From owner-svn-src-head@FreeBSD.ORG Mon May 25 02:42:18 2015 Return-Path: Delivered-To: svn-src-head@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [8.8.178.115]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id 9EA61F71; Mon, 25 May 2015 02:42:18 +0000 (UTC) (envelope-from julian@freebsd.org) Received: from vps1.elischer.org (vps1.elischer.org [204.109.63.16]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client CN "vps1.elischer.org", Issuer "CA Cert Signing Authority" (not verified)) by mx1.freebsd.org (Postfix) with ESMTPS id 7377DF50; Mon, 25 May 2015 02:42:18 +0000 (UTC) (envelope-from julian@freebsd.org) Received: from jre-mbp.elischer.org (ppp121-45-225-35.lns20.per1.internode.on.net [121.45.225.35]) (authenticated bits=0) by vps1.elischer.org (8.14.9/8.14.9) with ESMTP id t4P2gC7H059269 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES128-SHA bits=128 verify=NO); Sun, 24 May 2015 19:42:15 -0700 (PDT) (envelope-from julian@freebsd.org) Message-ID: <55628BFE.5070306@freebsd.org> Date: Mon, 25 May 2015 10:42:06 +0800 From: Julian Elischer User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:31.0) Gecko/20100101 Thunderbird/31.7.0 MIME-Version: 1.0 To: Dmitry Chagin , src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r283441 - in head/sys: amd64/linux amd64/linux32 compat/linux conf i386/linux modules/linux modules/linux64 References: <201505241641.t4OGfeAX094242@svn.freebsd.org> In-Reply-To: <201505241641.t4OGfeAX094242@svn.freebsd.org> Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit X-BeenThere: svn-src-head@freebsd.org X-Mailman-Version: 2.1.20 Precedence: list List-Id: SVN commit messages for the src tree for head/-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 25 May 2015 02:42:18 -0000 On 5/25/15 12:41 AM, Dmitry Chagin wrote: > Author: dchagin > Date: Sun May 24 16:41:39 2015 > New Revision: 283441 > URL: https://svnweb.freebsd.org/changeset/base/283441 > > Log: > Implement epoll family system calls. This is a tiny wrapper > around kqueue() to implement epoll subset of functionality. > The kqueue user data are 32bit on i386 which is not enough for > epoll user data, so we keep user data in the proc emuldata. Have you considered making the in-kernel representation just have more room? > > Initial patch developed by rdivacky@ in 2007, then extended > by Yuri Victorovich @ r255672 and finished by me > in collaboration with mjg@ and jillies@. > > Differential Revision: https://reviews.freebsd.org/D1092 > > Added: > head/sys/compat/linux/linux_event.c (contents, props changed) > head/sys/compat/linux/linux_event.h (contents, props changed) > Modified: > head/sys/amd64/linux/linux_dummy.c > head/sys/amd64/linux/syscalls.master > head/sys/amd64/linux32/linux32_dummy.c > head/sys/amd64/linux32/syscalls.master > head/sys/compat/linux/linux_emul.c > head/sys/compat/linux/linux_emul.h > head/sys/compat/linux/linux_util.c > head/sys/compat/linux/linux_util.h > head/sys/conf/files.amd64 > head/sys/conf/files.i386 > head/sys/conf/files.pc98 > head/sys/i386/linux/linux_dummy.c > head/sys/i386/linux/syscalls.master > head/sys/modules/linux/Makefile > head/sys/modules/linux64/Makefile > > Modified: head/sys/amd64/linux/linux_dummy.c > ============================================================================== > --- head/sys/amd64/linux/linux_dummy.c Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/amd64/linux/linux_dummy.c Sun May 24 16:41:39 2015 (r283441) > @@ -69,13 +69,10 @@ DUMMY(tuxcall); > DUMMY(security); > DUMMY(set_thread_area); > DUMMY(lookup_dcookie); > -DUMMY(epoll_create); > DUMMY(epoll_ctl_old); > DUMMY(epoll_wait_old); > DUMMY(remap_file_pages); > DUMMY(semtimedop); > -DUMMY(epoll_ctl); > -DUMMY(epoll_wait); > DUMMY(mbind); > DUMMY(get_mempolicy); > DUMMY(set_mempolicy); > @@ -112,7 +109,6 @@ DUMMY(timerfd_settime); > DUMMY(timerfd_gettime); > DUMMY(signalfd4); > DUMMY(eventfd2); > -DUMMY(epoll_create1); > DUMMY(inotify_init1); > DUMMY(preadv); > DUMMY(pwritev); > > Modified: head/sys/amd64/linux/syscalls.master > ============================================================================== > --- head/sys/amd64/linux/syscalls.master Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/amd64/linux/syscalls.master Sun May 24 16:41:39 2015 (r283441) > @@ -373,7 +373,7 @@ > 210 AUE_NULL UNIMPL linux_io_cancel > 211 AUE_NULL UNIMPL linux_get_thread_area > 212 AUE_NULL STD { int linux_lookup_dcookie(void); } > -213 AUE_NULL STD { int linux_epoll_create(void); } > +213 AUE_NULL STD { int linux_epoll_create(l_int size); } > 214 AUE_NULL STD { int linux_epoll_ctl_old(void); } > 215 AUE_NULL STD { int linux_epoll_wait_old(void); } > 216 AUE_NULL STD { int linux_remap_file_pages(void); } > @@ -397,8 +397,10 @@ > 230 AUE_NULL STD { int linux_clock_nanosleep(clockid_t which, int flags, \ > struct l_timespec *rqtp, struct l_timespec *rmtp); } > 231 AUE_EXIT STD { int linux_exit_group(int error_code); } > -232 AUE_NULL STD { int linux_epoll_wait(void); } > -233 AUE_NULL STD { int linux_epoll_ctl(void); } > +232 AUE_NULL STD { int linux_epoll_wait(l_int epfd, struct epoll_event *events, \ > + l_int maxevents, l_int timeout); } > +233 AUE_NULL STD { int linux_epoll_ctl(l_int epfd, l_int op, l_int fd, \ > + struct epoll_event *event); } > 234 AUE_NULL STD { int linux_tgkill(int tgid, int pid, int sig); } > 235 AUE_UTIMES STD { int linux_utimes(char *fname, \ > struct l_timeval *tptr); } > @@ -466,7 +468,8 @@ > 278 AUE_NULL STD { int linux_vmsplice(void); } > 279 AUE_NULL STD { int linux_move_pages(void); } > 280 AUE_NULL STD { int linux_utimensat(void); } > -281 AUE_NULL STD { int linux_epoll_pwait(void); } > +281 AUE_NULL STD { int linux_epoll_pwait(l_int epfd, struct epoll_event *events, \ > + l_int maxevents, l_int timeout, l_sigset_t *mask); } > 282 AUE_NULL STD { int linux_signalfd(void); } > 283 AUE_NULL STD { int linux_timerfd(void); } > 284 AUE_NULL STD { int linux_eventfd(void); } > @@ -477,7 +480,7 @@ > l_uintptr_t namelen, int flags); } > 289 AUE_NULL STD { int linux_signalfd4(void); } > 290 AUE_NULL STD { int linux_eventfd2(void); } > -291 AUE_NULL STD { int linux_epoll_create1(void); } > +291 AUE_NULL STD { int linux_epoll_create1(l_int flags); } > 292 AUE_NULL STD { int linux_dup3(l_int oldfd, \ > l_int newfd, l_int flags); } > 293 AUE_NULL STD { int linux_pipe2(l_int *pipefds, l_int flags); } > > Modified: head/sys/amd64/linux32/linux32_dummy.c > ============================================================================== > --- head/sys/amd64/linux32/linux32_dummy.c Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/amd64/linux32/linux32_dummy.c Sun May 24 16:41:39 2015 (r283441) > @@ -68,9 +68,6 @@ DUMMY(pivot_root); > DUMMY(mincore); > DUMMY(ptrace); > DUMMY(lookup_dcookie); > -DUMMY(epoll_create); > -DUMMY(epoll_ctl); > -DUMMY(epoll_wait); > DUMMY(remap_file_pages); > DUMMY(fstatfs64); > DUMMY(mbind); > @@ -120,7 +117,6 @@ DUMMY(timerfd_gettime); > /* linux 2.6.27: */ > DUMMY(signalfd4); > DUMMY(eventfd2); > -DUMMY(epoll_create1); > DUMMY(inotify_init1); > /* linux 2.6.30: */ > DUMMY(preadv); > > Modified: head/sys/amd64/linux32/syscalls.master > ============================================================================== > --- head/sys/amd64/linux32/syscalls.master Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/amd64/linux32/syscalls.master Sun May 24 16:41:39 2015 (r283441) > @@ -430,9 +430,11 @@ > 251 AUE_NULL UNIMPL > 252 AUE_EXIT STD { int linux_exit_group(int error_code); } > 253 AUE_NULL STD { int linux_lookup_dcookie(void); } > -254 AUE_NULL STD { int linux_epoll_create(void); } > -255 AUE_NULL STD { int linux_epoll_ctl(void); } > -256 AUE_NULL STD { int linux_epoll_wait(void); } > +254 AUE_NULL STD { int linux_epoll_create(l_int size); } > +255 AUE_NULL STD { int linux_epoll_ctl(l_int epfd, l_int op, l_int fd, \ > + struct epoll_event *event); } > +256 AUE_NULL STD { int linux_epoll_wait(l_int epfd, struct epoll_event *events, \ > + l_int maxevents, l_int timeout); } > 257 AUE_NULL STD { int linux_remap_file_pages(void); } > 258 AUE_NULL STD { int linux_set_tid_address(int *tidptr); } > 259 AUE_NULL STD { int linux_timer_create(clockid_t clock_id, \ > @@ -527,7 +529,8 @@ > 317 AUE_NULL STD { int linux_move_pages(void); } > ; linux 2.6.19: > 318 AUE_NULL STD { int linux_getcpu(void); } > -319 AUE_NULL STD { int linux_epoll_pwait(void); } > +319 AUE_NULL STD { int linux_epoll_pwait(l_int epfd, struct epoll_event *events, \ > + l_int maxevents, l_int timeout, l_osigset_t *mask); } > ; linux 2.6.22: > 320 AUE_NULL STD { int linux_utimensat(void); } > 321 AUE_NULL STD { int linux_signalfd(void); } > @@ -541,7 +544,7 @@ > ; linux 2.6.27: > 327 AUE_NULL STD { int linux_signalfd4(void); } > 328 AUE_NULL STD { int linux_eventfd2(void); } > -329 AUE_NULL STD { int linux_epoll_create1(void); } > +329 AUE_NULL STD { int linux_epoll_create1(l_int flags); } > 330 AUE_NULL STD { int linux_dup3(l_int oldfd, \ > l_int newfd, l_int flags); } > 331 AUE_NULL STD { int linux_pipe2(l_int *pipefds, l_int flags); } > > Modified: head/sys/compat/linux/linux_emul.c > ============================================================================== > --- head/sys/compat/linux/linux_emul.c Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/compat/linux/linux_emul.c Sun May 24 16:41:39 2015 (r283441) > @@ -42,8 +42,6 @@ __FBSDID("$FreeBSD$"); > #include > #include > #include > -#include > -#include > > #include > #include > @@ -86,6 +84,7 @@ linux_proc_init(struct thread *td, struc > { > struct linux_emuldata *em; > struct linux_pemuldata *pem; > + struct epoll_emuldata *emd; > > if (newtd != NULL) { > /* non-exec call */ > @@ -93,8 +92,13 @@ linux_proc_init(struct thread *td, struc > em->pdeath_signal = 0; > em->robust_futexes = NULL; > if (flags & LINUX_CLONE_THREAD) { > + LINUX_CTR1(proc_init, "thread newtd(%d)", > + newtd->td_tid); > + > em->em_tid = newtd->td_tid; > } else { > + LINUX_CTR1(proc_init, "fork newtd(%d)", > + newtd->td_proc->p_pid); > > em->em_tid = newtd->td_proc->p_pid; > > @@ -105,12 +109,24 @@ linux_proc_init(struct thread *td, struc > newtd->td_emuldata = em; > } else { > /* exec */ > + LINUX_CTR1(proc_init, "exec newtd(%d)", > + td->td_proc->p_pid); > > /* lookup the old one */ > em = em_find(td); > KASSERT(em != NULL, ("proc_init: emuldata not found in exec case.\n")); > > em->em_tid = td->td_proc->p_pid; > + > + /* epoll should be destroyed in a case of exec. */ > + pem = pem_find(td->td_proc); > + KASSERT(pem != NULL, ("proc_exit: proc emuldata not found.\n")); > + > + if (pem->epoll != NULL) { > + emd = pem->epoll; > + pem->epoll = NULL; > + free(emd, M_EPOLL); > + } > } > > em->child_clear_tid = NULL; > @@ -121,6 +137,7 @@ void > linux_proc_exit(void *arg __unused, struct proc *p) > { > struct linux_pemuldata *pem; > + struct epoll_emuldata *emd; > struct thread *td = curthread; > > if (__predict_false(SV_CURPROC_ABI() != SV_ABI_LINUX)) > @@ -133,6 +150,12 @@ linux_proc_exit(void *arg __unused, stru > > p->p_emuldata = NULL; > > + if (pem->epoll != NULL) { > + emd = pem->epoll; > + pem->epoll = NULL; > + free(emd, M_EPOLL); > + } > + > sx_destroy(&pem->pem_sx); > free(pem, M_LINUX); > } > @@ -141,6 +164,7 @@ int > linux_common_execve(struct thread *td, struct image_args *eargs) > { > struct linux_pemuldata *pem; > + struct epoll_emuldata *emd; > struct linux_emuldata *em; > struct proc *p; > int error; > @@ -180,6 +204,12 @@ linux_common_execve(struct thread *td, s > p->p_emuldata = NULL; > PROC_UNLOCK(p); > > + if (pem->epoll != NULL) { > + emd = pem->epoll; > + pem->epoll = NULL; > + free(emd, M_EPOLL); > + } > + > free(em, M_TEMP); > free(pem, M_LINUX); > } > @@ -197,6 +227,7 @@ linux_proc_exec(void *arg __unused, stru > */ > if (__predict_false((imgp->sysent->sv_flags & SV_ABI_MASK) == > SV_ABI_LINUX)) { > + > if (SV_PROC_ABI(p) == SV_ABI_LINUX) > linux_proc_init(td, NULL, 0); > else > > Modified: head/sys/compat/linux/linux_emul.h > ============================================================================== > --- head/sys/compat/linux/linux_emul.h Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/compat/linux/linux_emul.h Sun May 24 16:41:39 2015 (r283441) > @@ -60,9 +60,12 @@ int linux_common_execve(struct thread *, > /* process emuldata flags */ > #define LINUX_XDEPR_REQUEUEOP 0x00000001 /* uses deprecated > futex REQUEUE op*/ > +#define LINUX_XUNSUP_EPOLL 0x00000002 /* unsupported epoll events */ > + > struct linux_pemuldata { > uint32_t flags; /* process emuldata flags */ > struct sx pem_sx; /* lock for this struct */ > + void *epoll; /* epoll data */ > }; > > #define LINUX_PEM_XLOCK(p) sx_xlock(&(p)->pem_sx) > > Added: head/sys/compat/linux/linux_event.c > ============================================================================== > --- /dev/null 00:00:00 1970 (empty, because file is newly added) > +++ head/sys/compat/linux/linux_event.c Sun May 24 16:41:39 2015 (r283441) > @@ -0,0 +1,500 @@ > +/*- > + * Copyright (c) 2007 Roman Divacky > + * Copyright (c) 2014 Dmitry Chagin > + * All rights reserved. > + * > + * 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. > + */ > + > +#include > +__FBSDID("$FreeBSD$"); > + > +#include "opt_compat.h" > + > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > + > +#ifdef COMPAT_LINUX32 > +#include > +#include > +#else > +#include > +#include > +#endif > + > +#include > +#include > +#include > +#include > + > +/* > + * epoll defines 'struct epoll_event' with the field 'data' as 64 bits > + * on all architectures. But on 32 bit architectures BSD 'struct kevent' only > + * has 32 bit opaque pointer as 'udata' field. So we can't pass epoll supplied > + * data verbatuim. Therefore we allocate 64-bit memory block to pass > + * user supplied data for every file descriptor. > + */ > + > +typedef uint64_t epoll_udata_t; > + > +struct epoll_emuldata { > + uint32_t fdc; /* epoll udata max index */ > + epoll_udata_t udata[1]; /* epoll user data vector */ > +}; > + > +#define EPOLL_DEF_SZ 16 > +#define EPOLL_SIZE(fdn) \ > + (sizeof(struct epoll_emuldata)+(fdn) * sizeof(epoll_udata_t)) > + > +struct epoll_event { > + uint32_t events; > + epoll_udata_t data; > +} > +#if defined(__amd64__) > +__attribute__((packed)) > +#endif > +; > + > +#define LINUX_MAX_EVENTS (INT_MAX / sizeof(struct epoll_event)) > + > +static void epoll_fd_install(struct thread *td, int fd, epoll_udata_t udata); > +static int epoll_to_kevent(struct thread *td, struct file *epfp, > + int fd, struct epoll_event *l_event, int *kev_flags, > + struct kevent *kevent, int *nkevents); > +static void kevent_to_epoll(struct kevent *kevent, struct epoll_event *l_event); > +static int epoll_kev_copyout(void *arg, struct kevent *kevp, int count); > +static int epoll_kev_copyin(void *arg, struct kevent *kevp, int count); > +static int epoll_delete_event(struct thread *td, struct file *epfp, > + int fd, int filter); > +static int epoll_delete_all_events(struct thread *td, struct file *epfp, > + int fd); > + > +struct epoll_copyin_args { > + struct kevent *changelist; > +}; > + > +struct epoll_copyout_args { > + struct epoll_event *leventlist; > + struct proc *p; > + uint32_t count; > + int error; > +}; > + > + > +static void > +epoll_fd_install(struct thread *td, int fd, epoll_udata_t udata) > +{ > + struct linux_pemuldata *pem; > + struct epoll_emuldata *emd; > + struct proc *p; > + > + p = td->td_proc; > + > + pem = pem_find(p); > + KASSERT(pem != NULL, ("epoll proc emuldata not found.\n")); > + > + LINUX_PEM_XLOCK(pem); > + if (pem->epoll == NULL) { > + emd = malloc(EPOLL_SIZE(fd), M_EPOLL, M_WAITOK); > + emd->fdc = fd; > + pem->epoll = emd; > + } else { > + emd = pem->epoll; > + if (fd > emd->fdc) { > + emd = realloc(emd, EPOLL_SIZE(fd), M_EPOLL, M_WAITOK); > + emd->fdc = fd; > + pem->epoll = emd; > + } > + } > + emd->udata[fd] = udata; > + LINUX_PEM_XUNLOCK(pem); > +} > + > +static int > +epoll_create_common(struct thread *td, int flags) > +{ > + int error; > + > + error = kern_kqueue(td, flags); > + if (error) > + return (error); > + > + epoll_fd_install(td, EPOLL_DEF_SZ, 0); > + > + return (0); > +} > + > +int > +linux_epoll_create(struct thread *td, struct linux_epoll_create_args *args) > +{ > + > + /* > + * args->size is unused. Linux just tests it > + * and then forgets it as well. > + */ > + if (args->size <= 0) > + return (EINVAL); > + > + return (epoll_create_common(td, 0)); > +} > + > +int > +linux_epoll_create1(struct thread *td, struct linux_epoll_create1_args *args) > +{ > + int flags; > + > + if ((args->flags & ~(LINUX_O_CLOEXEC)) != 0) > + return (EINVAL); > + > + flags = 0; > + if ((args->flags & LINUX_O_CLOEXEC) != 0) > + flags |= O_CLOEXEC; > + > + return (epoll_create_common(td, flags)); > +} > + > +/* Structure converting function from epoll to kevent. */ > +static int > +epoll_to_kevent(struct thread *td, struct file *epfp, > + int fd, struct epoll_event *l_event, int *kev_flags, > + struct kevent *kevent, int *nkevents) > +{ > + uint32_t levents = l_event->events; > + struct linux_pemuldata *pem; > + struct proc *p; > + > + /* flags related to how event is registered */ > + if ((levents & LINUX_EPOLLONESHOT) != 0) > + *kev_flags |= EV_ONESHOT; > + if ((levents & LINUX_EPOLLET) != 0) > + *kev_flags |= EV_CLEAR; > + > + /* flags related to what event is registered */ > + if ((levents & LINUX_EPOLL_EVRD) != 0) { > + EV_SET(kevent++, fd, EVFILT_READ, *kev_flags, 0, 0, 0); > + ++(*nkevents); > + } > + if ((levents & LINUX_EPOLL_EVWR) != 0) { > + EV_SET(kevent++, fd, EVFILT_WRITE, *kev_flags, 0, 0, 0); > + ++(*nkevents); > + } > + > + if ((levents & ~(LINUX_EPOLL_EVSUP)) != 0) { > + p = td->td_proc; > + > + pem = pem_find(p); > + KASSERT(pem != NULL, ("epoll proc emuldata not found.\n")); > + KASSERT(pem->epoll != NULL, ("epoll proc epolldata not found.\n")); > + > + LINUX_PEM_XLOCK(pem); > + if ((pem->flags & LINUX_XUNSUP_EPOLL) == 0) { > + pem->flags |= LINUX_XUNSUP_EPOLL; > + LINUX_PEM_XUNLOCK(pem); > + linux_msg(td, "epoll_ctl unsupported flags: 0x%x\n", > + levents); > + } else > + LINUX_PEM_XUNLOCK(pem); > + return (EINVAL); > + } > + > + return (0); > +} > + > +/* > + * Structure converting function from kevent to epoll. In a case > + * this is called on error in registration we store the error in > + * event->data and pick it up later in linux_epoll_ctl(). > + */ > +static void > +kevent_to_epoll(struct kevent *kevent, struct epoll_event *l_event) > +{ > + > + if ((kevent->flags & EV_ERROR) != 0) > + return; > + > + switch (kevent->filter) { > + case EVFILT_READ: > + l_event->events = LINUX_EPOLLIN|LINUX_EPOLLRDNORM|LINUX_EPOLLPRI; > + break; > + case EVFILT_WRITE: > + l_event->events = LINUX_EPOLLOUT|LINUX_EPOLLWRNORM; > + break; > + } > +} > + > +/* > + * Copyout callback used by kevent. This converts kevent > + * events to epoll events and copies them back to the > + * userspace. This is also called on error on registering > + * of the filter. > + */ > +static int > +epoll_kev_copyout(void *arg, struct kevent *kevp, int count) > +{ > + struct epoll_copyout_args *args; > + struct linux_pemuldata *pem; > + struct epoll_emuldata *emd; > + struct epoll_event *eep; > + int error, fd, i; > + > + args = (struct epoll_copyout_args*) arg; > + eep = malloc(sizeof(*eep) * count, M_EPOLL, M_WAITOK | M_ZERO); > + > + pem = pem_find(args->p); > + KASSERT(pem != NULL, ("epoll proc emuldata not found.\n")); > + LINUX_PEM_SLOCK(pem); > + emd = pem->epoll; > + KASSERT(emd != NULL, ("epoll proc epolldata not found.\n")); > + > + for (i = 0; i < count; i++) { > + kevent_to_epoll(&kevp[i], &eep[i]); > + > + fd = kevp[i].ident; > + KASSERT(fd <= emd->fdc, ("epoll user data vector" > + " is too small.\n")); > + eep[i].data = emd->udata[fd]; > + } > + LINUX_PEM_SUNLOCK(pem); > + > + error = copyout(eep, args->leventlist, count * sizeof(*eep)); > + if (error == 0) { > + args->leventlist += count; > + args->count += count; > + } else if (args->error == 0) > + args->error = error; > + > + free(eep, M_EPOLL); > + return (error); > +} > + > +/* > + * Copyin callback used by kevent. This copies already > + * converted filters from kernel memory to the kevent > + * internal kernel memory. Hence the memcpy instead of > + * copyin. > + */ > +static int > +epoll_kev_copyin(void *arg, struct kevent *kevp, int count) > +{ > + struct epoll_copyin_args *args; > + > + args = (struct epoll_copyin_args*) arg; > + > + memcpy(kevp, args->changelist, count * sizeof(*kevp)); > + args->changelist += count; > + > + return (0); > +} > + > +/* > + * Load epoll filter, convert it to kevent filter > + * and load it into kevent subsystem. > + */ > +int > +linux_epoll_ctl(struct thread *td, struct linux_epoll_ctl_args *args) > +{ > + struct file *epfp, *fp; > + struct epoll_copyin_args ciargs; > + struct kevent kev[2]; > + struct kevent_copyops k_ops = { &ciargs, > + NULL, > + epoll_kev_copyin}; > + struct epoll_event le; > + cap_rights_t rights; > + int kev_flags; > + int nchanges = 0; > + int error; > + > + if (args->op != LINUX_EPOLL_CTL_DEL) { > + error = copyin(args->event, &le, sizeof(le)); > + if (error != 0) > + return (error); > + } > + > + error = fget(td, args->epfd, > + cap_rights_init(&rights, CAP_KQUEUE_CHANGE), &epfp); > + if (error != 0) > + return (error); > + if (epfp->f_type != DTYPE_KQUEUE) > + goto leave1; > + > + /* Protect user data vector from incorrectly supplied fd. */ > + error = fget(td, args->fd, cap_rights_init(&rights, CAP_POLL_EVENT), &fp); > + if (error != 0) > + goto leave1; > + > + /* Linux disallows spying on himself */ > + if (epfp == fp) { > + error = EINVAL; > + goto leave0; > + } > + > + ciargs.changelist = kev; > + > + switch (args->op) { > + case LINUX_EPOLL_CTL_MOD: > + /* > + * We don't memorize which events were set for this FD > + * on this level, so just delete all we could have set: > + * EVFILT_READ and EVFILT_WRITE, ignoring any errors > + */ > + error = epoll_delete_all_events(td, epfp, args->fd); > + if (error) > + goto leave0; > + /* FALLTHROUGH */ > + > + case LINUX_EPOLL_CTL_ADD: > + kev_flags = EV_ADD | EV_ENABLE; > + break; > + > + case LINUX_EPOLL_CTL_DEL: > + /* CTL_DEL means unregister this fd with this epoll */ > + error = epoll_delete_all_events(td, epfp, args->fd); > + goto leave0; > + > + default: > + error = EINVAL; > + goto leave0; > + } > + > + error = epoll_to_kevent(td, epfp, args->fd, &le, &kev_flags, > + kev, &nchanges); > + if (error) > + goto leave0; > + > + epoll_fd_install(td, args->fd, le.data); > + > + error = kern_kevent_fp(td, epfp, nchanges, 0, &k_ops, NULL); > + > +leave0: > + fdrop(fp, td); > + > +leave1: > + fdrop(epfp, td); > + return (error); > +} > + > +/* > + * Wait for a filter to be triggered on the epoll file descriptor. > + */ > +int > +linux_epoll_wait(struct thread *td, struct linux_epoll_wait_args *args) > +{ > + struct file *epfp; > + struct timespec ts, *tsp; > + cap_rights_t rights; > + struct epoll_copyout_args coargs; > + struct kevent_copyops k_ops = { &coargs, > + epoll_kev_copyout, > + NULL}; > + int error; > + > + if (args->maxevents <= 0 || args->maxevents > LINUX_MAX_EVENTS) > + return (EINVAL); > + > + error = fget(td, args->epfd, > + cap_rights_init(&rights, CAP_KQUEUE_EVENT), &epfp); > + if (error != 0) > + return (error); > + > + coargs.leventlist = args->events; > + coargs.p = td->td_proc; > + coargs.count = 0; > + coargs.error = 0; > + > + if (args->timeout != -1) { > + if (args->timeout < 0) { > + error = EINVAL; > + goto leave; > + } > + /* Convert from milliseconds to timespec. */ > + ts.tv_sec = args->timeout / 1000; > + ts.tv_nsec = (args->timeout % 1000) * 1000000; > + tsp = &ts; > + } else { > + tsp = NULL; > + } > + > + error = kern_kevent_fp(td, epfp, 0, args->maxevents, &k_ops, tsp); > + if (error == 0 && coargs.error != 0) > + error = coargs.error; > + > + /* > + * kern_kevent might return ENOMEM which is not expected from epoll_wait. > + * Maybe we should translate that but I don't think it matters at all. > + */ > + if (error == 0) > + td->td_retval[0] = coargs.count; > +leave: > + fdrop(epfp, td); > + return (error); > +} > + > +static int > +epoll_delete_event(struct thread *td, struct file *epfp, int fd, int filter) > +{ > + struct epoll_copyin_args ciargs; > + struct kevent kev; > + struct kevent_copyops k_ops = { &ciargs, > + NULL, > + epoll_kev_copyin}; > + int error; > + > + ciargs.changelist = &kev; > + EV_SET(&kev, fd, filter, EV_DELETE | EV_DISABLE, 0, 0, 0); > + > + error = kern_kevent_fp(td, epfp, 1, 0, &k_ops, NULL); > + > + /* > + * here we ignore ENONT, because we don't keep track of events here > + */ > + if (error == ENOENT) > + error = 0; > + return (error); > +} > + > +static int > +epoll_delete_all_events(struct thread *td, struct file *epfp, int fd) > +{ > + int error1, error2; > + > + error1 = epoll_delete_event(td, epfp, fd, EVFILT_READ); > + error2 = epoll_delete_event(td, epfp, fd, EVFILT_WRITE); > + > + /* report any errors we got */ > + return (error1 == 0 ? error2 : error1); > +} > > Added: head/sys/compat/linux/linux_event.h > ============================================================================== > --- /dev/null 00:00:00 1970 (empty, because file is newly added) > +++ head/sys/compat/linux/linux_event.h Sun May 24 16:41:39 2015 (r283441) > @@ -0,0 +1,58 @@ > +/*- > + * Copyright (c) 2007 Roman Divacky > + * Copyright (c) 2014 Dmitry Chagin > + * All rights reserved. > + * > + * 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. > + * > + * $FreeBSD$ > + */ > + > +#ifndef _LINUX_EVENT_H_ > +#define _LINUX_EVENT_H_ > + > +#define LINUX_EPOLLIN 0x001 > +#define LINUX_EPOLLPRI 0x002 > +#define LINUX_EPOLLOUT 0x004 > +#define LINUX_EPOLLRDNORM 0x040 > +#define LINUX_EPOLLRDBAND 0x080 > +#define LINUX_EPOLLWRNORM 0x100 > +#define LINUX_EPOLLWRBAND 0x200 > +#define LINUX_EPOLLMSG 0x400 > +#define LINUX_EPOLLERR 0x008 > +#define LINUX_EPOLLHUP 0x010 > +#define LINUX_EPOLLRDHUP 0x2000 > +#define LINUX_EPOLLWAKEUP 1u<<29 > +#define LINUX_EPOLLONESHOT 1u<<30 > +#define LINUX_EPOLLET 1u<<31 > + > +#define LINUX_EPOLL_EVRD (LINUX_EPOLLIN|LINUX_EPOLLRDNORM \ > + |LINUX_EPOLLHUP|LINUX_EPOLLPRI) > +#define LINUX_EPOLL_EVWR (LINUX_EPOLLOUT|LINUX_EPOLLWRNORM) > +#define LINUX_EPOLL_EVSUP (LINUX_EPOLLET|LINUX_EPOLLONESHOT \ > + |LINUX_EPOLL_EVRD|LINUX_EPOLL_EVWR) > + > +#define LINUX_EPOLL_CTL_ADD 1 > +#define LINUX_EPOLL_CTL_DEL 2 > +#define LINUX_EPOLL_CTL_MOD 3 > + > +#endif /* !_LINUX_EVENT_H_ */ > > Modified: head/sys/compat/linux/linux_util.c > ============================================================================== > --- head/sys/compat/linux/linux_util.c Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/compat/linux/linux_util.c Sun May 24 16:41:39 2015 (r283441) > @@ -54,6 +54,7 @@ __FBSDID("$FreeBSD$"); > #include > > MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures"); > +MALLOC_DEFINE(M_EPOLL, "lepoll", "Linux events structures"); > MALLOC_DEFINE(M_FUTEX, "futex", "Linux futexes"); > MALLOC_DEFINE(M_FUTEX_WP, "futex wp", "Linux futex waiting proc"); > > > Modified: head/sys/compat/linux/linux_util.h > ============================================================================== > --- head/sys/compat/linux/linux_util.h Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/compat/linux/linux_util.h Sun May 24 16:41:39 2015 (r283441) > @@ -45,6 +45,7 @@ > #include > > MALLOC_DECLARE(M_LINUX); > +MALLOC_DECLARE(M_EPOLL); > MALLOC_DECLARE(M_FUTEX); > MALLOC_DECLARE(M_FUTEX_WP); > > > Modified: head/sys/conf/files.amd64 > ============================================================================== > --- head/sys/conf/files.amd64 Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/conf/files.amd64 Sun May 24 16:41:39 2015 (r283441) > @@ -509,6 +509,7 @@ compat/linux/linux_uid16.c optional comp > compat/linux/linux_util.c optional compat_linux32 > compat/linux/linux_vdso.c optional compat_linux32 > compat/linux/linux_common.c optional compat_linux32 > +compat/linux/linux_event.c optional compat_linux32 > dev/amr/amr_linux.c optional compat_linux32 amr > dev/mfi/mfi_linux.c optional compat_linux32 mfi > # > > Modified: head/sys/conf/files.i386 > ============================================================================== > --- head/sys/conf/files.i386 Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/conf/files.i386 Sun May 24 16:41:39 2015 (r283441) > @@ -81,6 +81,7 @@ hptrr_lib.o optional hptrr \ > cddl/contrib/opensolaris/common/atomic/i386/opensolaris_atomic.S optional zfs compile-with "${ZFS_S}" > compat/linprocfs/linprocfs.c optional linprocfs > compat/linsysfs/linsysfs.c optional linsysfs > +compat/linux/linux_event.c optional compat_linux > compat/linux/linux_emul.c optional compat_linux > compat/linux/linux_file.c optional compat_linux > compat/linux/linux_fork.c optional compat_linux > > Modified: head/sys/conf/files.pc98 > ============================================================================== > --- head/sys/conf/files.pc98 Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/conf/files.pc98 Sun May 24 16:41:39 2015 (r283441) > @@ -41,6 +41,7 @@ ukbdmap.h optional ukbd_dflt_keymap \ > cddl/contrib/opensolaris/common/atomic/i386/opensolaris_atomic.S optional zfs compile-with "${ZFS_S}" > compat/linprocfs/linprocfs.c optional linprocfs > compat/linsysfs/linsysfs.c optional linsysfs > +compat/linux/linux_event.c optional compat_linux > compat/linux/linux_emul.c optional compat_linux > compat/linux/linux_file.c optional compat_linux > compat/linux/linux_fork.c optional compat_linux > > Modified: head/sys/i386/linux/linux_dummy.c > ============================================================================== > --- head/sys/i386/linux/linux_dummy.c Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/i386/linux/linux_dummy.c Sun May 24 16:41:39 2015 (r283441) > @@ -70,9 +70,6 @@ DUMMY(setfsgid); > DUMMY(pivot_root); > DUMMY(mincore); > DUMMY(lookup_dcookie); > -DUMMY(epoll_create); > -DUMMY(epoll_ctl); > -DUMMY(epoll_wait); > DUMMY(remap_file_pages); > DUMMY(fstatfs64); > DUMMY(mbind); > @@ -116,7 +113,6 @@ DUMMY(timerfd_gettime); > /* linux 2.6.27: */ > DUMMY(signalfd4); > DUMMY(eventfd2); > -DUMMY(epoll_create1); > DUMMY(inotify_init1); > /* linux 2.6.30: */ > DUMMY(preadv); > > Modified: head/sys/i386/linux/syscalls.master > ============================================================================== > --- head/sys/i386/linux/syscalls.master Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/i386/linux/syscalls.master Sun May 24 16:41:39 2015 (r283441) > @@ -432,9 +432,11 @@ > 251 AUE_NULL UNIMPL > 252 AUE_EXIT STD { int linux_exit_group(int error_code); } > 253 AUE_NULL STD { int linux_lookup_dcookie(void); } > -254 AUE_NULL STD { int linux_epoll_create(void); } > -255 AUE_NULL STD { int linux_epoll_ctl(void); } > -256 AUE_NULL STD { int linux_epoll_wait(void); } > +254 AUE_NULL STD { int linux_epoll_create(l_int size); } > +255 AUE_NULL STD { int linux_epoll_ctl(l_int epfd, l_int op, l_int fd, \ > + struct epoll_event *event); } > +256 AUE_NULL STD { int linux_epoll_wait(l_int epfd, struct epoll_event *events, \ > + l_int maxevents, l_int timeout); } > 257 AUE_NULL STD { int linux_remap_file_pages(void); } > 258 AUE_NULL STD { int linux_set_tid_address(int *tidptr); } > 259 AUE_NULL STD { int linux_timer_create(clockid_t clock_id, \ > @@ -535,7 +537,8 @@ > 317 AUE_NULL STD { int linux_move_pages(void); } > ; linux 2.6.19: > 318 AUE_NULL STD { int linux_getcpu(void); } > -319 AUE_NULL STD { int linux_epoll_pwait(void); } > +319 AUE_NULL STD { int linux_epoll_pwait(l_int epfd, struct epoll_event *events, \ > + l_int maxevents, l_int timeout, l_osigset_t *mask); } > ; linux 2.6.22: > 320 AUE_NULL STD { int linux_utimensat(void); } > 321 AUE_NULL STD { int linux_signalfd(void); } > @@ -549,7 +552,7 @@ > ; linux 2.6.27: > 327 AUE_NULL STD { int linux_signalfd4(void); } > 328 AUE_NULL STD { int linux_eventfd2(void); } > -329 AUE_NULL STD { int linux_epoll_create1(void); } > +329 AUE_NULL STD { int linux_epoll_create1(l_int flags); } > 330 AUE_NULL STD { int linux_dup3(l_int oldfd, \ > l_int newfd, l_int flags); } > 331 AUE_NULL STD { int linux_pipe2(l_int *pipefds, l_int flags); } > > Modified: head/sys/modules/linux/Makefile > ============================================================================== > --- head/sys/modules/linux/Makefile Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/modules/linux/Makefile Sun May 24 16:41:39 2015 (r283441) > @@ -10,7 +10,7 @@ CFLAGS+=-DCOMPAT_FREEBSD32 -DCOMPAT_LINU > VDSO= linux${SFX}_vdso > > KMOD= linux > -SRCS= linux_fork.c linux${SFX}_dummy.c linux_file.c \ > +SRCS= linux_fork.c linux${SFX}_dummy.c linux_file.c linux_event.c \ > linux_futex.c linux_getcwd.c linux_ioctl.c linux_ipc.c \ > linux${SFX}_machdep.c linux_misc.c linux_signal.c \ > linux_socket.c linux_stats.c linux_sysctl.c linux${SFX}_sysent.c \ > > Modified: head/sys/modules/linux64/Makefile > ============================================================================== > --- head/sys/modules/linux64/Makefile Sun May 24 16:36:29 2015 (r283440) > +++ head/sys/modules/linux64/Makefile Sun May 24 16:41:39 2015 (r283441) > @@ -5,7 +5,7 @@ > VDSO= linux_vdso > > KMOD= linux64 > -SRCS= linux_fork.c linux_dummy.c linux_file.c \ > +SRCS= linux_fork.c linux_dummy.c linux_file.c linux_event.c \ > linux_futex.c linux_getcwd.c linux_ioctl.c linux_ipc.c \ > linux_machdep.c linux_misc.c linux_signal.c \ > linux_socket.c linux_stats.c linux_sysctl.c linux_sysent.c \ > >