files arch/$ARCH/include/asm/int.h to
include/asm-generic/int.h and makes the types {u,s}{8,16,32}
be aliases of the fixed sized integer types [u]int{8,16,32}_t.
This makes it possible to use single set of integer typedefs
in all architectural ports.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This patch splits the file arch/x86/vdso-pie.c into machine-dependent
and machine-independent parts by moving the routines vdso_fill_symtable(),
vdso_proxify(), and vdso_remap() to the file pie/vdso.c.
The ARM version of the routines is moved to the source pie/vdso-stub.c
to provide the vDSO proxy stub implementation for architectures
that don't provide the vDSO.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Looks-good-to: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Christopher Covington <cov@codeaurora.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This patch moves the enum VDSO_SYMBOL_* and macros VDSO_SYMBOL_*_NAME
to the x86 specific header since different architectures export
different symbols from their vDSOs.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Looks-good-to: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Christopher Covington <cov@codeaurora.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Supported machine architectures provide TLS stogares of different sizes:
the size of the TLS storage in x86-64 is 24 bytes, ARM --- 4 bytes
and upcoming AArch64 --- 8 bytes. This means every supported architecture
needs a specific type to store the value of the TLS register.
This patch reworks the insterface of the routines arch_get_tls()
and restore_tls() passing them the TLS storage by pointer
rather than by value to simplify the TLS stub for x86.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Christopher Covington <cov@codeaurora.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The implementation of bit operations for ARM isn't actually
architecture-specific so it would rather be shared with
the upcoming port for AArch64 that won't provide optimized
implementation of bit operations.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Reviewed-by: Christopher Covington <cov@codeaurora.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This patch should simplify the routine interface change
in the following patch.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Reviewed-by: Christopher Covington <cov@codeaurora.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
At the moment we are using 4K pages all the time,
so instead of copying code over all archs we're
supporting -- add asm-generic/page.h header.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Without this a user may encounter the following error.
arch/arm/include/asm/atomic.h:38:2: error: implicit declaration of function ‘smp_mb’
With the #error pragma a user is a least pointed to the specific
section in the file that's causing the problem
arch/arm/include/asm/atomic.h:21:2: error: #error ARM architecture version (CONFIG_ARMV*) not set or unsupported.
Acked-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We have a mess of uintX_t and uX usage. Drop off uintX_t ones.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Andrew Vagin <avagin@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Remove space before tab characters.
Found by git grep ' ' (Space, Ctrl-V, Tab in shell).
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Move prctl constants into own header. They are
arch independent.
The x86 specific ARCH_SET_GS should be fetched from <asm/prctl.h>
(instead of defining own ones).
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We really don't need it spread over all headers. The file
handlers are used in fsnotify only, declare it there.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This limit is not exported by a kernel, so no need
to "explicitly" define it pretending that it's immutable.
Still I kept this name in comment just to give reader
a tip where CR_SCM_MAX_FD came from.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
fcntl data is arch independent, so move it out of include/asm/type.h
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
It has been a mistake to introduce this struct since the FPU state
is stored deep inside an ARM sigframe. A remnant of the struct
is left to make the code compatible with the universal declaration
of the routine sigreturn_prep_fpu_frame().
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The ARM VFP header is a part of the VFP frame that makes
it unreasonable to have frame initialization split into
two routines. This patch moerges the ARM version of
the routine sigreturn_prep_fpu_frame() into restore_fpu()
and reduces the former into a stub.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
An fpu_state_t instance shouldn't be stored in an ARM sigframe
that makes it necessary to generalize the way the FPU state is
accessed in the routine construct_sigframe()
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Use same code as provided in kernel. In first place
we used own prototypes in case of simplicity (they
all were based on "lock xadd" instruction. There is
no more need for that and we can switch to well known
kernel's api.
Because kernel uses plain int type to carry atomic
counters I had to add explicit u32 type for futexes,
as well as a couple of fixes for new api usage.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This patch modifies the ARM syscall glue as follows:
* the macros syscallX that produce different code
for syscalls with different number of arguments
are replaced by the generic syscall thunk copied
from libc;
* the syscall tables generation script is simplified
since the syscall argument number analysis is redundant;
* the ARMv7 specific instruction MOVW is eliminated.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
A struct sigframe* instead of fpu_state_t should be passed
to the routine restore_fpu() since FPU registers are stored
in different fields of the sigframe in different architectures.
An architecture-specific implementation of the routine restore_fpu()
should know details of this layout instead of construct_sigframe().
This change makes it possible to move ARM FPU restoration
from sigreturn_prep_fpu_frame() (where it caused a segfault
since the pointer fpu_stat has become invalid in the dumper
address space) to restore_fpu()
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Acked-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The names of these functions should be swapped to avoid confusion.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Acked-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The macro REG_SYSCALL_NR is designated to get the value of the CPU register
storing syscall number in a machine-independent way.
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Acked-by: Andrew Vagin <avagin@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
parasite_fixup_vdso is not worth calling on arm at all,
moreover it might try to read pfn for addresses greater
than TASK_SIZE, leading to dump interruption. Lets make
it arch dependent.
[alekskartashov@: Fix compiling on arm]
Reported-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Tested-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This patch adds macro for getting IP register which differs according to
architecture.
Signed-off-by: Chanho Park <chanho61.park@samsung.com>
Signed-off-by: Myungjoo Ham <myungjoo.ham@samsung.com>
Acked-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
rhel6 compiler is not happy about preliminary
declaration of VmaEntry type yielding error on
building attempt, thus include protobuf header.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Otherwise a task can start to handle a signal and registers can be
changed.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
It will be used for executed parasite as a daemon.
What we have previously -- the stack has been preallocated in parasite
blob itself and bootstrap procedure calculated the value needed for %rsp.
With this patch applied we provide every thread own stack as:
- find out how many threads are present
- calculate the summary size of all stacks
- when we ask dumpee to provide us memory area needed to run
parasite code, we pass summary size needed for everything
- when parasite code is asked to run we calculate %rsp needed
taking into account the thread number (ie offsets) and then
setup proper %rsp via ptrace call, instead of calculating it
in bootstrap parasite code
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
When parasite daemon mode will be implemented we get deprived of ability
to fetch registers at the late moment of dumping as we were, thus just
bind CoreEntry to pstree item and allocate CoreEntry'ies for every
thread found, once process tree is in seized state.
Then immediately fill CoreEntry'ies with registers. We use prctl
opcode for that but fetch a complete set of registers including
FPU state, and convert them into protobuf format.
Zombie tasks remains untouched, we allocate CoreEntry for them
right at moment of dumping becuase we don't need registers there
to be written on disk.
This way get_task_regs no longer need parasite_ctl argument
and it's zapped.
Still parasite_ctl has own copy of general registers set but
this is because we need them to be in cpu native format unlike
ones kept in CoreEntry.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The first one fills sigframe and the second one restores another
registers.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
