- 12 Jul, 2017 40 commits
-
-
Rik van Riel authored
Use the ascii-armor canary to prevent unterminated C string overflows from being able to successfully overwrite the canary, even if they somehow obtain the canary value. Inspired by execshield ascii-armor and Daniel Micay's linux-hardened tree. Link: http://lkml.kernel.org/r/20170524155751.424-4-riel@redhat.comSigned-off-by:
Rik van Riel <riel@redhat.com> Acked-by:
Kees Cook <keescook@chromium.org> Cc: Daniel Micay <danielmicay@gmail.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Ingo Molnar <mingo@kernel.org> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Yoshinori Sato <ysato@users.sourceforge.jp> Signed-off-by:
Andrew Morton <akpm@linux-foundation.org> Signed-off-by:
Linus Torvalds <torvalds@linux-foundation.org>
-
Rik van Riel authored
Use the ascii-armor canary to prevent unterminated C string overflows from being able to successfully overwrite the canary, even if they somehow obtain the canary value. Inspired by execshield ascii-armor and Daniel Micay's linux-hardened tree. Link: http://lkml.kernel.org/r/20170524155751.424-3-riel@redhat.comSigned-off-by:
-
Rik van Riel authored
Patch series "stackprotector: ascii armor the stack canary", v2. Zero out the first byte of the stack canary value on 64 bit systems, in order to mitigate unterminated C string overflows. The null byte both prevents C string functions from reading the canary, and from writing it if the canary value were guessed or obtained through some other means. Reducing the entropy by 8 bits is acceptable on 64-bit systems, which will still have 56 bits of entropy left, but not on 32 bit systems, so the "ascii armor" canary is only implemented on 64-bit systems. Inspired by the "ascii armor" code in execshield and Daniel Micay's linux-hardened tree. Also see https://github.com/thestinger/linux-hardened/ This patch (of 5): Introduce get_random_canary(), which provides a random unsigned long canary value with the first byte zeroed out on 64 bit architectures, in order to mitigate non-terminated C string overflows. The null byte both prevents C string functions from reading the canary, and from writing it if the canary value were guessed or obtained through some other means. Reducing the entropy by 8 bits is acceptable on 64-bit systems, which will still have 56 bits of entropy left, but not on 32 bit systems, so the "ascii armor" canary is only implemented on 64-bit systems. Inspired by the "ascii armor" code in the old execshield patches, and Daniel Micay's linux-hardened tree. Link: http://lkml.kernel.org/r/20170524155751.424-2-riel@redhat.comSigned-off-by:
-
Kees Cook authored
When building the sh architecture, the compiler doesn't realize that BUG() doesn't return, so it will complain about functions using BUG() that are marked with the noreturn attribute: lib/string.c: In function 'fortify_panic': >> lib/string.c:986:1: warning: 'noreturn' function does return } ^ Link: http://lkml.kernel.org/r/20170627192050.GA66784@beastSigned-off-by: -
Daniel Micay authored
This adds support for compiling with a rough equivalent to the glibc _FORTIFY_SOURCE=1 feature, providing compile-time and runtime buffer overflow checks for string.h functions when the compiler determines the size of the source or destination buffer at compile-time. Unlike glibc, it covers buffer reads in addition to writes. GNU C __builtin_*_chk intrinsics are avoided because they would force a much more complex implementation. They aren't designed to detect read overflows and offer no real benefit when using an implementation based on inline checks. Inline checks don't add up to much code size and allow full use of the regular string intrinsics while avoiding the need for a bunch of _chk functions and per-arch assembly to avoid wrapper overhead. This detects various overflows at compile-time in various drivers and some non-x86 core kernel code. There will likely be issues caught in regular use at runtime too. Future improvements left out of initial implementation for simplicity, as it's all quite optional and can be done incrementally: * Some of the fortified string functions (strncpy, strcat), don't yet place a limit on reads from the source based on __builtin_object_size of the source buffer. * Extending coverage to more string functions like strlcat. * It should be possible to optionally use __builtin_object_size(x, 1) for some functions (C strings) to detect intra-object overflows (like glibc's _FORTIFY_SOURCE=2), but for now this takes the conservative approach to avoid likely compatibility issues. * The compile-time checks should be made available via a separate config option which can be enabled by default (or always enabled) once enough time has passed to get the issues it catches fixed. Kees said: "This is great to have. While it was out-of-tree code, it would have blocked at least CVE-2016-3858 from being exploitable (improper size argument to strlcpy()). I've sent a number of fixes for out-of-bounds-reads that this detected upstream already" [arnd@arndb.de: x86: fix fortified memcpy] Link: http://lkml.kernel.org/r/20170627150047.660360-1-arnd@arndb.de [keescook@chromium.org: avoid panic() in favor of BUG()] Link: http://lkml.kernel.org/r/20170626235122.GA25261@beast [keescook@chromium.org: move from -mm, add ARCH_HAS_FORTIFY_SOURCE, tweak Kconfig help] Link: http://lkml.kernel.org/r/20170526095404.20439-1-danielmicay@gmail.com Link: http://lkml.kernel.org/r/1497903987-21002-8-git-send-email-keescook@chromium.orgSigned-off-by:
Daniel Micay <danielmicay@gmail.com> Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Acked-by:
-
Daniel Axtens authored
Testing the fortified string functions[1] would cause a kernel panic on boot in test_feature_fixups() due to a buffer overflow in memcmp. This boils down to things like this: extern unsigned int ftr_fixup_test1; extern unsigned int ftr_fixup_test1_orig; check(memcmp(&ftr_fixup_test1, &ftr_fixup_test1_orig, size) == 0); We know that these are asm labels so it is safe to read up to 'size' bytes at those addresses. However, because we have passed the address of a single unsigned int to memcmp, the compiler believes the underlying object is in fact a single unsigned int. So if size > sizeof(unsigned int), there will be a panic at runtime. We can fix this by changing the types: instead of calling the asm labels unsigned ints, call them unsigned int[]s. Therefore the size isn't incorrectly determined at compile time and we get a regular unsafe memcmp and no panic. [1] http://openwall.com/lists/kernel-hardening/2017/05/09/2 Link: http://lkml.kernel.org/r/1497903987-21002-7-git-send-email-keescook@chromium.orgSigned-off-by:
Daniel Axtens <dja@axtens.net> Signed-off-by:
Michael Ellerman <mpe@ellerman.id.au> Tested-by:
Andrew Donnellan <andrew.donnellan@au1.ibm.com> Reviewed-by:
-
Daniel Axtens authored
prom_init is a bit special; in theory it should be able to be linked separately to the kernel. To keep this from getting too complex, the symbols that prom_init.c uses are checked. Fortification adds symbols, and it gets quite messy as it includes things like panic(). So just don't fortify prom_init.c for now. Link: http://lkml.kernel.org/r/1497903987-21002-6-git-send-email-keescook@chromium.orgSigned-off-by:
-
Kees Cook authored
This fixes a over-read condition detected by FORTIFY_SOURCE for this line: memcpy(SKB_TO_PKT(skb), &ack_pkt, sizeof(skb->cb)); The error was: In file included from ./include/linux/bitmap.h:8:0, from ./include/linux/cpumask.h:11, from ./include/linux/mm_types_task.h:13, from ./include/linux/mm_types.h:4, from ./include/linux/kmemcheck.h:4, from ./include/linux/skbuff.h:18, from drivers/infiniband/sw/rxe/rxe_resp.c:34: In function 'memcpy', inlined from 'send_atomic_ack.constprop' at drivers/infiniband/sw/rxe/rxe_resp.c:998:2, inlined from 'acknowledge' at drivers/infiniband/sw/rxe/rxe_resp.c:1026:3, inlined from 'rxe_responder' at drivers/infiniband/sw/rxe/rxe_resp.c:1286:10: ./include/linux/string.h:309:4: error: call to '__read_overflow2' declared with attribute error: detected read beyond size of object passed as 2nd parameter __read_overflow2(); Daniel Micay noted that struct rxe_pkt_info is 32 bytes on 32-bit architectures, but skb->cb is still 64. The memcpy() over-reads 32 bytes. This fixes it by zeroing the unused bytes in skb->cb. Link: http://lkml.kernel.org/r/1497903987-21002-5-git-send-email-keescook@chromium.orgSigned-off-by: -
Kees Cook authored
Defining kexec_purgatory as a zero-length char array upsets compile time size checking. Since this is built on a per-arch basis, define it as an unsized char array (like is done for other similar things, e.g. linker sections). This silences the warning generated by the future CONFIG_FORTIFY_SOURCE, which did not like the memcmp() of a "0 byte" array. This drops the __weak and uses an extern instead, since both users define kexec_purgatory. Link: http://lkml.kernel.org/r/1497903987-21002-4-git-send-email-keescook@chromium.orgSigned-off-by:
"Eric W. Biederman" <ebiederm@xmission.com> Cc: Daniel Micay <danielmicay@gmail.com> Signed-off-by:
-
Kees Cook authored
This avoids CONFIG_FORTIFY_SOURCE from being enabled during the EFI stub build, as adding a panic() implementation may not work well. This can be adjusted in the future. Link: http://lkml.kernel.org/r/1497903987-21002-2-git-send-email-keescook@chromium.orgSigned-off-by:
Ard Biesheuvel <ard.biesheuvel@linaro.org> Acked-by:
Mark Rutland <mark.rutland@arm.com> Cc: Matt Fleming <matt@codeblueprint.co.uk> Signed-off-by:
-
Nicholas Piggin authored
Implement an arch-speicfic watchdog rather than use the perf-based hardlockup detector. The new watchdog takes the soft-NMI directly, rather than going through perf. Perf interrupts are to be made maskable in future, so that would prevent the perf detector from working in those regions. Additionally, implement a SMP based detector where all CPUs watch one another by pinging a shared cpumask. This is because powerpc Book3S does not have a true periodic local NMI, but some platforms do implement a true NMI IPI. If a CPU is stuck with interrupts hard disabled, the soft-NMI watchdog does not work, but the SMP watchdog will. Even on platforms without a true NMI IPI to get a good trace from the stuck CPU, other CPUs will notice the lockup sufficiently to report it and panic. [npiggin@gmail.com: honor watchdog disable at boot/hotplug] Link: http://lkml.kernel.org/r/20170621001346.5bb337c9@roar.ozlabs.ibm.com [npiggin@gmail.com: fix false positive warning at CPU unplug] Link: http://lkml.kernel.org/r/20170630080740.20766-1-npiggin@gmail.com [akpm@linux-foundation.org: coding-style fixes] Link: http://lkml.kernel.org/r/20170616065715.18390-6-npiggin@gmail.comSigned-off-by:
Nicholas Piggin <npiggin@gmail.com> Reviewed-by:
Don Zickus <dzickus@redhat.com> Tested-by: Babu Moger <babu.moger@oracle.com> [sparc] Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by:
-
Nicholas Piggin authored
After reconfiguring watchdog sysctls etc., architecture specific watchdogs may not get all their parameters updated. watchdog_nmi_reconfigure() can be implemented to pull the new values in and set the arch NMI watchdog. [npiggin@gmail.com: add code comments] Link: http://lkml.kernel.org/r/20170617125933.774d3858@roar.ozlabs.ibm.com [arnd@arndb.de: hide unused function] Link: http://lkml.kernel.org/r/20170620204854.966601-1-arnd@arndb.de Link: http://lkml.kernel.org/r/20170616065715.18390-5-npiggin@gmail.comSigned-off-by:
-
Nicholas Piggin authored
Split SOFTLOCKUP_DETECTOR from LOCKUP_DETECTOR, and split HARDLOCKUP_DETECTOR_PERF from HARDLOCKUP_DETECTOR. LOCKUP_DETECTOR implies the general boot, sysctl, and programming interfaces for the lockup detectors. An architecture that wants to use a hard lockup detector must define HAVE_HARDLOCKUP_DETECTOR_PERF or HAVE_HARDLOCKUP_DETECTOR_ARCH. Alternatively an arch can define HAVE_NMI_WATCHDOG, which provides the minimum arch_touch_nmi_watchdog, and it otherwise does its own thing and does not implement the LOCKUP_DETECTOR interfaces. sparc is unusual in that it has started to implement some of the interfaces, but not fully yet. It should probably be converted to a full HAVE_HARDLOCKUP_DETECTOR_ARCH. [npiggin@gmail.com: fix] Link: http://lkml.kernel.org/r/20170617223522.66c0ad88@roar.ozlabs.ibm.com Link: http://lkml.kernel.org/r/20170616065715.18390-4-npiggin@gmail.comSigned-off-by:
Babu Moger <babu.moger@oracle.com> Tested-by: Babu Moger <babu.moger@oracle.com> [sparc] Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by:
-
Nicholas Piggin authored
For architectures that define HAVE_NMI_WATCHDOG, instead of having them provide the complete touch_nmi_watchdog() function, just have them provide arch_touch_nmi_watchdog(). This gives the generic code more flexibility in implementing this function, and arch implementations don't miss out on touching the softlockup watchdog or other generic details. Link: http://lkml.kernel.org/r/20170616065715.18390-3-npiggin@gmail.comSigned-off-by:
-
Nicholas Piggin authored
Patch series "Improve watchdog config for arch watchdogs", v4. A series to make the hardlockup watchdog more easily replaceable by arch code. The last patch provides some justification for why we want to do this (existing sparc watchdog is another that could benefit). This patch (of 5): Remove unused declaration. Link: http://lkml.kernel.org/r/20170616065715.18390-2-npiggin@gmail.comSigned-off-by:
-
Michal Hocko authored
xt_alloc_table_info() basically opencodes kvmalloc() so use the library function instead. Link: http://lkml.kernel.org/r/20170531155145.17111-4-mhocko@kernel.orgSigned-off-by:
Michal Hocko <mhocko@suse.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu> Cc: Florian Westphal <fw@strlen.de> Signed-off-by:
-
Manfred Spraul authored
Now that ipc_rcu_alloc() and ipc_rcu_free() are removed, document when it is valid to use ipc_getref() and ipc_putref(). Link: http://lkml.kernel.org/r/20170525185107.12869-21-manfred@colorfullife.comSigned-off-by:
Manfred Spraul <manfred@colorfullife.com> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Kees Cook <keescook@chromium.org> Signed-off-by:
-
Kees Cook authored
The remaining users of __sem_free() can simply call kvfree() instead for better readability. [manfred@colorfullife.com: Rediff to keep rcu protection for security_sem_alloc()] Link: http://lkml.kernel.org/r/20170525185107.12869-20-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
There is nothing special about the msg_alloc/free routines any more, so remove them to make code more readable. [manfred@colorfullife.com: Rediff to keep rcu protection for security_msg_queue_alloc()] Link: http://lkml.kernel.org/r/20170525185107.12869-19-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
There is nothing special about the shm_alloc/free routines any more, so remove them to make code more readable. [manfred@colorfullife.com: Rediff, to continue to keep rcu for free calls after a successful security_shm_alloc()] Link: http://lkml.kernel.org/r/20170525185107.12869-18-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
Only after ipc_addid() has succeeded will refcounting be used, so move initialization into ipc_addid() and remove from open-coded *_alloc() routines. Link: http://lkml.kernel.org/r/20170525185107.12869-17-manfred@colorfullife.comSigned-off-by:
-
Manfred Spraul authored
Loosely based on a patch from Kees Cook <keescook@chromium.org>: - id and retval can be merged - if ipc_addid() fails, then use call_rcu() directly. The difference is that call_rcu is used for failed ipc_addid() calls, to continue to guaranteed an rcu delay for security_msg_queue_free(). Link: http://lkml.kernel.org/r/20170525185107.12869-16-manfred@colorfullife.comSigned-off-by:
-
Manfred Spraul authored
Loosely based on a patch from Kees Cook <keescook@chromium.org>: - id and error can be merged - if operations before ipc_addid() fail, then use call_rcu() directly. The difference is that call_rcu is used for failures after security_shm_alloc(), to continue to guaranteed an rcu delay for security_sem_free(). Link: http://lkml.kernel.org/r/20170525185107.12869-15-manfred@colorfullife.comSigned-off-by:
-
Manfred Spraul authored
Loosely based on a patch from Kees Cook <keescook@chromium.org>: - id and retval can be merged - if ipc_addid() fails, then use call_rcu() directly. The difference is that call_rcu is used for failed ipc_addid() calls, to continue to guaranteed an rcu delay for security_sem_free(). Link: http://lkml.kernel.org/r/20170525185107.12869-14-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
No callers remain for ipc_rcu_alloc(). Drop the function. [manfred@colorfullife.com: Rediff because the memset was temporarily inside ipc_rcu_free()] Link: http://lkml.kernel.org/r/20170525185107.12869-13-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
Instead of using ipc_rcu_alloc() which only performs the refcount bump, open code it. This also allows for msg_queue structure layout to be randomized in the future. Link: http://lkml.kernel.org/r/20170525185107.12869-12-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
Instead of using ipc_rcu_alloc() which only performs the refcount bump, open code it. This also allows for shmid_kernel structure layout to be randomized in the future. Link: http://lkml.kernel.org/r/20170525185107.12869-11-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
Instead of using ipc_rcu_alloc() which only performs the refcount bump, open code it to perform better sem-specific checks. This also allows for sem_array structure layout to be randomized in the future. [manfred@colorfullife.com: Rediff, because the memset was temporarily inside ipc_rcu_alloc()] Link: http://lkml.kernel.org/r/20170525185107.12869-10-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
There are no more callers of ipc_rcu_free(), so remove it. Link: http://lkml.kernel.org/r/20170525185107.12869-9-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
Avoid using ipc_rcu_free, since it just re-finds the original structure pointer. For the pre-list-init failure path, there is no RCU needed, since it was just allocated. It can be directly freed. Link: http://lkml.kernel.org/r/20170525185107.12869-8-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
Avoid using ipc_rcu_free, since it just re-finds the original structure pointer. For the pre-list-init failure path, there is no RCU needed, since it was just allocated. It can be directly freed. Link: http://lkml.kernel.org/r/20170525185107.12869-7-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
Avoid using ipc_rcu_free, since it just re-finds the original structure pointer. For the pre-list-init failure path, there is no RCU needed, since it was just allocated. It can be directly freed. Link: http://lkml.kernel.org/r/20170525185107.12869-6-manfred@colorfullife.comSigned-off-by:
-
Kees Cook authored
The only users of ipc_alloc() were ipc_rcu_alloc() and the on-heap sem_io fall-back memory. Better to just open-code these to make things easier to read. [manfred@colorfullife.com: Rediff due to inclusion of memset() into ipc_rcu_alloc()] Link: http://lkml.kernel.org/r/20170525185107.12869-5-manfred@colorfullife.comSigned-off-by:
-
Manfred Spraul authored
sem_ctime is initialized to the semget() time and then updated at every semctl() that changes the array. Thus it does not represent the time of the last change. Especially, semop() calls are only stored in sem_otime, not in sem_ctime. This is already described in ipc/sem.c, I just overlooked that there is a comment in include/linux/sem.h and man semctl(2) as well. So: Correct wrong comments. Link: http://lkml.kernel.org/r/20170515171912.6298-4-manfred@colorfullife.comSigned-off-by:
-
Manfred Spraul authored
ipc has two management structures that exist for every id: - struct kern_ipc_perm, it contains e.g. the permissions. - struct ipc_rcu, it contains the rcu head for rcu handling and the refcount. The patch merges both structures. As a bonus, we may save one cacheline, because both structures are cacheline aligned. In addition, it reduces the number of casts, instead most codepaths can use container_of. To simplify code, the ipc_rcu_alloc initializes the allocation to 0. [manfred@colorfullife.com: really include the memset() into ipc_alloc_rcu()] Link: http://lkml.kernel.org/r/564f8612-0601-b267-514f-a9f650ec9b32@colorfullife.com Link: http://lkml.kernel.org/r/20170525185107.12869-3-manfred@colorfullife.comSigned-off-by:
-
Manfred Spraul authored
sma->sem_base is initialized with sma->sem_base = (struct sem *) &sma[1]; The current code has four problems: - There is an unnecessary pointer dereference - sem_base is not needed. - Alignment for struct sem only works by chance. - The current code causes false positive for static code analysis. - This is a cast between different non-void types, which the future randstruct GCC plugin warns on. And, as bonus, the code size gets smaller: Before: 0 .text 00003770 After: 0 .text 0000374e [manfred@colorfullife.com: s/[0]/[]/, per hch] Link: http://lkml.kernel.org/r/20170525185107.12869-2-manfred@colorfullife.com Link: http://lkml.kernel.org/r/20170515171912.6298-2-manfred@colorfullife.comSigned-off-by: -
Dmitry Vyukov authored
Add /proc/self/task/<current-tid>/fail-nth file that allows failing 0-th, 1-st, 2-nd and so on calls systematically. Excerpt from the added documentation: "Write to this file of integer N makes N-th call in the current task fail (N is 0-based). Read from this file returns a single char 'Y' or 'N' that says if the fault setup with a previous write to this file was injected or not, and disables the fault if it wasn't yet injected. Note that this file enables all types of faults (slab, futex, etc). This setting takes precedence over all other generic settings like probability, interval, times, etc. But per-capability settings (e.g. fail_futex/ignore-private) take precedence over it. This feature is intended for systematic testing of faults in a single system call. See an example below" Why add a new setting: 1. Existing settings are global rather than per-task. So parallel testing is not possible. 2. attr->interval is close but it depends on attr->count which is non reset to 0, so interval does not work as expected. 3. Trying to model this with existing settings requires manipulations of all of probability, interval, times, space, task-filter and unexposed count and per-task make-it-fail files. 4. Existing settings are per-failure-type, and the set of failure types is potentially expanding. 5. make-it-fail can't be changed by unprivileged user and aggressive stress testing better be done from an unprivileged user. Similarly, this would require opening the debugfs files to the unprivileged user, as he would need to reopen at least times file (not possible to pre-open before dropping privs). The proposed interface solves all of the above (see the example). We want to integrate this into syzkaller fuzzer. A prototype has found 10 bugs in kernel in first day of usage: https://groups.google.com/forum/#!searchin/syzkaller/%22FAULT_INJECTION%22%7Csort:relevance I've made the current interface work with all types of our sandboxes. For setuid the secret sauce was prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) to make /proc entries non-root owned. So I am fine with the current version of the code. [akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/20170328130128.101773-1-dvyukov@google.comSigned-off-by:
Dmitry Vyukov <dvyukov@google.com> Cc: Akinobu Mita <akinobu.mita@gmail.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by:
-
Cyrill Gorcunov authored
kcmp syscall is build iif CONFIG_CHECKPOINT_RESTORE is selected, so wrap appropriate helpers in epoll code with the config to build it conditionally. Link: http://lkml.kernel.org/r/20170513083456.GG1881@uranus.lanSigned-off-by:
Cyrill Gorcunov <gorcunov@openvz.org> Reported-by:
Andrew Morton <akpm@linuxfoundation.org> Cc: Andrey Vagin <avagin@openvz.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Michael Kerrisk <mtk.manpages@gmail.com> Cc: Jason Baron <jbaron@akamai.com> Cc: Andy Lutomirski <luto@amacapital.net> Signed-off-by:
-
Cyrill Gorcunov authored
With current epoll architecture target files are addressed with file_struct and file descriptor number, where the last is not unique. Moreover files can be transferred from another process via unix socket, added into queue and closed then so we won't find this descriptor in the task fdinfo list. Thus to checkpoint and restore such processes CRIU needs to find out where exactly the target file is present to add it into epoll queue. For this sake one can use kcmp call where some particular target file from the queue is compared with arbitrary file passed as an argument. Because epoll target files can have same file descriptor number but different file_struct a caller should explicitly specify the offset within. To test if some particular file is matching entry inside epoll one have to - fill kcmp_epoll_slot structure with epoll file descriptor, target file number and target file offset (in case if only one target is present then it should be 0) - call kcmp as kcmp(pid1, pid2, KCMP_EPOLL_TFD, fd, &kcmp_epoll_slot) - the kernel fetch file pointer matching file descriptor @fd of pid1 - lookups for file struct in epoll queue of pid2 and returns traditional 0,1,2 result for sorting purpose Link: http://lkml.kernel.org/r/20170424154423.511592110@gmail.comSigned-off-by:Andrey Vagin <avagin@openvz.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Michael Kerrisk <mtk.manpages@gmail.com> Cc: Jason Baron <jbaron@akamai.com> Cc: Andy Lutomirski <luto@amacapital.net> Signed-off-by:
-
Cyrill Gorcunov authored
Since it is possbile to have same number in tfd field (say file added, closed, then nother file dup'ed to same number and added back) it is imposible to distinguish such target files solely by their numbers. Strictly speaking regular applications don't need to recognize these targets at all but for checkpoint/restore sake we need to collect targets to be able to push them back on restore stage in a proper order. Thus lets add file position, inode and device number where this target lays. This three fields can be used as a primary key for sorting, and together with kcmp help CRIU can find out an exact file target (from the whole set of processes being checkpointed). Link: http://lkml.kernel.org/r/20170424154423.436491881@gmail.comSigned-off-by:
Andrei Vagin <avagin@virtuozzo.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Cc: Michael Kerrisk <mtk.manpages@gmail.com> Cc: Jason Baron <jbaron@akamai.com> Cc: Andy Lutomirski <luto@amacapital.net> Signed-off-by:
-
