From fcf634098c00dd9cd247447368495f0b79be12d1 Mon Sep 17 00:00:00 2001 From: Christopher Yeoh Date: Mon, 31 Oct 2011 17:06:39 -0700 Subject: Cross Memory Attach The basic idea behind cross memory attach is to allow MPI programs doing intra-node communication to do a single copy of the message rather than a double copy of the message via shared memory. The following patch attempts to achieve this by allowing a destination process, given an address and size from a source process, to copy memory directly from the source process into its own address space via a system call. There is also a symmetrical ability to copy from the current process's address space into a destination process's address space. - Use of /proc/pid/mem has been considered, but there are issues with using it: - Does not allow for specifying iovecs for both src and dest, assuming preadv or pwritev was implemented either the area read from or written to would need to be contiguous. - Currently mem_read allows only processes who are currently ptrace'ing the target and are still able to ptrace the target to read from the target. This check could possibly be moved to the open call, but its not clear exactly what race this restriction is stopping (reason appears to have been lost) - Having to send the fd of /proc/self/mem via SCM_RIGHTS on unix domain socket is a bit ugly from a userspace point of view, especially when you may have hundreds if not (eventually) thousands of processes that all need to do this with each other - Doesn't allow for some future use of the interface we would like to consider adding in the future (see below) - Interestingly reading from /proc/pid/mem currently actually involves two copies! (But this could be fixed pretty easily) As mentioned previously use of vmsplice instead was considered, but has problems. Since you need the reader and writer working co-operatively if the pipe is not drained then you block. Which requires some wrapping to do non blocking on the send side or polling on the receive. In all to all communication it requires ordering otherwise you can deadlock. And in the example of many MPI tasks writing to one MPI task vmsplice serialises the copying. There are some cases of MPI collectives where even a single copy interface does not get us the performance gain we could. For example in an MPI_Reduce rather than copy the data from the source we would like to instead use it directly in a mathops (say the reduce is doing a sum) as this would save us doing a copy. We don't need to keep a copy of the data from the source. I haven't implemented this, but I think this interface could in the future do all this through the use of the flags - eg could specify the math operation and type and the kernel rather than just copying the data would apply the specified operation between the source and destination and store it in the destination. Although we don't have a "second user" of the interface (though I've had some nibbles from people who may be interested in using it for intra process messaging which is not MPI). This interface is something which hardware vendors are already doing for their custom drivers to implement fast local communication. And so in addition to this being useful for OpenMPI it would mean the driver maintainers don't have to fix things up when the mm changes. There was some discussion about how much faster a true zero copy would go. Here's a link back to the email with some testing I did on that: http://marc.info/?l=linux-mm&m=130105930902915&w=2 There is a basic man page for the proposed interface here: http://ozlabs.org/~cyeoh/cma/process_vm_readv.txt This has been implemented for x86 and powerpc, other architecture should mainly (I think) just need to add syscall numbers for the process_vm_readv and process_vm_writev. There are 32 bit compatibility versions for 64-bit kernels. For arch maintainers there are some simple tests to be able to quickly verify that the syscalls are working correctly here: http://ozlabs.org/~cyeoh/cma/cma-test-20110718.tgz Signed-off-by: Chris Yeoh Cc: Ingo Molnar Cc: "H. Peter Anvin" Cc: Thomas Gleixner Cc: Arnd Bergmann Cc: Paul Mackerras Cc: Benjamin Herrenschmidt Cc: David Howells Cc: James Morris Cc: Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/sys_ni.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'kernel') diff --git a/kernel/sys_ni.c b/kernel/sys_ni.c index a9a5de07c4f1..47bfa16430d7 100644 --- a/kernel/sys_ni.c +++ b/kernel/sys_ni.c @@ -145,6 +145,10 @@ cond_syscall(sys_io_submit); cond_syscall(sys_io_cancel); cond_syscall(sys_io_getevents); cond_syscall(sys_syslog); +cond_syscall(sys_process_vm_readv); +cond_syscall(sys_process_vm_writev); +cond_syscall(compat_sys_process_vm_readv); +cond_syscall(compat_sys_process_vm_writev); /* arch-specific weak syscall entries */ cond_syscall(sys_pciconfig_read); -- cgit v1.2.1 From c9f01245b6a7d77d17deaa71af10f6aca14fa24e Mon Sep 17 00:00:00 2001 From: David Rientjes Date: Mon, 31 Oct 2011 17:07:15 -0700 Subject: oom: remove oom_disable_count This removes mm->oom_disable_count entirely since it's unnecessary and currently buggy. The counter was intended to be per-process but it's currently decremented in the exit path for each thread that exits, causing it to underflow. The count was originally intended to prevent oom killing threads that share memory with threads that cannot be killed since it doesn't lead to future memory freeing. The counter could be fixed to represent all threads sharing the same mm, but it's better to remove the count since: - it is possible that the OOM_DISABLE thread sharing memory with the victim is waiting on that thread to exit and will actually cause future memory freeing, and - there is no guarantee that a thread is disabled from oom killing just because another thread sharing its mm is oom disabled. Signed-off-by: David Rientjes Reported-by: Oleg Nesterov Reviewed-by: Oleg Nesterov Cc: Ying Han Cc: KOSAKI Motohiro Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/exit.c | 2 -- kernel/fork.c | 10 +--------- 2 files changed, 1 insertion(+), 11 deletions(-) (limited to 'kernel') diff --git a/kernel/exit.c b/kernel/exit.c index 2913b3509d42..d0b7d988f873 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -681,8 +681,6 @@ static void exit_mm(struct task_struct * tsk) enter_lazy_tlb(mm, current); /* We don't want this task to be frozen prematurely */ clear_freeze_flag(tsk); - if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) - atomic_dec(&mm->oom_disable_count); task_unlock(tsk); mm_update_next_owner(mm); mmput(mm); diff --git a/kernel/fork.c b/kernel/fork.c index 8e6b6f4fb272..70d76191afb9 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -501,7 +501,6 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) mm->cached_hole_size = ~0UL; mm_init_aio(mm); mm_init_owner(mm, p); - atomic_set(&mm->oom_disable_count, 0); if (likely(!mm_alloc_pgd(mm))) { mm->def_flags = 0; @@ -816,8 +815,6 @@ good_mm: /* Initializing for Swap token stuff */ mm->token_priority = 0; mm->last_interval = 0; - if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) - atomic_inc(&mm->oom_disable_count); tsk->mm = mm; tsk->active_mm = mm; @@ -1391,13 +1388,8 @@ bad_fork_cleanup_io: bad_fork_cleanup_namespaces: exit_task_namespaces(p); bad_fork_cleanup_mm: - if (p->mm) { - task_lock(p); - if (p->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) - atomic_dec(&p->mm->oom_disable_count); - task_unlock(p); + if (p->mm) mmput(p->mm); - } bad_fork_cleanup_signal: if (!(clone_flags & CLONE_THREAD)) free_signal_struct(p->signal); -- cgit v1.2.1 From bc3e53f682d93df677dbd5006a404722b3adfe18 Mon Sep 17 00:00:00 2001 From: Christoph Lameter Date: Mon, 31 Oct 2011 17:07:30 -0700 Subject: mm: distinguish between mlocked and pinned pages Some kernel components pin user space memory (infiniband and perf) (by increasing the page count) and account that memory as "mlocked". The difference between mlocking and pinning is: A. mlocked pages are marked with PG_mlocked and are exempt from swapping. Page migration may move them around though. They are kept on a special LRU list. B. Pinned pages cannot be moved because something needs to directly access physical memory. They may not be on any LRU list. I recently saw an mlockalled process where mm->locked_vm became bigger than the virtual size of the process (!) because some memory was accounted for twice: Once when the page was mlocked and once when the Infiniband layer increased the refcount because it needt to pin the RDMA memory. This patch introduces a separate counter for pinned pages and accounts them seperately. Signed-off-by: Christoph Lameter Cc: Mike Marciniszyn Cc: Roland Dreier Cc: Sean Hefty Cc: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/events/core.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index d1a1bee35228..12a0287e0358 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -3544,7 +3544,7 @@ static void perf_mmap_close(struct vm_area_struct *vma) struct ring_buffer *rb = event->rb; atomic_long_sub((size >> PAGE_SHIFT) + 1, &user->locked_vm); - vma->vm_mm->locked_vm -= event->mmap_locked; + vma->vm_mm->pinned_vm -= event->mmap_locked; rcu_assign_pointer(event->rb, NULL); mutex_unlock(&event->mmap_mutex); @@ -3625,7 +3625,7 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) lock_limit = rlimit(RLIMIT_MEMLOCK); lock_limit >>= PAGE_SHIFT; - locked = vma->vm_mm->locked_vm + extra; + locked = vma->vm_mm->pinned_vm + extra; if ((locked > lock_limit) && perf_paranoid_tracepoint_raw() && !capable(CAP_IPC_LOCK)) { @@ -3651,7 +3651,7 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) atomic_long_add(user_extra, &user->locked_vm); event->mmap_locked = extra; event->mmap_user = get_current_user(); - vma->vm_mm->locked_vm += event->mmap_locked; + vma->vm_mm->pinned_vm += event->mmap_locked; unlock: if (!ret) -- cgit v1.2.1 From f445027e4e692bd885118460b292d08027fd5501 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 31 Oct 2011 17:11:15 -0700 Subject: stop_machine: make stop_machine safe and efficient to call early Make stop_machine() safe to call early in boot, before SMP has been set up, by simply calling the callback function directly if there's only one CPU online. [ Fixes from AKPM: - add comment - local_irq_flags, not save_flags - also call hard_irq_disable() for systems which need it Tejun suggested using an explicit flag rather than just looking at the online cpu count. ] Cc: Tejun Heo Acked-by: Rusty Russell Cc: Peter Zijlstra Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Steven Rostedt Acked-by: Tejun Heo Cc: Konrad Rzeszutek Wilk Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/stop_machine.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) (limited to 'kernel') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index ba5070ce5765..5b0951aa0496 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -41,6 +41,7 @@ struct cpu_stopper { }; static DEFINE_PER_CPU(struct cpu_stopper, cpu_stopper); +static bool stop_machine_initialized = false; static void cpu_stop_init_done(struct cpu_stop_done *done, unsigned int nr_todo) { @@ -386,6 +387,8 @@ static int __init cpu_stop_init(void) cpu_stop_cpu_callback(&cpu_stop_cpu_notifier, CPU_ONLINE, bcpu); register_cpu_notifier(&cpu_stop_cpu_notifier); + stop_machine_initialized = true; + return 0; } early_initcall(cpu_stop_init); @@ -485,6 +488,25 @@ int __stop_machine(int (*fn)(void *), void *data, const struct cpumask *cpus) .num_threads = num_online_cpus(), .active_cpus = cpus }; + if (!stop_machine_initialized) { + /* + * Handle the case where stop_machine() is called + * early in boot before stop_machine() has been + * initialized. + */ + unsigned long flags; + int ret; + + WARN_ON_ONCE(smdata.num_threads != 1); + + local_irq_save(flags); + hard_irq_disable(); + ret = (*fn)(data); + local_irq_restore(flags); + + return ret; + } + /* Set the initial state and stop all online cpus. */ set_state(&smdata, STOPMACHINE_PREPARE); return stop_cpus(cpu_online_mask, stop_machine_cpu_stop, &smdata); -- cgit v1.2.1 From 4ff819515b203f937cc6c8a0215a37a68d1ee71f Mon Sep 17 00:00:00 2001 From: Vasily Averin Date: Mon, 31 Oct 2011 17:11:18 -0700 Subject: watchdog: move watchdog_*_all_cpus under CONFIG_SYSCTL Fix compilation warnings for CONFIG_SYSCTL=n: fixed compilation warnings in case of disabled CONFIG_SYSCTL kernel/watchdog.c:483:13: warning: `watchdog_enable_all_cpus' defined but not used kernel/watchdog.c:500:13: warning: `watchdog_disable_all_cpus' defined but not used these functions are static and are used only in sysctl handler, so move them inside #ifdef CONFIG_SYSCTL too Signed-off-by: Vasily Averin Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/watchdog.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/watchdog.c b/kernel/watchdog.c index d680381b0e9c..1d7bca7f4f52 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -481,6 +481,8 @@ static void watchdog_disable(int cpu) } } +/* sysctl functions */ +#ifdef CONFIG_SYSCTL static void watchdog_enable_all_cpus(void) { int cpu; @@ -510,8 +512,6 @@ static void watchdog_disable_all_cpus(void) } -/* sysctl functions */ -#ifdef CONFIG_SYSCTL /* * proc handler for /proc/sys/kernel/nmi_watchdog,watchdog_thresh */ -- cgit v1.2.1 From 73efc0394e148d0e15583e13712637831f926720 Mon Sep 17 00:00:00 2001 From: Dan Ballard Date: Mon, 31 Oct 2011 17:11:20 -0700 Subject: kernel/sysctl.c: add cap_last_cap to /proc/sys/kernel Userspace needs to know the highest valid capability of the running kernel, which right now cannot reliably be retrieved from the header files only. The fact that this value cannot be determined properly right now creates various problems for libraries compiled on newer header files which are run on older kernels. They assume capabilities are available which actually aren't. libcap-ng is one example. And we ran into the same problem with systemd too. Now the capability is exported in /proc/sys/kernel/cap_last_cap. [akpm@linux-foundation.org: make cap_last_cap const, per Ulrich] Signed-off-by: Dan Ballard Cc: Randy Dunlap Cc: Ingo Molnar Cc: Lennart Poettering Cc: Kay Sievers Cc: Ulrich Drepper Cc: James Morris Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/sysctl.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'kernel') diff --git a/kernel/sysctl.c b/kernel/sysctl.c index 2d2ecdcc8cdb..c49d66658ec0 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -57,6 +57,7 @@ #include #include #include +#include #include #include @@ -134,6 +135,7 @@ static int minolduid; static int min_percpu_pagelist_fract = 8; static int ngroups_max = NGROUPS_MAX; +static const int cap_last_cap = CAP_LAST_CAP; #ifdef CONFIG_INOTIFY_USER #include @@ -740,6 +742,13 @@ static struct ctl_table kern_table[] = { .mode = 0444, .proc_handler = proc_dointvec, }, + { + .procname = "cap_last_cap", + .data = (void *)&cap_last_cap, + .maxlen = sizeof(int), + .mode = 0444, + .proc_handler = proc_dointvec, + }, #if defined(CONFIG_LOCKUP_DETECTOR) { .procname = "watchdog", -- cgit v1.2.1 From 0eca6b7c78fd997e02bd9850e608102382b7822e Mon Sep 17 00:00:00 2001 From: Yanmin Zhang Date: Mon, 31 Oct 2011 17:11:25 -0700 Subject: printk: add module parameter ignore_loglevel to control ignore_loglevel We are enabling some power features on medfield. To test suspend-2-RAM conveniently, we need turn on/off ignore_loglevel frequently without rebooting. Add a module parameter, so users can change it by: /sys/module/printk/parameters/ignore_loglevel Signed-off-by: Yanmin Zhang Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/printk.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'kernel') diff --git a/kernel/printk.c b/kernel/printk.c index b7da18391c38..e62f949ec140 100644 --- a/kernel/printk.c +++ b/kernel/printk.c @@ -532,6 +532,9 @@ static int __init ignore_loglevel_setup(char *str) } early_param("ignore_loglevel", ignore_loglevel_setup); +module_param_named(ignore_loglevel, ignore_loglevel, bool, S_IRUGO | S_IWUSR); +MODULE_PARM_DESC(ignore_loglevel, "ignore loglevel setting, to" + "print all kernel messages to the console."); /* * Write out chars from start to end - 1 inclusive -- cgit v1.2.1 From 134620f7a865b3bc9e3d56d460603592b70ede21 Mon Sep 17 00:00:00 2001 From: Yanmin Zhang Date: Mon, 31 Oct 2011 17:11:27 -0700 Subject: printk: add console_suspend module parameter We are enabling some power features on medfield. To test suspend-2-RAM conveniently, we need turn on/off console_suspend_enabled frequently. Add a module parameter, so users could change it by: /sys/module/printk/parameters/console_suspend Signed-off-by: Yanmin Zhang Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/printk.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'kernel') diff --git a/kernel/printk.c b/kernel/printk.c index e62f949ec140..6d9dedd11450 100644 --- a/kernel/printk.c +++ b/kernel/printk.c @@ -1111,6 +1111,10 @@ static int __init console_suspend_disable(char *str) return 1; } __setup("no_console_suspend", console_suspend_disable); +module_param_named(console_suspend, console_suspend_enabled, + bool, S_IRUGO | S_IWUSR); +MODULE_PARM_DESC(console_suspend, "suspend console during suspend" + " and hibernate operations"); /** * suspend_console - suspend the console subsystem -- cgit v1.2.1 From 48e41899e4a3592746e5263c14681bf5c1393563 Mon Sep 17 00:00:00 2001 From: William Douglas Date: Mon, 31 Oct 2011 17:11:29 -0700 Subject: printk: fix bounds checking for log_prefix Currently log_prefix is testing that the first character of the log level and facility is less than '0' and greater than '9' (which is always false). It should be testing to see if the character less than '0' or greater than '9' instead. This patch makes that change. The code being changed worked because strtoul bombs out (endp isn't updated) and 0 is returned anyway. Signed-off-by: William Douglas Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/printk.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/printk.c b/kernel/printk.c index 6d9dedd11450..286d2c7be52c 100644 --- a/kernel/printk.c +++ b/kernel/printk.c @@ -595,7 +595,7 @@ static size_t log_prefix(const char *p, unsigned int *level, char *special) /* multi digit including the level and facility number */ char *endp = NULL; - if (p[1] < '0' && p[1] > '9') + if (p[1] < '0' || p[1] > '9') return 0; lev = (simple_strtoul(&p[1], &endp, 10) & 7); -- cgit v1.2.1 From ae29bc92da01a2e9d278a9a58c3b307d41cc0254 Mon Sep 17 00:00:00 2001 From: William Douglas Date: Mon, 31 Oct 2011 17:11:31 -0700 Subject: printk: remove bounds checking for log_prefix Currently log_prefix is testing that the first character of the log level and facility is less than '0' and greater than '9' (which is always false). Since the code being updated works because strtoul bombs out (endp isn't updated) and 0 is returned anyway just remove the check and don't change the behavior of the function. Signed-off-by: William Douglas Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/printk.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'kernel') diff --git a/kernel/printk.c b/kernel/printk.c index 286d2c7be52c..1455a0d4eedd 100644 --- a/kernel/printk.c +++ b/kernel/printk.c @@ -595,9 +595,6 @@ static size_t log_prefix(const char *p, unsigned int *level, char *special) /* multi digit including the level and facility number */ char *endp = NULL; - if (p[1] < '0' || p[1] > '9') - return 0; - lev = (simple_strtoul(&p[1], &endp, 10) & 7); if (endp == NULL || endp[0] != '>') return 0; -- cgit v1.2.1 From 50e1499f468fd74c6db95deb2e1e6bfee578ae70 Mon Sep 17 00:00:00 2001 From: Andy Shevchenko Date: Mon, 31 Oct 2011 17:12:51 -0700 Subject: kgdb: follow rename pack_hex_byte() to hex_byte_pack() There is no functional change. Signed-off-by: Andy Shevchenko Acked-by: Jesper Nilsson Cc: David Howells Cc: Koichi Yasutake Cc: Jason Wessel Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/debug/gdbstub.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/debug/gdbstub.c b/kernel/debug/gdbstub.c index 34872482315e..c22d8c28ad84 100644 --- a/kernel/debug/gdbstub.c +++ b/kernel/debug/gdbstub.c @@ -217,7 +217,7 @@ void gdbstub_msg_write(const char *s, int len) /* Pack in hex chars */ for (i = 0; i < wcount; i++) - bufptr = pack_hex_byte(bufptr, s[i]); + bufptr = hex_byte_pack(bufptr, s[i]); *bufptr = '\0'; /* Move up */ @@ -249,7 +249,7 @@ char *kgdb_mem2hex(char *mem, char *buf, int count) if (err) return NULL; while (count > 0) { - buf = pack_hex_byte(buf, *tmp); + buf = hex_byte_pack(buf, *tmp); tmp++; count--; } @@ -411,14 +411,14 @@ static char *pack_threadid(char *pkt, unsigned char *id) limit = id + (BUF_THREAD_ID_SIZE / 2); while (id < limit) { if (!lzero || *id != 0) { - pkt = pack_hex_byte(pkt, *id); + pkt = hex_byte_pack(pkt, *id); lzero = 0; } id++; } if (lzero) - pkt = pack_hex_byte(pkt, 0); + pkt = hex_byte_pack(pkt, 0); return pkt; } @@ -486,7 +486,7 @@ static void gdb_cmd_status(struct kgdb_state *ks) dbg_remove_all_break(); remcom_out_buffer[0] = 'S'; - pack_hex_byte(&remcom_out_buffer[1], ks->signo); + hex_byte_pack(&remcom_out_buffer[1], ks->signo); } static void gdb_get_regs_helper(struct kgdb_state *ks) @@ -954,7 +954,7 @@ int gdb_serial_stub(struct kgdb_state *ks) /* Reply to host that an exception has occurred */ ptr = remcom_out_buffer; *ptr++ = 'T'; - ptr = pack_hex_byte(ptr, ks->signo); + ptr = hex_byte_pack(ptr, ks->signo); ptr += strlen(strcpy(ptr, "thread:")); int_to_threadref(thref, shadow_pid(current->pid)); ptr = pack_threadid(ptr, thref); -- cgit v1.2.1