freebsd-dev/sys/kern/uipc_sem.c

1066 lines
24 KiB
C
Raw Normal View History

/*-
* Copyright (c) 2002 Alfred Perlstein <alfred@FreeBSD.org>
* Copyright (c) 2003-2005 SPARTA, Inc.
* Copyright (c) 2005 Robert N. M. Watson
* All rights reserved.
*
* This software was developed for the FreeBSD Project in part by Network
* Associates Laboratories, the Security Research Division of Network
* Associates, Inc. under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"),
* as part of the DARPA CHATS research program.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
2003-06-11 00:56:59 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_compat.h"
#include "opt_posix.h"
#include <sys/param.h>
#include <sys/capsicum.h>
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
#include <sys/condvar.h>
#include <sys/fcntl.h>
#include <sys/file.h>
#include <sys/filedesc.h>
#include <sys/fnv_hash.h>
#include <sys/kernel.h>
#include <sys/ksem.h>
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/mutex.h>
#include <sys/priv.h>
#include <sys/proc.h>
#include <sys/posix4.h>
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
#include <sys/_semaphore.h>
#include <sys/stat.h>
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
#include <sys/syscall.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
#include <sys/sysent.h>
#include <sys/sysproto.h>
#include <sys/systm.h>
#include <sys/sx.h>
#include <sys/user.h>
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
#include <sys/vnode.h>
#include <security/mac/mac_framework.h>
FEATURE(p1003_1b_semaphores, "POSIX P1003.1B semaphores support");
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
/*
* TODO
*
* - Resource limits?
* - Replace global sem_lock with mtx_pool locks?
* - Add a MAC check_create() hook for creating new named semaphores.
*/
#ifndef SEM_MAX
#define SEM_MAX 30
#endif
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
#ifdef SEM_DEBUG
#define DP(x) printf x
#else
#define DP(x)
#endif
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct ksem_mapping {
char *km_path;
Fnv32_t km_fnv;
struct ksem *km_ksem;
LIST_ENTRY(ksem_mapping) km_link;
};
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
static MALLOC_DEFINE(M_KSEM, "ksem", "semaphore file descriptor");
static LIST_HEAD(, ksem_mapping) *ksem_dictionary;
static struct sx ksem_dict_lock;
static struct mtx ksem_count_lock;
static struct mtx sem_lock;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
static u_long ksem_hash;
static int ksem_dead;
#define KSEM_HASH(fnv) (&ksem_dictionary[(fnv) & ksem_hash])
static int nsems = 0;
SYSCTL_DECL(_p1003_1b);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
SYSCTL_INT(_p1003_1b, OID_AUTO, nsems, CTLFLAG_RD, &nsems, 0,
"Number of active kernel POSIX semaphores");
static int kern_sem_wait(struct thread *td, semid_t id, int tryflag,
struct timespec *abstime);
static int ksem_access(struct ksem *ks, struct ucred *ucred);
static struct ksem *ksem_alloc(struct ucred *ucred, mode_t mode,
unsigned int value);
static int ksem_create(struct thread *td, const char *path,
semid_t *semidp, mode_t mode, unsigned int value,
int flags, int compat32);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
static void ksem_drop(struct ksem *ks);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
static int ksem_get(struct thread *td, semid_t id, cap_rights_t *rightsp,
struct file **fpp);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
static struct ksem *ksem_hold(struct ksem *ks);
static void ksem_insert(char *path, Fnv32_t fnv, struct ksem *ks);
static struct ksem *ksem_lookup(char *path, Fnv32_t fnv);
static void ksem_module_destroy(void);
static int ksem_module_init(void);
static int ksem_remove(char *path, Fnv32_t fnv, struct ucred *ucred);
static int sem_modload(struct module *module, int cmd, void *arg);
static fo_stat_t ksem_stat;
static fo_close_t ksem_closef;
static fo_chmod_t ksem_chmod;
static fo_chown_t ksem_chown;
static fo_fill_kinfo_t ksem_fill_kinfo;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
/* File descriptor operations. */
static struct fileops ksem_ops = {
.fo_read = invfo_rdwr,
.fo_write = invfo_rdwr,
.fo_truncate = invfo_truncate,
.fo_ioctl = invfo_ioctl,
.fo_poll = invfo_poll,
.fo_kqfilter = invfo_kqfilter,
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
.fo_stat = ksem_stat,
.fo_close = ksem_closef,
.fo_chmod = ksem_chmod,
.fo_chown = ksem_chown,
.fo_sendfile = invfo_sendfile,
.fo_fill_kinfo = ksem_fill_kinfo,
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
.fo_flags = DFLAG_PASSABLE
};
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
FEATURE(posix_sem, "POSIX semaphores");
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
static int
ksem_stat(struct file *fp, struct stat *sb, struct ucred *active_cred,
struct thread *td)
{
struct ksem *ks;
#ifdef MAC
int error;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
#endif
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
ks = fp->f_data;
#ifdef MAC
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = mac_posixsem_check_stat(active_cred, fp->f_cred, ks);
if (error)
return (error);
#endif
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
/*
* Attempt to return sanish values for fstat() on a semaphore
* file descriptor.
*/
bzero(sb, sizeof(*sb));
mtx_lock(&sem_lock);
sb->st_atim = ks->ks_atime;
sb->st_ctim = ks->ks_ctime;
sb->st_mtim = ks->ks_mtime;
sb->st_birthtim = ks->ks_birthtime;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
sb->st_uid = ks->ks_uid;
sb->st_gid = ks->ks_gid;
sb->st_mode = S_IFREG | ks->ks_mode; /* XXX */
mtx_unlock(&sem_lock);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
return (0);
}
static int
ksem_chmod(struct file *fp, mode_t mode, struct ucred *active_cred,
struct thread *td)
{
struct ksem *ks;
int error;
error = 0;
ks = fp->f_data;
mtx_lock(&sem_lock);
#ifdef MAC
error = mac_posixsem_check_setmode(active_cred, ks, mode);
if (error != 0)
goto out;
#endif
error = vaccess(VREG, ks->ks_mode, ks->ks_uid, ks->ks_gid, VADMIN,
active_cred, NULL);
if (error != 0)
goto out;
ks->ks_mode = mode & ACCESSPERMS;
out:
mtx_unlock(&sem_lock);
return (error);
}
static int
ksem_chown(struct file *fp, uid_t uid, gid_t gid, struct ucred *active_cred,
struct thread *td)
{
struct ksem *ks;
int error;
error = 0;
ks = fp->f_data;
mtx_lock(&sem_lock);
#ifdef MAC
error = mac_posixsem_check_setowner(active_cred, ks, uid, gid);
if (error != 0)
goto out;
#endif
if (uid == (uid_t)-1)
uid = ks->ks_uid;
if (gid == (gid_t)-1)
gid = ks->ks_gid;
if (((uid != ks->ks_uid && uid != active_cred->cr_uid) ||
(gid != ks->ks_gid && !groupmember(gid, active_cred))) &&
(error = priv_check_cred(active_cred, PRIV_VFS_CHOWN, 0)))
goto out;
ks->ks_uid = uid;
ks->ks_gid = gid;
out:
mtx_unlock(&sem_lock);
return (error);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
static int
ksem_closef(struct file *fp, struct thread *td)
{
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct ksem *ks;
ks = fp->f_data;
fp->f_data = NULL;
ksem_drop(ks);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
return (0);
}
static int
ksem_fill_kinfo(struct file *fp, struct kinfo_file *kif, struct filedesc *fdp)
{
struct ksem *ks;
kif->kf_type = KF_TYPE_SEM;
ks = fp->f_data;
mtx_lock(&sem_lock);
kif->kf_un.kf_sem.kf_sem_value = ks->ks_value;
kif->kf_un.kf_sem.kf_sem_mode = S_IFREG | ks->ks_mode; /* XXX */
mtx_unlock(&sem_lock);
if (ks->ks_path != NULL) {
sx_slock(&ksem_dict_lock);
if (ks->ks_path != NULL)
strlcpy(kif->kf_path, ks->ks_path, sizeof(kif->kf_path));
sx_sunlock(&ksem_dict_lock);
}
return (0);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
/*
* ksem object management including creation and reference counting
* routines.
*/
static struct ksem *
ksem_alloc(struct ucred *ucred, mode_t mode, unsigned int value)
{
struct ksem *ks;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
mtx_lock(&ksem_count_lock);
if (nsems == p31b_getcfg(CTL_P1003_1B_SEM_NSEMS_MAX) || ksem_dead) {
mtx_unlock(&ksem_count_lock);
return (NULL);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
nsems++;
mtx_unlock(&ksem_count_lock);
ks = malloc(sizeof(*ks), M_KSEM, M_WAITOK | M_ZERO);
ks->ks_uid = ucred->cr_uid;
ks->ks_gid = ucred->cr_gid;
ks->ks_mode = mode;
ks->ks_value = value;
cv_init(&ks->ks_cv, "ksem");
vfs_timestamp(&ks->ks_birthtime);
ks->ks_atime = ks->ks_mtime = ks->ks_ctime = ks->ks_birthtime;
refcount_init(&ks->ks_ref, 1);
#ifdef MAC
mac_posixsem_init(ks);
mac_posixsem_create(ucred, ks);
#endif
return (ks);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
static struct ksem *
ksem_hold(struct ksem *ks)
{
refcount_acquire(&ks->ks_ref);
return (ks);
}
static void
ksem_drop(struct ksem *ks)
{
if (refcount_release(&ks->ks_ref)) {
#ifdef MAC
mac_posixsem_destroy(ks);
#endif
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
cv_destroy(&ks->ks_cv);
free(ks, M_KSEM);
mtx_lock(&ksem_count_lock);
nsems--;
mtx_unlock(&ksem_count_lock);
}
}
/*
* Determine if the credentials have sufficient permissions for read
* and write access.
*/
static int
ksem_access(struct ksem *ks, struct ucred *ucred)
{
int error;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = vaccess(VREG, ks->ks_mode, ks->ks_uid, ks->ks_gid,
VREAD | VWRITE, ucred, NULL);
if (error)
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = priv_check_cred(ucred, PRIV_SEM_WRITE, 0);
return (error);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
/*
* Dictionary management. We maintain an in-kernel dictionary to map
* paths to semaphore objects. We use the FNV hash on the path to
* store the mappings in a hash table.
*/
static struct ksem *
ksem_lookup(char *path, Fnv32_t fnv)
{
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct ksem_mapping *map;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
LIST_FOREACH(map, KSEM_HASH(fnv), km_link) {
if (map->km_fnv != fnv)
continue;
if (strcmp(map->km_path, path) == 0)
return (map->km_ksem);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
return (NULL);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
static void
ksem_insert(char *path, Fnv32_t fnv, struct ksem *ks)
{
struct ksem_mapping *map;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
map = malloc(sizeof(struct ksem_mapping), M_KSEM, M_WAITOK);
map->km_path = path;
map->km_fnv = fnv;
map->km_ksem = ksem_hold(ks);
ks->ks_path = path;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
LIST_INSERT_HEAD(KSEM_HASH(fnv), map, km_link);
}
static int
ksem_remove(char *path, Fnv32_t fnv, struct ucred *ucred)
{
struct ksem_mapping *map;
int error;
LIST_FOREACH(map, KSEM_HASH(fnv), km_link) {
if (map->km_fnv != fnv)
continue;
if (strcmp(map->km_path, path) == 0) {
#ifdef MAC
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = mac_posixsem_check_unlink(ucred, map->km_ksem);
if (error)
return (error);
#endif
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = ksem_access(map->km_ksem, ucred);
if (error)
return (error);
map->km_ksem->ks_path = NULL;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
LIST_REMOVE(map, km_link);
ksem_drop(map->km_ksem);
free(map->km_path, M_KSEM);
free(map, M_KSEM);
return (0);
}
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
return (ENOENT);
}
static int
ksem_create_copyout_semid(struct thread *td, semid_t *semidp, int fd,
int compat32)
{
semid_t semid;
#ifdef COMPAT_FREEBSD32
int32_t semid32;
#endif
void *ptr;
size_t ptrs;
#ifdef COMPAT_FREEBSD32
if (compat32) {
semid32 = fd;
ptr = &semid32;
ptrs = sizeof(semid32);
} else {
#endif
semid = fd;
ptr = &semid;
ptrs = sizeof(semid);
compat32 = 0; /* silence gcc */
#ifdef COMPAT_FREEBSD32
}
#endif
return (copyout(ptr, semidp, ptrs));
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
/* Other helper routines. */
static int
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
ksem_create(struct thread *td, const char *name, semid_t *semidp, mode_t mode,
unsigned int value, int flags, int compat32)
{
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct filedesc *fdp;
struct ksem *ks;
struct file *fp;
char *path;
Fnv32_t fnv;
int error, fd;
if (value > SEM_VALUE_MAX)
return (EINVAL);
fdp = td->td_proc->p_fd;
mode = (mode & ~fdp->fd_cmask) & ACCESSPERMS;
error = falloc(td, &fp, &fd, O_CLOEXEC);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (error) {
if (name == NULL)
error = ENOSPC;
return (error);
}
/*
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
* Go ahead and copyout the file descriptor now. This is a bit
* premature, but it is a lot easier to handle errors as opposed
* to later when we've possibly created a new semaphore, etc.
*/
error = ksem_create_copyout_semid(td, semidp, fd, compat32);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (error) {
fdclose(fdp, fp, fd, td);
fdrop(fp, td);
return (error);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (name == NULL) {
/* Create an anonymous semaphore. */
ks = ksem_alloc(td->td_ucred, mode, value);
if (ks == NULL)
error = ENOSPC;
else
ks->ks_flags |= KS_ANONYMOUS;
} else {
path = malloc(MAXPATHLEN, M_KSEM, M_WAITOK);
error = copyinstr(name, path, MAXPATHLEN, NULL);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
/* Require paths to start with a '/' character. */
if (error == 0 && path[0] != '/')
error = EINVAL;
if (error) {
fdclose(fdp, fp, fd, td);
fdrop(fp, td);
free(path, M_KSEM);
return (error);
}
fnv = fnv_32_str(path, FNV1_32_INIT);
sx_xlock(&ksem_dict_lock);
ks = ksem_lookup(path, fnv);
if (ks == NULL) {
/* Object does not exist, create it if requested. */
if (flags & O_CREAT) {
ks = ksem_alloc(td->td_ucred, mode, value);
if (ks == NULL)
error = ENFILE;
else {
ksem_insert(path, fnv, ks);
path = NULL;
}
} else
error = ENOENT;
} else {
/*
* Object already exists, obtain a new
* reference if requested and permitted.
*/
if ((flags & (O_CREAT | O_EXCL)) ==
(O_CREAT | O_EXCL))
error = EEXIST;
else {
#ifdef MAC
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = mac_posixsem_check_open(td->td_ucred,
ks);
if (error == 0)
#endif
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = ksem_access(ks, td->td_ucred);
}
if (error == 0)
ksem_hold(ks);
#ifdef INVARIANTS
else
ks = NULL;
#endif
}
sx_xunlock(&ksem_dict_lock);
if (path)
free(path, M_KSEM);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (error) {
KASSERT(ks == NULL, ("ksem_create error with a ksem"));
fdclose(fdp, fp, fd, td);
fdrop(fp, td);
return (error);
}
KASSERT(ks != NULL, ("ksem_create w/o a ksem"));
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
finit(fp, FREAD | FWRITE, DTYPE_SEM, ks, &ksem_ops);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
fdrop(fp, td);
return (0);
}
static int
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
ksem_get(struct thread *td, semid_t id, cap_rights_t *rightsp,
struct file **fpp)
{
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct ksem *ks;
struct file *fp;
int error;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = fget(td, id, rightsp, &fp);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (error)
return (EINVAL);
if (fp->f_type != DTYPE_SEM) {
fdrop(fp, td);
return (EINVAL);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
ks = fp->f_data;
if (ks->ks_flags & KS_DEAD) {
fdrop(fp, td);
return (EINVAL);
}
*fpp = fp;
return (0);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
/* System calls. */
#ifndef _SYS_SYSPROTO_H_
struct ksem_init_args {
unsigned int value;
semid_t *idp;
};
#endif
int
sys_ksem_init(struct thread *td, struct ksem_init_args *uap)
{
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
return (ksem_create(td, NULL, uap->idp, S_IRWXU | S_IRWXG, uap->value,
0, 0));
}
#ifndef _SYS_SYSPROTO_H_
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct ksem_open_args {
char *name;
int oflag;
mode_t mode;
unsigned int value;
semid_t *idp;
};
#endif
int
sys_ksem_open(struct thread *td, struct ksem_open_args *uap)
{
DP((">>> ksem_open start, pid=%d\n", (int)td->td_proc->p_pid));
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if ((uap->oflag & ~(O_CREAT | O_EXCL)) != 0)
return (EINVAL);
return (ksem_create(td, uap->name, uap->idp, uap->mode, uap->value,
uap->oflag, 0));
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
#ifndef _SYS_SYSPROTO_H_
struct ksem_unlink_args {
char *name;
};
#endif
int
sys_ksem_unlink(struct thread *td, struct ksem_unlink_args *uap)
{
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
char *path;
Fnv32_t fnv;
int error;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
error = copyinstr(uap->name, path, MAXPATHLEN, NULL);
if (error) {
free(path, M_TEMP);
return (error);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
fnv = fnv_32_str(path, FNV1_32_INIT);
sx_xlock(&ksem_dict_lock);
error = ksem_remove(path, fnv, td->td_ucred);
sx_xunlock(&ksem_dict_lock);
free(path, M_TEMP);
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct ksem_close_args {
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
semid_t id;
};
#endif
int
sys_ksem_close(struct thread *td, struct ksem_close_args *uap)
{
struct ksem *ks;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct file *fp;
int error;
/* No capability rights required to close a semaphore. */
error = ksem_get(td, uap->id, 0, &fp);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (error)
return (error);
ks = fp->f_data;
if (ks->ks_flags & KS_ANONYMOUS) {
fdrop(fp, td);
return (EINVAL);
}
error = kern_close(td, uap->id);
fdrop(fp, td);
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct ksem_post_args {
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
semid_t id;
};
#endif
int
sys_ksem_post(struct thread *td, struct ksem_post_args *uap)
{
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct file *fp;
struct ksem *ks;
int error;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = ksem_get(td, uap->id,
cap_rights_init(&rights, CAP_SEM_POST), &fp);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (error)
return (error);
ks = fp->f_data;
mtx_lock(&sem_lock);
#ifdef MAC
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = mac_posixsem_check_post(td->td_ucred, fp->f_cred, ks);
if (error)
goto err;
#endif
if (ks->ks_value == SEM_VALUE_MAX) {
error = EOVERFLOW;
goto err;
}
++ks->ks_value;
if (ks->ks_waiters > 0)
cv_signal(&ks->ks_cv);
error = 0;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
vfs_timestamp(&ks->ks_ctime);
err:
mtx_unlock(&sem_lock);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
fdrop(fp, td);
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct ksem_wait_args {
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
semid_t id;
};
#endif
int
sys_ksem_wait(struct thread *td, struct ksem_wait_args *uap)
{
return (kern_sem_wait(td, uap->id, 0, NULL));
}
#ifndef _SYS_SYSPROTO_H_
struct ksem_timedwait_args {
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
semid_t id;
const struct timespec *abstime;
};
#endif
int
sys_ksem_timedwait(struct thread *td, struct ksem_timedwait_args *uap)
{
struct timespec abstime;
struct timespec *ts;
int error;
/*
* We allow a null timespec (wait forever).
*/
if (uap->abstime == NULL)
ts = NULL;
else {
error = copyin(uap->abstime, &abstime, sizeof(abstime));
if (error != 0)
return (error);
if (abstime.tv_nsec >= 1000000000 || abstime.tv_nsec < 0)
return (EINVAL);
ts = &abstime;
}
return (kern_sem_wait(td, uap->id, 0, ts));
}
#ifndef _SYS_SYSPROTO_H_
struct ksem_trywait_args {
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
semid_t id;
};
#endif
int
sys_ksem_trywait(struct thread *td, struct ksem_trywait_args *uap)
{
return (kern_sem_wait(td, uap->id, 1, NULL));
}
static int
kern_sem_wait(struct thread *td, semid_t id, int tryflag,
struct timespec *abstime)
{
struct timespec ts1, ts2;
struct timeval tv;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct file *fp;
struct ksem *ks;
int error;
DP((">>> kern_sem_wait entered! pid=%d\n", (int)td->td_proc->p_pid));
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = ksem_get(td, id, cap_rights_init(&rights, CAP_SEM_WAIT), &fp);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (error)
return (error);
ks = fp->f_data;
mtx_lock(&sem_lock);
DP((">>> kern_sem_wait critical section entered! pid=%d\n",
(int)td->td_proc->p_pid));
#ifdef MAC
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = mac_posixsem_check_wait(td->td_ucred, fp->f_cred, ks);
if (error) {
DP(("kern_sem_wait mac failed\n"));
goto err;
}
#endif
2002-10-07 04:13:21 +00:00
DP(("kern_sem_wait value = %d, tryflag %d\n", ks->ks_value, tryflag));
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
vfs_timestamp(&ks->ks_atime);
while (ks->ks_value == 0) {
ks->ks_waiters++;
if (tryflag != 0)
error = EAGAIN;
else if (abstime == NULL)
error = cv_wait_sig(&ks->ks_cv, &sem_lock);
else {
for (;;) {
ts1 = *abstime;
getnanotime(&ts2);
timespecsub(&ts1, &ts2);
TIMESPEC_TO_TIMEVAL(&tv, &ts1);
if (tv.tv_sec < 0) {
error = ETIMEDOUT;
break;
}
error = cv_timedwait_sig(&ks->ks_cv,
&sem_lock, tvtohz(&tv));
if (error != EWOULDBLOCK)
break;
}
}
ks->ks_waiters--;
if (error)
goto err;
}
ks->ks_value--;
DP(("kern_sem_wait value post-decrement = %d\n", ks->ks_value));
error = 0;
err:
mtx_unlock(&sem_lock);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
fdrop(fp, td);
DP(("<<< kern_sem_wait leaving, pid=%d, error = %d\n",
(int)td->td_proc->p_pid, error));
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct ksem_getvalue_args {
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
semid_t id;
int *val;
};
#endif
int
sys_ksem_getvalue(struct thread *td, struct ksem_getvalue_args *uap)
{
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct file *fp;
struct ksem *ks;
int error, val;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = ksem_get(td, uap->id,
cap_rights_init(&rights, CAP_SEM_GETVALUE), &fp);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (error)
return (error);
ks = fp->f_data;
mtx_lock(&sem_lock);
#ifdef MAC
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = mac_posixsem_check_getvalue(td->td_ucred, fp->f_cred, ks);
if (error) {
mtx_unlock(&sem_lock);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
fdrop(fp, td);
return (error);
}
#endif
val = ks->ks_value;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
vfs_timestamp(&ks->ks_atime);
mtx_unlock(&sem_lock);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
fdrop(fp, td);
error = copyout(&val, uap->val, sizeof(val));
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct ksem_destroy_args {
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
semid_t id;
};
#endif
int
sys_ksem_destroy(struct thread *td, struct ksem_destroy_args *uap)
{
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
struct file *fp;
struct ksem *ks;
int error;
/* No capability rights required to close a semaphore. */
error = ksem_get(td, uap->id, 0, &fp);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
if (error)
return (error);
ks = fp->f_data;
if (!(ks->ks_flags & KS_ANONYMOUS)) {
fdrop(fp, td);
return (EINVAL);
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
mtx_lock(&sem_lock);
if (ks->ks_waiters != 0) {
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
mtx_unlock(&sem_lock);
error = EBUSY;
goto err;
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
ks->ks_flags |= KS_DEAD;
mtx_unlock(&sem_lock);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = kern_close(td, uap->id);
err:
fdrop(fp, td);
return (error);
}
static struct syscall_helper_data ksem_syscalls[] = {
SYSCALL_INIT_HELPER(ksem_init),
SYSCALL_INIT_HELPER(ksem_open),
SYSCALL_INIT_HELPER(ksem_unlink),
SYSCALL_INIT_HELPER(ksem_close),
SYSCALL_INIT_HELPER(ksem_post),
SYSCALL_INIT_HELPER(ksem_wait),
SYSCALL_INIT_HELPER(ksem_timedwait),
SYSCALL_INIT_HELPER(ksem_trywait),
SYSCALL_INIT_HELPER(ksem_getvalue),
SYSCALL_INIT_HELPER(ksem_destroy),
SYSCALL_INIT_LAST
};
#ifdef COMPAT_FREEBSD32
#include <compat/freebsd32/freebsd32.h>
#include <compat/freebsd32/freebsd32_proto.h>
#include <compat/freebsd32/freebsd32_signal.h>
#include <compat/freebsd32/freebsd32_syscall.h>
#include <compat/freebsd32/freebsd32_util.h>
int
freebsd32_ksem_init(struct thread *td, struct freebsd32_ksem_init_args *uap)
{
return (ksem_create(td, NULL, uap->idp, S_IRWXU | S_IRWXG, uap->value,
0, 1));
}
int
freebsd32_ksem_open(struct thread *td, struct freebsd32_ksem_open_args *uap)
{
if ((uap->oflag & ~(O_CREAT | O_EXCL)) != 0)
return (EINVAL);
return (ksem_create(td, uap->name, uap->idp, uap->mode, uap->value,
uap->oflag, 1));
}
int
freebsd32_ksem_timedwait(struct thread *td,
struct freebsd32_ksem_timedwait_args *uap)
{
struct timespec32 abstime32;
struct timespec *ts, abstime;
int error;
/*
* We allow a null timespec (wait forever).
*/
if (uap->abstime == NULL)
ts = NULL;
else {
error = copyin(uap->abstime, &abstime32, sizeof(abstime32));
if (error != 0)
return (error);
CP(abstime32, abstime, tv_sec);
CP(abstime32, abstime, tv_nsec);
if (abstime.tv_nsec >= 1000000000 || abstime.tv_nsec < 0)
return (EINVAL);
ts = &abstime;
}
return (kern_sem_wait(td, uap->id, 0, ts));
}
static struct syscall_helper_data ksem32_syscalls[] = {
SYSCALL32_INIT_HELPER(freebsd32_ksem_init),
SYSCALL32_INIT_HELPER(freebsd32_ksem_open),
SYSCALL32_INIT_HELPER_COMPAT(ksem_unlink),
SYSCALL32_INIT_HELPER_COMPAT(ksem_close),
SYSCALL32_INIT_HELPER_COMPAT(ksem_post),
SYSCALL32_INIT_HELPER_COMPAT(ksem_wait),
SYSCALL32_INIT_HELPER(freebsd32_ksem_timedwait),
SYSCALL32_INIT_HELPER_COMPAT(ksem_trywait),
SYSCALL32_INIT_HELPER_COMPAT(ksem_getvalue),
SYSCALL32_INIT_HELPER_COMPAT(ksem_destroy),
SYSCALL_INIT_LAST
};
#endif
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
static int
ksem_module_init(void)
{
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
int error;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
mtx_init(&sem_lock, "sem", NULL, MTX_DEF);
mtx_init(&ksem_count_lock, "ksem count", NULL, MTX_DEF);
sx_init(&ksem_dict_lock, "ksem dictionary");
ksem_dictionary = hashinit(1024, M_KSEM, &ksem_hash);
p31b_setcfg(CTL_P1003_1B_SEMAPHORES, 200112L);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
p31b_setcfg(CTL_P1003_1B_SEM_NSEMS_MAX, SEM_MAX);
p31b_setcfg(CTL_P1003_1B_SEM_VALUE_MAX, SEM_VALUE_MAX);
error = syscall_helper_register(ksem_syscalls, SY_THR_STATIC_KLD);
if (error)
return (error);
#ifdef COMPAT_FREEBSD32
error = syscall32_helper_register(ksem32_syscalls, SY_THR_STATIC_KLD);
if (error)
return (error);
#endif
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
return (0);
}
static void
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
ksem_module_destroy(void)
{
#ifdef COMPAT_FREEBSD32
syscall32_helper_unregister(ksem32_syscalls);
#endif
syscall_helper_unregister(ksem_syscalls);
p31b_setcfg(CTL_P1003_1B_SEMAPHORES, 0);
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
hashdestroy(ksem_dictionary, M_KSEM, ksem_hash);
sx_destroy(&ksem_dict_lock);
mtx_destroy(&ksem_count_lock);
mtx_destroy(&sem_lock);
p31b_unsetcfg(CTL_P1003_1B_SEM_VALUE_MAX);
p31b_unsetcfg(CTL_P1003_1B_SEM_NSEMS_MAX);
}
static int
sem_modload(struct module *module, int cmd, void *arg)
{
int error = 0;
switch (cmd) {
case MOD_LOAD:
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
error = ksem_module_init();
if (error)
ksem_module_destroy();
break;
case MOD_UNLOAD:
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
mtx_lock(&ksem_count_lock);
if (nsems != 0) {
error = EOPNOTSUPP;
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
mtx_unlock(&ksem_count_lock);
break;
}
Rework the lifetime management of the kernel implementation of POSIX semaphores. Specifically, semaphores are now represented as new file descriptor type that is set to close on exec. This removes the need for all of the manual process reference counting (and fork, exec, and exit event handlers) as the normal file descriptor operations handle all of that for us nicely. It is also suggested as one possible implementation in the spec and at least one other OS (OS X) uses this approach. Some bugs that were fixed as a result include: - References to a named semaphore whose name is removed still work after the sem_unlink() operation. Prior to this patch, if a semaphore's name was removed, valid handles from sem_open() would get EINVAL errors from sem_getvalue(), sem_post(), etc. This fixes that. - Unnamed semaphores created with sem_init() were not cleaned up when a process exited or exec'd. They were only cleaned up if the process did an explicit sem_destroy(). This could result in a leak of semaphore objects that could never be cleaned up. - On the other hand, if another process guessed the id (kernel pointer to 'struct ksem' of an unnamed semaphore (created via sem_init)) and had write access to the semaphore based on UID/GID checks, then that other process could manipulate the semaphore via sem_destroy(), sem_post(), sem_wait(), etc. - As part of the permission check (UID/GID), the umask of the proces creating the semaphore was not honored. Thus if your umask denied group read/write access but the explicit mode in the sem_init() call allowed it, the semaphore would be readable/writable by other users in the same group, for example. This includes access via the previous bug. - If the module refused to unload because there were active semaphores, then it might have deregistered one or more of the semaphore system calls before it noticed that there was a problem. I'm not sure if this actually happened as the order that modules are discovered by the kernel linker depends on how the actual .ko file is linked. One can make the order deterministic by using a single module with a mod_event handler that explicitly registers syscalls (and deregisters during unload after any checks). This also fixes a race where even if the sem_module unloaded first it would have destroyed locks that the syscalls might be trying to access if they are still executing when they are unloaded. XXX: By the way, deregistering system calls doesn't do any blocking to drain any threads from the calls. - Some minor fixes to errno values on error. For example, sem_init() isn't documented to return ENFILE or EMFILE if we run out of semaphores the way that sem_open() can. Instead, it should return ENOSPC in that case. Other changes: - Kernel semaphores now use a hash table to manage the namespace of named semaphores nearly in a similar fashion to the POSIX shared memory object file descriptors. Kernel semaphores can now also have names longer than 14 chars (up to MAXPATHLEN) and can include subdirectories in their pathname. - The UID/GID permission checks for access to a named semaphore are now done via vaccess() rather than a home-rolled set of checks. - Now that kernel semaphores have an associated file object, the various MAC checks for POSIX semaphores accept both a file credential and an active credential. There is also a new posixsem_check_stat() since it is possible to fstat() a semaphore file descriptor. - A small set of regression tests (using the ksem API directly) is present in src/tools/regression/posixsem. Reported by: kris (1) Tested by: kris Reviewed by: rwatson (lightly) MFC after: 1 month
2008-06-27 05:39:04 +00:00
ksem_dead = 1;
mtx_unlock(&ksem_count_lock);
ksem_module_destroy();
break;
case MOD_SHUTDOWN:
break;
default:
error = EINVAL;
break;
}
return (error);
}
static moduledata_t sem_mod = {
"sem",
&sem_modload,
NULL
};
DECLARE_MODULE(sem, sem_mod, SI_SUB_SYSV_SEM, SI_ORDER_FIRST);
MODULE_VERSION(sem, 1);