/* $NetBSD: irix_usema.c,v 1.9 2003/01/22 12:58:24 rafal Exp $ */ /*- * Copyright (c) 2002 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Emmanuel Dreyfus. * * 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. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the NetBSD * Foundation, Inc. and its contributors. * 4. Neither the name of The NetBSD Foundation nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. 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 FOUNDATION 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 __KERNEL_RCSID(0, "$NetBSD: irix_usema.c,v 1.9 2003/01/22 12:58:24 rafal Exp $"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include const struct cdevsw irix_usema_cdevsw = { nullopen, nullclose, noread, nowrite, noioctl, nostop, notty, nopoll, nommap, nokqfilter, }; /* * semaphore list, and operations on the list */ static LIST_HEAD(irix_usema_reclist, irix_usema_rec) irix_usema_reclist; static struct lock irix_usema_reclist_lock; static struct irix_usema_rec *iur_lookup_by_vn __P((struct vnode *)); static struct irix_usema_rec *iur_lookup_by_sem __P((struct irix_semaphore *)); static struct irix_usema_rec *iur_insert __P((struct irix_semaphore *, struct vnode *, struct proc *)); static void iur_remove __P((struct irix_usema_rec *)); static struct irix_waiting_proc_rec *iur_proc_queue __P((struct irix_usema_rec *, struct proc *)); static void iur_proc_dequeue __P((struct irix_usema_rec *, struct irix_waiting_proc_rec *)); static void iur_proc_release __P((struct irix_usema_rec *, struct irix_waiting_proc_rec *)); static int iur_proc_isreleased __P((struct irix_usema_rec *, struct proc *)); static struct irix_waiting_proc_rec *iur_proc_getfirst __P((struct irix_usema_rec *)); /* * In order to define a custom vnode operation vector for the usemaclone * driver, we need to define a dummy filesystem, featuring just a null * init function and the vnode operation vector. This is defined by * irix_usema_dummy_vfsops, and registered to the kernel using vfs_attach * at driver attach time, in irix_usemaattach(). */ struct vfsops irix_usema_dummy_vfsops = { "usema_dummy", NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, irix_usema_dummy_vfs_init, NULL, NULL, NULL, NULL, NULL, irix_usema_vnodeopv_descs, }; void irix_usema_dummy_vfs_init(void) { return; } /* Do nothing */ const struct vnodeopv_desc * const irix_usema_vnodeopv_descs[] = { &irix_usema_opv_desc, NULL, }; const struct vnodeopv_desc irix_usema_opv_desc = { &irix_usema_vnodeop_p, irix_usema_vnodeop_entries }; int (**irix_usema_vnodeop_p) __P((void *)); /* * Vnode operations on the usemaclone device */ const struct vnodeopv_entry_desc irix_usema_vnodeop_entries[] = { { &vop_default_desc, vn_default_error }, { &vop_lookup_desc, genfs_nullop }, { &vop_open_desc, genfs_nullop }, { &vop_close_desc, irix_usema_close }, { &vop_access_desc, irix_usema_access }, { &vop_getattr_desc, irix_usema_getattr }, { &vop_setattr_desc, irix_usema_setattr }, { &vop_ioctl_desc, irix_usema_ioctl }, { &vop_fcntl_desc, irix_usema_fcntl }, { &vop_poll_desc, irix_usema_poll }, { &vop_abortop_desc, genfs_abortop }, { &vop_lease_desc, genfs_nullop }, { &vop_lock_desc, genfs_lock }, { &vop_unlock_desc, genfs_unlock }, { &vop_islocked_desc, genfs_islocked }, { &vop_advlock_desc, genfs_nullop }, { &vop_fsync_desc, genfs_nullop }, { &vop_reclaim_desc, genfs_nullop }, { &vop_revoke_desc, genfs_revoke }, { &vop_inactive_desc, irix_usema_inactive }, { NULL, NULL}, }; struct irix_usema_softc { struct device irix_usema_dev; }; /* * Initialize the usema driver: prepare the chained lists * and attach the dummy filesystem we need to use custom vnode operations. */ void irix_usemaattach(parent, self, aux) struct device *parent; struct device *self; void *aux; { int error; lockinit(&irix_usema_reclist_lock, PZERO|PCATCH, "usema", 0, 0); LIST_INIT(&irix_usema_reclist); if ((error = vfs_attach(&irix_usema_dummy_vfsops)) != 0) panic("irix_usemaattach: vfs_attach() failed"); return; } /* * vnode operations on the device */ int irix_usema_ioctl(v) void *v; { struct vop_ioctl_args /* { struct vnode *a_vp; u_long a_command; caddr_t a_data; int a_fflag; struct ucred *a_cred; struct proc *a_p; } */ *ap = v; u_long cmd = ap->a_command; caddr_t data = ap->a_data; struct vnode *vp = ap->a_vp; struct irix_usema_rec *iur; struct irix_waiting_proc_rec *iwpr; struct irix_ioctl_usrdata iiu; register_t *retval; int error; #ifdef DEBUG_IRIX printf("irix_usema_ioctl(): vp = %p, cmd = %lx, data = %p\n", vp, cmd, data); #endif /* * Some ioctl commands need to set the ioctl return value. In * irix_sys_ioctl(), we copy the return value address and the * data argument to the stackgap in a struct irix_ioctl_usrdata. * The address of this structure is passed as the data argument * to the vnode layer. We therefore need to read this structure * to get the real data argument and the retval address. */ if ((error = copyin(data, &iiu, sizeof(iiu))) != 0) return error; data = iiu.iiu_data; retval = iiu.iiu_retval; switch (cmd) { case IRIX_UIOCABLOCKQ: /* semaphore has been blocked */ if ((iur = iur_lookup_by_vn(vp)) == NULL) return EBADF; (void *)iur_proc_queue(iur, ap->a_p); break; case IRIX_UIOCAUNBLOCKQ: /* semaphore has been unblocked */ if ((iur = iur_lookup_by_vn(vp)) == NULL) return EBADF; if ((iwpr = iur_proc_getfirst(iur)) != NULL) { iur_proc_release(iur, iwpr); wakeup((void *)&selwait); } break; case IRIX_UIOCGETCOUNT: /* get semaphore value */ if ((iur = iur_lookup_by_vn(vp)) == NULL) return EBADF; *retval = -iur->iur_waiting_count; break; case IRIX_UIOCIDADDR: { /* register address of sem. owner field */ struct irix_usema_idaddr iui; struct irix_semaphore *isp; if ((error = copyin(data, &iui, sizeof(iui))) != 0) return error; /* * iui.iui_oidp points to the is_oid field of struct * irix_semaphore. We want the structre address itself. */ isp = NULL; isp = (struct irix_semaphore *)((u_long)(isp) - (u_long)(&isp->is_oid) + (u_long)iui.iui_oidp); if ((iur_insert(isp, vp, ap->a_p)) == NULL) return EFAULT; break; } default: printf("Warning: unimplemented IRIX usema ioctl command %ld\n", (cmd & 0xff)); break; } return 0; } int irix_usema_poll(v) void *v; { struct vop_poll_args /* { struct vnode *a_vp; int a_events; struct proc *a_p; } */ *ap = v; int events = ap->a_events; struct vnode *vp = ap->a_vp; struct irix_usema_rec *iur; int check = POLLIN|POLLRDNORM|POLLRDBAND|POLLPRI; #ifdef DEBUG_IRIX printf("irix_usema_poll() vn = %p, events = 0x%x\n", vp, events); #endif if ((events & check) == 0) return 0; if ((iur = iur_lookup_by_vn(vp)) == NULL) return 0; if (iur_proc_isreleased(iur, ap->a_p) == 0) return 0; return (events & check); } int irix_usema_close(v) void *v; { struct vop_close_args /* { struct vnode *a_vp; int a_fflag; struct ucred *a_cred; struct proc *a_p; } */ *ap = v; struct vnode *vp = ap->a_vp; struct vnode *rvp; struct irix_usema_rec *iur; int error; #ifdef DEBUG_IRIX printf("irix_usema_close() vn = %p\n", vp); #endif simple_lock(&vp->v_interlock); /* vp is a vnode duplicated from rvp. eventually also close rvp */ rvp = (struct vnode *)(vp->v_data); vrele(rvp); /* for vref() in irix_sys_open() */ vp->v_data = NULL; if (ap->a_fflag & FWRITE) rvp->v_writecount--; vn_lock(rvp, LK_EXCLUSIVE | LK_RETRY); error = VOP_CLOSE(rvp, ap->a_fflag, ap->a_cred, ap->a_p); vput(rvp); if ((iur = iur_lookup_by_vn(vp)) != NULL) iur_remove(iur); simple_unlock(&vp->v_interlock); return error; } /* * Try to apply setattr to the original vnode, not the duplicated one, * but still return 0 in case of failure (IRIX libc rely on this). */ int irix_usema_setattr(v) void *v; { struct vop_setattr_args /* { struct vnode *a_vp; struct vattr *a_vap; struct ucred *a_cred; struct proc *a_p; } */ *ap = v; struct vnode *vp = (struct vnode *)(ap->a_vp->v_data); int error; #ifdef DEBUG_IRIX printf("irix_usema_setattr()\n"); #endif error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred, ap->a_p); /* Silently ignore any error */ return 0; } int irix_usema_inactive(v) void *v; { struct vop_inactive_args /* { struct vnode *a_vp; struct proc *a_p; } */ *ap = v; VOP_UNLOCK(ap->a_vp, 0); vrecycle(ap->a_vp, NULL, ap->a_p); return 0; } /* * For fcntl, access and getattr vnode operations, we want to do the * operation on the original vnode, not the duplicated one. */ #define ___CONCAT(x,y) __CONCAT(x,y) #define __CONCAT3(x,y,z) ___CONCAT(__CONCAT(x,y),z) #define IRIX_USEMA_VNOP_WRAP(op) \ int \ __CONCAT(irix_usema_,op)(v) \ void *v; \ { \ struct __CONCAT3(vop_,op,_args) *ap = v; \ struct vnode *vp = (struct vnode *)(ap->a_vp->v_data); \ struct __CONCAT3(vop_,op,_args) a; \ \ (void)memcpy(&a, ap, sizeof(a)); \ a.a_vp = vp; \ \ return VCALL(vp,VOFFSET(__CONCAT(vop_,op)),&a); \ } IRIX_USEMA_VNOP_WRAP(access) IRIX_USEMA_VNOP_WRAP(getattr) IRIX_USEMA_VNOP_WRAP(fcntl) /* * The usync_ctnl system call is not part of the usema driver, * but it is closely related to it. */ int irix_sys_usync_cntl(l, v, retval) struct lwp *l; void *v; register_t *retval; { struct irix_sys_usync_cntl_args /* { syscallarg(int) cmd; syscallarg(void *) arg; } */ *uap = v; struct proc *p = l->l_proc; int error; struct irix_usync_arg iua; struct irix_usema_rec *iur; struct irix_waiting_proc_rec *iwpr; switch (SCARG(uap, cmd)) { case IRIX_USYNC_BLOCK: if ((error = copyin(SCARG(uap, arg), &iua, sizeof(iua))) != 0) return error; if ((iur = iur_insert(iua.iua_sem, NULL, p)) == NULL) return EFAULT; iwpr = iur_proc_queue(iur, p); (void)tsleep(iwpr, PZERO, "irix_usema", 0); break; case IRIX_USYNC_INTR_BLOCK: if ((error = copyin(SCARG(uap, arg), &iua, sizeof(iua))) != 0) return error; if ((iur = iur_insert(iua.iua_sem, NULL, p)) == NULL) return EFAULT; iwpr = iur_proc_queue(iur, p); (void)tsleep(iwpr, PZERO|PCATCH, "irix_usema", 0); break; case IRIX_USYNC_UNBLOCK_ALL: if ((error = copyin(SCARG(uap, arg), &iua, sizeof(iua))) != 0) return error; if ((iur = iur_lookup_by_sem(iua.iua_sem)) == 0) return EINVAL; (void)lockmgr(&iur->iur_lock, LK_SHARED, NULL); TAILQ_FOREACH(iwpr, &iur->iur_waiting_p, iwpr_list) { wakeup((void *)iwpr); iur_proc_dequeue(iur, iwpr); } iur_remove(iur); (void)lockmgr(&iur->iur_lock, LK_RELEASE, NULL); break; case IRIX_USYNC_UNBLOCK: if ((error = copyin(SCARG(uap, arg), &iua, sizeof(iua))) != 0) return error; if ((iur = iur_lookup_by_sem(iua.iua_sem)) == 0) return EINVAL; if ((iwpr = iur_proc_getfirst(iur)) != NULL) { wakeup((void *)iwpr); iur_proc_dequeue(iur, iwpr); } if ((iwpr = iur_proc_getfirst(iur)) == NULL) iur_remove(iur); break; case IRIX_USYNC_GET_STATE: if ((error = copyin(SCARG(uap, arg), &iua, sizeof(iua))) != 0) return error; if ((iur = iur_lookup_by_sem(iua.iua_sem)) == NULL) return 0; /* Not blocked, return 0 */ *retval = -iur->iur_waiting_count; break; default: printf("Warning: unimplemented IRIX usync_cntl command %d\n", SCARG(uap, cmd)); return EINVAL; } return 0; } /* Operations on irix_usema_reclist */ static struct irix_usema_rec * iur_lookup_by_vn(vp) struct vnode *vp; { struct irix_usema_rec *iur; (void)lockmgr(&irix_usema_reclist_lock, LK_SHARED, NULL); LIST_FOREACH(iur, &irix_usema_reclist, iur_list) if (iur->iur_vn == vp) break; (void)lockmgr(&irix_usema_reclist_lock, LK_RELEASE, NULL); return iur; } static struct irix_usema_rec * iur_lookup_by_sem(sem) struct irix_semaphore *sem; { struct irix_usema_rec *iur; struct irix_semaphore is; int error; if ((error = copyin(sem, &is, sizeof(is))) != 0) return NULL; (void)lockmgr(&irix_usema_reclist_lock, LK_SHARED, NULL); LIST_FOREACH(iur, &irix_usema_reclist, iur_list) if (iur->iur_sem == sem && iur->iur_shid == is.is_shid) break; (void)lockmgr(&irix_usema_reclist_lock, LK_RELEASE, NULL); return iur; } static struct irix_usema_rec * iur_insert(sem, vp, p) struct irix_semaphore *sem; struct vnode *vp; struct proc *p; { struct irix_usema_rec *iur; struct irix_semaphore is; int error; if ((iur = iur_lookup_by_sem(sem)) != NULL) return iur; if ((error = copyin(sem, &is, sizeof(is))) != 0) return NULL; iur = malloc(sizeof(struct irix_usema_rec), M_DEVBUF, M_WAITOK); iur->iur_sem = sem; iur->iur_vn = vp; iur->iur_shid = is.is_shid; iur->iur_p = p; iur->iur_waiting_count = 0; lockinit(&iur->iur_lock, PZERO|PCATCH, "_usema", 0, 0); TAILQ_INIT(&iur->iur_waiting_p); TAILQ_INIT(&iur->iur_released_p); (void)lockmgr(&irix_usema_reclist_lock, LK_EXCLUSIVE, NULL); LIST_INSERT_HEAD(&irix_usema_reclist, iur, iur_list); (void)lockmgr(&irix_usema_reclist_lock, LK_RELEASE, NULL); return iur; } static void iur_remove(iur) struct irix_usema_rec *iur; { struct irix_waiting_proc_rec *iwpr; (void)lockmgr(&iur->iur_lock, LK_EXCLUSIVE, NULL); waiting_restart: TAILQ_FOREACH(iwpr, &iur->iur_waiting_p, iwpr_list) { TAILQ_REMOVE(&iur->iur_waiting_p, iwpr, iwpr_list); free(iwpr, M_DEVBUF); /* iwpr is now invalid, restart */ goto waiting_restart; } released_restart: TAILQ_FOREACH(iwpr, &iur->iur_released_p, iwpr_list) { TAILQ_REMOVE(&iur->iur_released_p, iwpr, iwpr_list); free(iwpr, M_DEVBUF); /* iwpr is now invalid, restart */ goto released_restart; } (void)lockmgr(&irix_usema_reclist_lock, LK_EXCLUSIVE, NULL); LIST_REMOVE(iur, iur_list); (void)lockmgr(&irix_usema_reclist_lock, LK_RELEASE, NULL); (void)lockmgr(&iur->iur_lock, LK_RELEASE, NULL); free(iur, M_DEVBUF); return; } static struct irix_waiting_proc_rec * iur_proc_queue(iur, p) struct irix_usema_rec *iur; struct proc *p; { struct irix_waiting_proc_rec *iwpr; /* Do we have this iwpr on the released list? If we do, reuse it */ (void)lockmgr(&iur->iur_lock, LK_SHARED, NULL); TAILQ_FOREACH(iwpr, &iur->iur_released_p, iwpr_list) { if (iwpr->iwpr_p == p) { (void)lockmgr(&iur->iur_lock, LK_UPGRADE, NULL); TAILQ_REMOVE(&iur->iur_released_p, iwpr, iwpr_list); goto got_iwpr; } } /* Otherwise, create a new one */ iwpr = malloc(sizeof(struct irix_waiting_proc_rec), M_DEVBUF, M_WAITOK); iwpr->iwpr_p = p; (void)lockmgr(&iur->iur_lock, LK_UPGRADE, NULL); got_iwpr: TAILQ_INSERT_TAIL(&iur->iur_waiting_p, iwpr, iwpr_list); iur->iur_waiting_count++; (void)lockmgr(&iur->iur_lock, LK_RELEASE, NULL); return iwpr; } static void iur_proc_dequeue(iur, iwpr) struct irix_usema_rec *iur; struct irix_waiting_proc_rec *iwpr; { (void)lockmgr(&iur->iur_lock, LK_EXCLUSIVE, NULL); iur->iur_waiting_count--; TAILQ_REMOVE(&iur->iur_waiting_p, iwpr, iwpr_list); (void)lockmgr(&iur->iur_lock, LK_RELEASE, NULL); free(iwpr, M_DEVBUF); return; } static void iur_proc_release(iur, iwpr) struct irix_usema_rec *iur; struct irix_waiting_proc_rec *iwpr; { (void)lockmgr(&iur->iur_lock, LK_EXCLUSIVE, NULL); iur->iur_waiting_count--; TAILQ_REMOVE(&iur->iur_waiting_p, iwpr, iwpr_list); TAILQ_INSERT_TAIL(&iur->iur_released_p, iwpr, iwpr_list); (void)lockmgr(&iur->iur_lock, LK_RELEASE, NULL); return; } static int iur_proc_isreleased(iur, p) struct irix_usema_rec *iur; struct proc *p; { struct irix_waiting_proc_rec *iwpr; int res = 0; (void)lockmgr(&iur->iur_lock, LK_SHARED, NULL); TAILQ_FOREACH(iwpr, &iur->iur_released_p, iwpr_list) { if (iwpr->iwpr_p == p) { res = 1; break; } } (void)lockmgr(&iur->iur_lock, LK_RELEASE, NULL); return res; } static struct irix_waiting_proc_rec * iur_proc_getfirst(iur) struct irix_usema_rec *iur; { struct irix_waiting_proc_rec *iwpr; (void)lockmgr(&iur->iur_lock, LK_SHARED, NULL); iwpr = TAILQ_FIRST(&iur->iur_waiting_p); (void)lockmgr(&iur->iur_lock, LK_RELEASE, NULL); return iwpr; } /* * Cleanup irix_usema_usync_cntl() allocations * if new_p is NULL, free any structure allocated for process p * otherwise change ownership of structure allocated for process p to new_p */ void irix_usema_exit_cleanup(p, new_p) struct proc *p; struct proc *new_p; { struct irix_usema_rec *iur; #ifdef DEBUG_IRIX printf("irix_usema_exit_cleanup(): p = %p, new_p = %p\n", p, new_p); #endif remove_restart: (void)lockmgr(&irix_usema_reclist_lock, LK_SHARED, NULL); LIST_FOREACH(iur, &irix_usema_reclist, iur_list) { if (iur->iur_p != p) continue; if (new_p == NULL) { /* * Release the lock now since iur_remove() needs to * acquire an exclusive lock. */ (void)lockmgr(&irix_usema_reclist_lock, LK_RELEASE, NULL); iur_remove(iur); /* * iur is now invalid and we lost the lock, restart */ goto remove_restart; } else { (void)lockmgr(&irix_usema_reclist_lock, LK_UPGRADE, NULL); iur->iur_p = new_p; (void)lockmgr(&irix_usema_reclist_lock, LK_DOWNGRADE, NULL); } } (void)lockmgr(&irix_usema_reclist_lock, LK_RELEASE, NULL); return; } #ifdef DEBUG_IRIX /* * This dumps all in-kernel information about processes waiting for * semaphores and process that have been released by an operation * on a semaphore. * When called from ddb, curproc is NULL, and this panic lockmgr(). */ void irix_usema_debug(void) { struct irix_usema_rec *iur; struct irix_waiting_proc_rec *iwpr; if (curproc != NULL) (void)lockmgr(&irix_usema_reclist_lock, LK_SHARED, NULL); LIST_FOREACH(iur, &irix_usema_reclist, iur_list) { printf("iur %p\n", iur); printf(" iur->iur_vn = %p\n", iur->iur_vn); printf(" iur->iur_sem = %p\n", iur->iur_sem); printf(" iur->iur_shid = 0x%08x\n", iur->iur_shid); printf(" iur->iur_p = %p\n", iur->iur_p); printf(" iur->iur_waiting_count = %d\n", iur->iur_waiting_count); printf(" Waiting processes\n"); if (curproc != NULL) (void)lockmgr(&iur->iur_lock, LK_SHARED, NULL); TAILQ_FOREACH(iwpr, &iur->iur_waiting_p, iwpr_list) { printf(" iwpr %p: iwpr->iwpr_p = %p (pid %d)\n", iwpr, iwpr->iwpr_p, iwpr->iwpr_p->p_pid); } printf(" Released processes\n"); TAILQ_FOREACH(iwpr, &iur->iur_released_p, iwpr_list) { printf(" iwpr %p: iwpr->iwpr_p = %p (pid %d)\n", iwpr, iwpr->iwpr_p, iwpr->iwpr_p->p_pid); } if (curproc != NULL) (void)lockmgr(&iur->iur_lock, LK_RELEASE, NULL); } if (curproc != NULL) (void)lockmgr(&irix_usema_reclist_lock, LK_RELEASE, NULL); } #endif /* DEBUG_IRIX */