- 03 Jul, 2013 40 commits
-
-
Ed Cashin authored
Some users have a large AoE target while others like to use many AoE targets at the same time. In the latter case, there is an opportunity to greatly improve aggregate throughput by allowing different threads to complete the I/O associated with each target. For 36 targets, 4 KiB read throughput roughly doubles, for example, with these changes in place. Signed-off-by:
Ed Cashin <ecashin@coraid.com> Signed-off-by:
Andrew Morton <akpm@linux-foundation.org> Signed-off-by:
Linus Torvalds <torvalds@linux-foundation.org>
-
Paul Clements authored
Currently, when a disconnect is requested by the user (via NBD_DISCONNECT ioctl) the return from NBD_DO_IT is undefined (it is usually one of several error codes). This means that nbd-client does not know if a manual disconnect was performed or whether a network error occurred. Because of this, nbd-client's persist mode (which tries to reconnect after error, but not after manual disconnect) does not always work correctly. This change fixes this by causing NBD_DO_IT to always return 0 if a user requests a disconnect. This means that nbd-client can correctly either persist the connection (if an error occurred) or disconnect (if the user requested it). Signed-off-by:
Paul Clements <paul.clements@steeleye.com> Acked-by:
Rob Landley <rob@landley.net> Cc: <stable@vger.kernel.org> Signed-off-by:
-
Michal Belczyk authored
The NBD_CLEAR_QUE ioctl has been deprecated for quite some time (its job is now done by two other ioctls). We should stop trying to make bogus assertions in it. Also, user-level code should remove calls to NBD_CLEAR_QUE, ASAP. Signed-off-by:
Michal Belczyk <belczyk@bsd.krakow.pl> Signed-off-by:
-
Raphael S. Carvalho authored
Move statement to static initilization of init_pid_ns. Signed-off-by:
Raphael S. Carvalho <raphael.scarv@gmail.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Acked-by:
Serge Hallyn <serge.hallyn@canonical.com> Signed-off-by:
-
Alexandre Bounine authored
Change endpoint device name format to use a component tag value instead of device destination ID. RapidIO specification defines a component tag to be a unique identifier for devices in a network. RapidIO switches already use component tag as part of their device name and also use it for device identification when processing error management event notifications. Forming an endpoint's device name using its component tag instead of destination ID allows to keep sysfs device directories unchanged in case if a routing process dynamically changes endpoint's destination ID as a result of route optimization. This change should not affect any existing users because a valid device destination ID always should be obtained by reading "destid" attribute and not by parsing device name. This patch also removes switchid member from struct rio_switch because it simply duplicates the component tag and does not have other use than in device name generation. Signed-off-by:
Alexandre Bounine <alexandre.bounine@idt.com> Cc: Matt Porter <mporter@kernel.crashing.org> Cc: Li Yang <leoli@freescale.com> Cc: Kumar Gala <galak@kernel.crashing.org> Cc: Andre van Herk <andre.van.herk@Prodrive.nl> Cc: Micha Nelissen <micha.nelissen@Prodrive.nl> Cc: Stef van Os <stef.van.os@Prodrive.nl> Signed-off-by:
-
Alexandre Bounine authored
Update RapidIO documentation files to reflect modularization changes. Signed-off-by:
-
Alexandre Bounine authored
Add RapidIO-specific modalias generation to enable udev notifications about RapidIO-specific events. The RapidIO modalias string format is shown below: "rapidio:vNNNNdNNNNavNNNNadNNNN" Where: v - Device Vendor ID (16 bit), d - Device ID (16 bit), av - Assembly Vendor ID (16 bit), ad - Assembly ID (16 bit), as they are reported in corresponding Capability Registers (CARs) of each RapidIO device. Signed-off-by:
-
Alexandre Bounine authored
Add a configuration option to build RapidIO subsystem core code as a loadable kernel module. Currently this option is available only for x86-based platforms, with the additional patch for PowerPC planned to be provided later. This patch replaces kernel command line parameter "riohdid=" with its module-specific analog "rapidio.hdid=". Signed-off-by:
-
Alexandre Bounine authored
This patch adds an option to build device driver for Tsi721 PCIe-to-SRIO bridge device as a kernel module. Currently this module cannot be unloaded because the existing RapidIO subsystem code does not support dynamic removal of local RapidIO controllers (TODO). Signed-off-by:
-
Alexandre Bounine authored
Update enumeration/discovery method registration mechanism to allow loading enumeration/discovery methods before all mports are registered. Existing statically linked RapidIO subsystem expects that all available RapidIO mport devices are initialized and registered before the enumeration/discovery method is registered. Switching to loadable mport device drivers creates situation when mport device driver can be loaded after enumeration/discovery method is attached (e.g., loadable mport driver in a system with statically linked RapidIO core and enumerator). This also will happen in a system with hot-pluggable RapidIO controllers. To remove the dependency on the initialization/registration order this patch introduces enumeration/discovery registration mechanism that supports arbitrary registration order of mports and enumerator/discovery methods. The following registration rules are implemented: - only one enumeration/discovery method can be registered for given mport ID (including RIO_MPORT_ANY); - when new enumeration/discovery methods tries to attach to the registered mport device, method with matching mport ID will replace a default method previously registered for given mport (if any); - enumeration/discovery method with target ID=RIO_MPORT_ANY will be attached only to mports that do not have another enumerator attached to them; - when new mport device is registered with RapidIO subsystem, registration routine searches for the enumeration/discovery method with the best matching mport ID; Signed-off-by:
-
Alexandre Bounine authored
Rework probe/remove routines to prevent rionet driver from monopolizing target RapidIO devices. Fix conflict with modular RapidIO switch drivers. Using one of RapidIO messaging channels rionet driver provides a service layer common to all endpoint devices in a system's RapidIO network. These devices may also require their own specific device driver which will be blocked from attaching to the target device by rionet (or block rionet if loaded earlier). To avoid conflict with device-specific drivers, the rionet driver is reworked to be registered as a subsystem interface on the RapidIO bus. The reworked rio_remove_dev() and rionet_exit() routines also include handling of individual rionet peer device removal which was not supported before. Signed-off-by:
-
Alexandre Bounine authored
Rework RapidIO switch drivers to add an option to build them as loadable kernel modules. This patch removes RapidIO-specific vmlinux section and converts switch drivers to be compatible with LDM driver registration method. To simplify registration of device-specific callback routines this patch introduces rio_switch_ops data structure. The sw_sysfs() callback is removed from the list of device-specific operations because under the new structure its functions can be handled by switch driver's probe() and remove() routines. If a specific switch device driver is not loaded the RapidIO subsystem core will use default standard-based operations to configure a switch. Because the current implementation of RapidIO enumeration/discovery method relies on availability of device-specific operations for error management, switch device drivers must be loaded before the RapidIO enumeration/discovery starts. This patch also moves several common routines from enumeration/discovery module into the RapidIO core code to make switch-specific operations accessible to all components of RapidIO subsystem. Signed-off-by:
-
Wu Fengguang authored
sparse warnings: drivers/rapidio/rio-scan.c:1143:5: sparse: symbol 'rio_enum_mport' was not declared. Should it be static? drivers/rapidio/rio-scan.c:1246:5: sparse: symbol 'rio_disc_mport' was not declared. Should it be static? Signed-off-by:
Fengguang Wu <fengguang.wu@intel.com> Cc: "Bounine, Alexandre" <Alexandre.Bounine@idt.com> Signed-off-by:
-
Alexandre Bounine authored
Remove the driver for Tsi500 Parallel RapidIO switch because this device has not been available for several years. Since the first introduction of Tsi500, the parallel RapidIO interface was replaced by the serial RapidIO (sRIO) and therefore there is no value in keeping this driver. Signed-off-by:
-
Jean Delvare authored
We print a dump stack after idr_remove warning. This is useful to find the faulty piece of code. Let's do the same for ida_remove, as it would be equally useful there. [akpm@linux-foundation.org: convert the open-coded printk+dump_stack into WARN()] Signed-off-by:
Jean Delvare <jdelvare@suse.de> Cc: Tejun Heo <tj@kernel.org> Cc: Takashi Iwai <tiwai@suse.de> Signed-off-by:
-
Zhang Yanfei authored
The only user of saved_max_pfn in s390 is read_oldmem interface but we have removed that interface, so saved_max_pfn is now unneeded in s390, and we needn't set it anymore. Signed-off-by:
Zhang Yanfei <zhangyanfei@cn.fujitsu.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Michael Holzheu <holzheu@linux.vnet.ibm.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Dave Hansen <dave@sr71.net> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Matt Fleming <matt.fleming@intel.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Tony Luck <tony.luck@intel.com> Cc: Vivek Goyal <vgoyal@redhat.com> Signed-off-by:
-
Zhang Yanfei authored
The only user of saved_max_pfn in ia64 is read_oldmem interface but we have removed that interface, so saved_max_pfn is now unneeded in ia64, and we needn't set it anymore. Signed-off-by:
-
Zhang Yanfei authored
saved_max_pfn is used to know the amount of memory that the previous kernel used. And for powerpc, we set saved_max_pfn by passing the kernel commandline parameter "savemaxmem=". The only user of saved_max_pfn in powerpc is read_oldmem interface. Since we have removed read_oldmem, we don't need this parameter anymore. Signed-off-by:
-
Zhang Yanfei authored
saved_max_pfn is used to know the amount of memory that the previous kernel used. And for powerpc, we set saved_max_pfn by passing the kernel commandline parameter "savemaxmem=". The only user of saved_max_pfn in mips is read_oldmem interface. Since we have removed read_oldmem, so we don't need this parameter anymore. Signed-off-by:
-
Zhang Yanfei authored
Signed-off-by:
-
Zhang Yanfei authored
/dev/oldmem provides the interface for us to access the "old memory" in the dump-capture kernel. Unfortunately, no one actually uses this interface. And this interface could actually cause some real problems if used on ia64 where the cached/uncached accesses are mixed. See the discussion from the link: https://lkml.org/lkml/2013/4/12/386. So Eric suggested that we should remove /dev/oldmem as an unused piece of code. [akpm@linux-foundation.org: mention /dev/oldmem obsolescence in devices.txt] Suggested-by:
"Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by:
-
Oleg Nesterov authored
924b42d5 ("Use boot based time for process start time and boot time in /proc") updated copy_process/do_task_stat but forgot about de_thread(). This breaks "ps axOT" if a sub-thread execs. Note: I think that task->start_time should die. Signed-off-by:
Oleg Nesterov <oleg@redhat.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Acked-by:
John Stultz <johnstul@us.ibm.com> Cc: Tomas Janousek <tjanouse@redhat.com> Cc: Tomas Smetana <tsmetana@redhat.com> Signed-off-by:
-
Oleg Nesterov authored
Trivial cleanup. do_execve_common() can use current_user() and avoid the unnecessary "struct cred *cred" var. Signed-off-by:
-
Oleg Nesterov authored
copy_process() does a lot of "chaotic" initializations and checks CLONE_THREAD twice before it takes tasklist. In particular it sets "p->group_leader = p" and then changes it again under tasklist if !thread_group_leader(p). This looks a bit confusing, lets create a single "if (CLONE_THREAD)" block which initializes ->exit_signal, ->group_leader, and ->tgid. Signed-off-by:
-
Oleg Nesterov authored
copy_process() adds the new child to thread_group/init_task.tasks list and then does attach_pid(child, PIDTYPE_PID). This means that the lockless next_thread() or next_task() can see this thread with the wrong pid. Say, "ls /proc/pid/task" can list the same inode twice. We could move attach_pid(child, PIDTYPE_PID) up, but in this case find_task_by_vpid() can find the new thread before it was fully initialized. And this is already true for PIDTYPE_PGID/PIDTYPE_SID, With this patch copy_process() initializes child->pids[*].pid first, then calls attach_pid() to insert the task into the pid->tasks list. attach_pid() no longer need the "struct pid*" argument, it is always called after pid_link->pid was already set. Signed-off-by:
-
Oleg Nesterov authored
Cleanup and preparation for the next changes. Move the "if (clone_flags & CLONE_THREAD)" code down under "if (likely(p->pid))" and turn it into into the "else" branch. This makes the process/thread initialization more symmetrical and removes one check. Signed-off-by:
-
Eric Paris authored
When a task is attempting to violate the RLIMIT_NPROC limit we have a check to see if the task is sufficiently priviledged. The check first looks at CAP_SYS_ADMIN, then CAP_SYS_RESOURCE, then if the task is uid=0. A result is that tasks which are allowed by the uid=0 check are first checked against the security subsystem. This results in the security subsystem auditting a denial for sys_admin and sys_resource and then the task passing the uid=0 check. This patch rearranges the code to first check uid=0, since if we pass that we shouldn't hit the security system at all. We then check sys_resource, since it is the smallest capability which will solve the problem. Lastly we check the fallback everything cap_sysadmin. We don't want to give this capability many places since it is so powerful. This will eliminate many of the false positive/needless denial messages we get when a root task tries to violate the nproc limit. (note that kthreads count against root, so on a sufficiently large machine we can actually get past the default limits before any userspace tasks are launched.) Signed-off-by:
Eric Paris <eparis@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by:
-
Zhao Hongjiang authored
For NUL terminated string, set '\0' at the end. Signed-off-by:
Zhao Hongjiang <zhaohongjiang@huawei.com> Signed-off-by:
-
Oleg Nesterov authored
Change uptime_proc_show() to use get_monotonic_boottime() instead of do_posix_clock_monotonic_gettime() + monotonic_to_bootbased(). Signed-off-by:
-
Oleg Nesterov authored
Move __set_special_pids() from exit.c to sys.c close to its single caller and make it static. And rename it to set_special_pids(), another helper with this name has gone away. Signed-off-by:
-
Oleg Nesterov authored
de_thread() can use change_pid() instead of detach + attach. This looks better and this ensures that, say, next_thread() can never see a task with ->pid == NULL. Signed-off-by:
-
Oleg Nesterov authored
"goto end" should not bypass the "Backward compatibility with core_uses_pid" code, move this label up. While at it, - It is ugly to copy '|' into cn->corename and then inc the pointer for argv_split(). Change format_corename() to increment pat_ptr instead. - Remove the dead "if (*pat_ptr == 0)" in format_corename(), we already checked it is not zero. Signed-off-by:
Neil Horman <nhorman@tuxdriver.com> Signed-off-by:
-
Oleg Nesterov authored
Imho, "atomic_t call_count" is ugly and should die. It buys nothing and in fact it can grow more than necessary, expand doesn't check if it was already incremented by another task. Kill it, and introduce "static int core_name_size" updated by expand_corename(). This is obviously racy too but harmless, and core_name_size never grows for no reason. We do not bother to to calculate the "right" new size, we simply do kmalloc(size_we_need) and use ksize() to rely on kmalloc_index's decision. Finally change format_corename() to use expand_corename(), krealloc(NULL) is fine. Signed-off-by:
-
Oleg Nesterov authored
The usage of cn_escape() looks really annoying, imho this sequence needs a wrapper. And it is buggy. If cn_printf() does expand_corename() cn_escape() writes to the freed memory. Introduce cn_esc_printf() which hopefully does this all right. It records the index before cn_vprintf(), not "char *" which is no longer valid (in general) after krealloc(). Signed-off-by:
-
Oleg Nesterov authored
cn_vprintf() looks really overcomplicated and sub-optimal. We do not need vsnprintf(NULL) to calculate the size we need, we can simply try to print into the current buffer and expand/retry only if necessary. Signed-off-by:
-
Oleg Nesterov authored
Turn cn_printf(...) into cn_vprintf(va_list args), reintroduce cn_printf() as a trivial wrapper. This simplifies the next change and cn_vprintf() will have more callers. Signed-off-by:
-
Oleg Nesterov authored
do_coredump() assumes that format_corename() can only fail if expand_corename() fails and frees cn->corename. This is not true, for example cn_print_exe_file() can fail and in this case nobody frees cn->corename. Change do_coredump() to always do kfree(cn->corename) after it calls format_corename() (NULL is fine), change expand_corename() to do nothing if kmalloc() fails. Signed-off-by:
-
Oleg Nesterov authored
call_usermodehelper_exec() does nothing but returns success if path[0] == 0. The only user which needs this strange feature is request_module(), it can check modprobe_path[0] itself like other users do if they want to detect the "disabled by admin" case. Kill it. Not only it looks strange, it can confuse other callers. And this allows us to revert 264b83c0 ("usermodehelper: check subprocess_info->path != NULL"), do_execve(NULL) is safe. Signed-off-by:
Rusty Russell <rusty@rustcorp.com.au> Cc: Lucas De Marchi <lucas.de.marchi@gmail.com> Signed-off-by:
-
Oleg Nesterov authored
sigprocmask() should die. None of the current callers actually need this strange interface. Change fs/eventpoll.c to use set_current_blocked(). This also means we should not worry about SIGKILL/SIGSTOP. Signed-off-by:
-
Andrey Vagin authored
crtools uses a parasite code for dumping processes. The parasite code is injected into a process with help PTRACE_SEIZE. Currently crtools blocks signals from a parasite code. If a process has pending signals, crtools wait while a process handles these signals. This method is not suitable for stopped tasks. A stopped task can have a few pending signals, when we will try to execute a parasite code, we will need to drop SIGSTOP, but all other signals must remain pending, because a state of processes must not be changed during checkpointing. This patch adds two ptrace commands to set/get signal-blocked mask. I think gdb can use this commands too. [akpm@linux-foundation.org: be consistent with brace layout] Signed-off-by:
Andrey Vagin <avagin@openvz.org> Reviewed-by:
-
