diff options
Diffstat (limited to 'usr/src/uts/common')
| -rw-r--r-- | usr/src/uts/common/brand/lx/autofs/lx_autofs.c | 3 | ||||
| -rw-r--r-- | usr/src/uts/common/brand/lx/procfs/lx_proc.h | 9 | ||||
| -rw-r--r-- | usr/src/uts/common/brand/lx/procfs/lx_prsubr.c | 259 | ||||
| -rw-r--r-- | usr/src/uts/common/brand/lx/procfs/lx_prvnops.c | 295 | ||||
| -rw-r--r-- | usr/src/uts/common/brand/lx/syscall/lx_mount.c | 167 | ||||
| -rw-r--r-- | usr/src/uts/common/brand/lx/syscall/lx_stat.c | 13 | ||||
| -rw-r--r-- | usr/src/uts/common/fs/lookup.c | 3 | ||||
| -rw-r--r-- | usr/src/uts/common/sys/file.h | 5 |
8 files changed, 528 insertions, 226 deletions
diff --git a/usr/src/uts/common/brand/lx/autofs/lx_autofs.c b/usr/src/uts/common/brand/lx/autofs/lx_autofs.c index 730deae80e..364215d026 100644 --- a/usr/src/uts/common/brand/lx/autofs/lx_autofs.c +++ b/usr/src/uts/common/brand/lx/autofs/lx_autofs.c @@ -2208,6 +2208,9 @@ lx_autofs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp, if (error != ENOENT) return (error); + if (flags & __FLXNOAUTO) + return (ENOENT); + if (data->lav_catatonic) return (ENOENT); diff --git a/usr/src/uts/common/brand/lx/procfs/lx_proc.h b/usr/src/uts/common/brand/lx/procfs/lx_proc.h index be95e7e471..f4d18fffc1 100644 --- a/usr/src/uts/common/brand/lx/procfs/lx_proc.h +++ b/usr/src/uts/common/brand/lx/procfs/lx_proc.h @@ -22,6 +22,7 @@ * Copyright 2009 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. * Copyright 2019 Joyent, Inc. + * Copyright 2021 OmniOS Community Edition (OmniOSce) Association. */ #ifndef _LX_PROC_H @@ -135,6 +136,8 @@ typedef enum lxpr_nodetype { LXPR_PID_TASK_IDDIR, /* /proc/<pid>/task/<tid> */ LXPR_PID_FDDIR, /* /proc/<pid>/fd */ LXPR_PID_FD_FD, /* /proc/<pid>/fd/nn */ + LXPR_PID_FDINFODIR, /* /proc/<pid>/fdinfo */ + LXPR_PID_FDINFO_FD, /* /proc/<pid>/fdinfo/nn */ LXPR_PID_UIDMAP, /* /proc/<pid>/uid_map */ LXPR_PID_TID_AUXV, /* /proc/<pid>/task/<tid>/auxv */ LXPR_PID_TID_CGROUP, /* /proc/<pid>/task/<tid>/cgroup */ @@ -158,6 +161,8 @@ typedef enum lxpr_nodetype { LXPR_PID_TID_STATUS, /* /proc/<pid>/task/<tid>/status */ LXPR_PID_TID_FDDIR, /* /proc/<pid>/task/<tid>/fd */ LXPR_PID_TID_FD_FD, /* /proc/<pid>/task/<tid>/fd/nn */ + LXPR_PID_TID_FDINFODIR, /* /proc/<pid>/task/<tid>/fdinfo */ + LXPR_PID_TID_FDINFO_FD, /* /proc/<pid>/task/<tid>/fdinfo/nn */ LXPR_PID_TID_UIDMAP, /* /proc/<pid>/task/<tid>/uid_map */ LXPR_CGROUPS, /* /proc/cgroups */ LXPR_CMDLINE, /* /proc/cmdline */ @@ -338,6 +343,8 @@ extern lxpr_node_t *lxpr_getnode(vnode_t *, lxpr_nodetype_t, proc_t *, int); extern void lxpr_freenode(lxpr_node_t *); extern vnode_t *lxpr_lookup_fdnode(vnode_t *, const char *); extern int lxpr_readlink_fdnode(lxpr_node_t *, char *, size_t); +extern vnode_t *lxpr_lookup_fdinfonode(vnode_t *, const char *); +extern int lxpr_open_flags_convert(offset_t, uint32_t); typedef struct lxpr_uiobuf { uio_t *uiop; @@ -370,6 +377,8 @@ extern proc_t *lxpr_lock_pid(lxpr_node_t *, pid_t, zombok_t, kthread_t **); extern void lxpr_unlock(proc_t *); extern netstack_t *lxpr_netstack(lxpr_node_t *); extern void lxpr_fixpid(zone_t *, proc_t *, pid_t *, pid_t *); +extern file_t *lxpr_getf(proc_t *, uint_t, short *); +extern void lxpr_releasef(proc_t *, uint_t); #ifdef __cplusplus } diff --git a/usr/src/uts/common/brand/lx/procfs/lx_prsubr.c b/usr/src/uts/common/brand/lx/procfs/lx_prsubr.c index aa778d1cc8..967d594913 100644 --- a/usr/src/uts/common/brand/lx/procfs/lx_prsubr.c +++ b/usr/src/uts/common/brand/lx/procfs/lx_prsubr.c @@ -22,6 +22,7 @@ * Copyright 2008 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. * Copyright 2019 Joyent, Inc. + * Copyright 2021 OmniOS Community Edition (OmniOSce) Association. */ /* @@ -35,7 +36,9 @@ #include <sys/vmsystm.h> #include <sys/prsystm.h> #include <sys/brand.h> +#include <sys/fcntl.h> #include <sys/lx_brand.h> +#include <sys/lx_fcntl.h> #include "lx_proc.h" @@ -352,6 +355,62 @@ lxpr_unlock(proc_t *p) mutex_exit(&p->p_lock); } +file_t * +lxpr_getf(proc_t *p, uint_t fd, short *flag) +{ + uf_entry_t *ufp; + uf_info_t *fip; + file_t *fp; + + ASSERT(MUTEX_HELD(&p->p_lock) && (p->p_proc_flag & P_PR_LOCK)); + + fip = P_FINFO(p); + + if (fd >= fip->fi_nfiles) + return (NULL); + + /* + * Drop p_lock, but keep the process P_PR_LOCK'd to prevent it from + * going away while we dereference into fi_list. + */ + mutex_exit(&p->p_lock); + mutex_enter(&fip->fi_lock); + UF_ENTER(ufp, fip, fd); + if ((fp = ufp->uf_file) != NULL && fp->f_count > 0) { + if (flag != NULL) + *flag = ufp->uf_flag; + ufp->uf_refcnt++; + } else { + fp = NULL; + } + UF_EXIT(ufp); + mutex_exit(&fip->fi_lock); + mutex_enter(&p->p_lock); + + return (fp); +} + +void +lxpr_releasef(proc_t *p, uint_t fd) +{ + uf_entry_t *ufp; + uf_info_t *fip; + + ASSERT(MUTEX_HELD(&p->p_lock) && (p->p_proc_flag & P_PR_LOCK)); + + fip = P_FINFO(p); + + mutex_exit(&p->p_lock); + mutex_enter(&fip->fi_lock); + UF_ENTER(ufp, fip, fd); + ASSERT3U(ufp->uf_refcnt, >, 0); + ufp->uf_refcnt--; + UF_EXIT(ufp); + mutex_exit(&fip->fi_lock); + mutex_enter(&p->p_lock); +} + + void lxpr_initnodecache() { @@ -366,7 +425,6 @@ lxpr_fininodecache() kmem_cache_destroy(lxpr_node_cache); } -/* ARGSUSED */ static int lxpr_node_constructor(void *buf, void *un, int kmflags) { @@ -383,7 +441,6 @@ lxpr_node_constructor(void *buf, void *un, int kmflags) return (0); } -/* ARGSUSED */ static void lxpr_node_destructor(void *buf, void *un) { @@ -586,8 +643,17 @@ lxpr_getnode(vnode_t *dp, lxpr_nodetype_t type, proc_t *p, int desc) vp->v_type = VLNK; break; + case LXPR_PID_FDINFO_FD: + case LXPR_PID_TID_FDINFO_FD: + ASSERT(p != NULL); + lxpnp->lxpr_mode = 0400; /* read by owner only */ + vp->v_type = VREG; + break; + case LXPR_PID_FDDIR: case LXPR_PID_TID_FDDIR: + case LXPR_PID_FDINFODIR: + case LXPR_PID_TID_FDINFODIR: ASSERT(p != NULL); vp->v_type = VDIR; lxpnp->lxpr_mode = 0500; /* read-search by owner only */ @@ -661,6 +727,27 @@ lxpr_freenode(lxpr_node_t *lxpnp) kmem_cache_free(lxpr_node_cache, lxpnp); } +static int +lxpr_parse_fdnode_num(const char *name) +{ + char *endptr = NULL; + long num; + int fd; + + if (ddi_strtol(name, &endptr, 10, &num) != 0) { + return (-1); + } else if (name[0] < '0' || name[0] > '9' || *endptr != '\0') { + /* + * ddi_strtol allows leading spaces and trailing garbage + * We do not tolerate such foolishness. + */ + return (-1); + } else if ((fd = (int)num) < 0) { + return (-1); + } + return (fd); +} + /* * Attempt to locate vnode for /proc/<pid>/fd/<#>. */ @@ -669,34 +756,20 @@ lxpr_lookup_fdnode(vnode_t *dvp, const char *name) { lxpr_node_t *lxdp = VTOLXP(dvp); lxpr_node_t *lxfp; - char *endptr = NULL; - long num; int fd; proc_t *p; vnode_t *vp = NULL; file_t *fp; - uf_entry_t *ufp; - uf_info_t *fip; ASSERT(lxdp->lxpr_type == LXPR_PID_FDDIR || lxdp->lxpr_type == LXPR_PID_TID_FDDIR); - if (ddi_strtol(name, &endptr, 10, &num) != 0) { - return (NULL); - } else if (name[0] < '0' || name[0] > '9' || *endptr != '\0') { - /* - * ddi_strtol allows leading spaces and trailing garbage - * We do not tolerate such foolishness. - */ + if ((fd = lxpr_parse_fdnode_num(name)) == -1) return (NULL); - } else if ((fd = (int)num) < 0) { - return (NULL); - } /* Lock the owner process */ - if ((p = lxpr_lock(lxdp, NO_ZOMB)) == NULL) { + if ((p = lxpr_lock(lxdp, NO_ZOMB)) == NULL) return (NULL); - } /* Not applicable to processes which are system-owned. */ if (p->p_as == &kas) { @@ -706,50 +779,38 @@ lxpr_lookup_fdnode(vnode_t *dvp, const char *name) lxfp = lxpr_getnode(dvp, LXPR_PID_FD_FD, p, fd); - /* - * Drop p_lock, but keep the process P_PR_LOCK'd to prevent it from - * going away while we dereference into fi_list. - */ - fip = P_FINFO(p); - mutex_exit(&p->p_lock); - mutex_enter(&fip->fi_lock); - if (fd < fip->fi_nfiles) { - UF_ENTER(ufp, fip, fd); - if ((fp = ufp->uf_file) != NULL) { - vp = fp->f_vnode; - VN_HOLD(vp); - } - UF_EXIT(ufp); + if ((fp = lxpr_getf(p, fd, NULL)) != NULL) { + vp = fp->f_vnode; + VN_HOLD(vp); + lxpr_releasef(p, fd); } - mutex_exit(&fip->fi_lock); if (vp == NULL) { - mutex_enter(&p->p_lock); lxpr_unlock(p); lxpr_freenode(lxfp); return (NULL); - } else { - /* - * Fill in the lxpr_node so future references will be able to - * find the underlying vnode. The vnode is held on the realvp. - */ - lxfp->lxpr_realvp = vp; - - /* - * For certain entries (sockets, pipes, etc), Linux expects a - * bogus-named symlink. If that's the case, report the type as - * VNON to bypass link-following elsewhere in the vfs system. - * - * See lxpr_readlink for more details. - */ - if (lxpr_readlink_fdnode(lxfp, NULL, 0) == 0) - LXPTOV(lxfp)->v_type = VNON; } - mutex_enter(&p->p_lock); + /* + * Fill in the lxpr_node so future references will be able to + * find the underlying vnode. The vnode is held on the realvp. + */ + lxfp->lxpr_realvp = vp; + vp = LXPTOV(lxfp); + + /* + * For certain entries (sockets, pipes, etc), Linux expects a + * bogus-named symlink. If that's the case, report the type as + * VNON to bypass link-following elsewhere in the vfs system. + * + * See lxpr_readlink for more details. + */ + if (lxpr_readlink_fdnode(lxfp, NULL, 0) == 0) + vp->v_type = VNON; + lxpr_unlock(p); - ASSERT(LXPTOV(lxfp) != NULL); - return (LXPTOV(lxfp)); + ASSERT(vp != NULL); + return (vp); } /* @@ -784,7 +845,95 @@ lxpr_readlink_fdnode(lxpr_node_t *lxpnp, char *bp, size_t len) } /* - * Translate a Linux core_pattern path to a native Illumos one, by replacing + * Attempt to locate vnode for /proc/<pid>/fdinfo/<#>. + */ +vnode_t * +lxpr_lookup_fdinfonode(vnode_t *dvp, const char *name) +{ + lxpr_node_t *lxdp = VTOLXP(dvp); + lxpr_node_t *lxfp; + proc_t *p; + int fd; + + ASSERT(lxdp->lxpr_type == LXPR_PID_FDINFODIR); + + if ((fd = lxpr_parse_fdnode_num(name)) == -1) + return (NULL); + + /* Lock the owner process */ + if ((p = lxpr_lock(lxdp, NO_ZOMB)) == NULL) + return (NULL); + + /* Not applicable to processes which are system-owned. */ + if (p->p_as == &kas) { + lxpr_unlock(p); + return (NULL); + } + + lxfp = lxpr_getnode(dvp, LXPR_PID_FDINFO_FD, p, fd); + + lxpr_unlock(p); + ASSERT(LXPTOV(lxfp) != NULL); + return (LXPTOV(lxfp)); +} + +/* + * Translate native file flags to Linux open flags. + */ +int +lxpr_open_flags_convert(offset_t uf_flag, uint32_t f_flag) +{ + int flags = 0; + + switch (f_flag & (FREAD | FWRITE)) { + case FREAD: + flags = LX_O_RDONLY; + break; + case FWRITE: + flags = LX_O_WRONLY; + break; + case FREAD | FWRITE: + flags = LX_O_RDWR; + break; + } + + if (f_flag & FNDELAY) + flags |= LX_O_NDELAY; + if (f_flag & FAPPEND) + flags |= LX_O_APPEND; + if (f_flag & FSYNC) + flags |= LX_O_SYNC; + if (f_flag & FNONBLOCK) + flags |= LX_O_NONBLOCK; + + if (f_flag & FCREAT) + flags |= LX_O_CREAT; + if (f_flag & FTRUNC) + flags |= LX_O_TRUNC; + if (f_flag & FEXCL) + flags |= LX_O_EXCL; + if (f_flag & FASYNC) + flags |= LX_O_ASYNC; + if (f_flag & FOFFMAX) + flags |= LX_O_LARGEFILE; + if (f_flag & FNOCTTY) + flags |= LX_O_NOCTTY; + if (f_flag & FNOFOLLOW) + flags |= LX_O_NOFOLLOW; + + if (f_flag & FDIRECT) + flags |= LX_O_DIRECT; + if (f_flag & __FLXPATH) + flags |= LX_O_PATH; + + if (uf_flag & FD_CLOEXEC) + flags |= LX_O_CLOEXEC; + + return (flags); +} + +/* + * Translate a Linux core_pattern path to a native illumos one, by replacing * the appropriate % escape sequences. * * Any % escape sequences that are not recognised are double-escaped so that @@ -854,7 +1003,7 @@ lxpr_core_path_l2s(const char *inp, char *outp, size_t outsz) } /* - * Translate an Illumos core pattern path back to Linux format. + * Translate an illumos core pattern path back to Linux format. */ int lxpr_core_path_s2l(const char *inp, char *outp, size_t outsz) diff --git a/usr/src/uts/common/brand/lx/procfs/lx_prvnops.c b/usr/src/uts/common/brand/lx/procfs/lx_prvnops.c index d573825652..6d4d2357eb 100644 --- a/usr/src/uts/common/brand/lx/procfs/lx_prvnops.c +++ b/usr/src/uts/common/brand/lx/procfs/lx_prvnops.c @@ -22,18 +22,11 @@ * Copyright 2010 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. * Copyright 2019 Joyent, Inc. - * Copyright 2020 OmniOS Community Edition (OmniOSce) Association. + * Copyright 2021 OmniOS Community Edition (OmniOSce) Association. */ /* * lx_proc -- a Linux-compatible /proc for the LX brand - * - * We have -- confusingly -- two implementations of Linux /proc. One is to - * support native (but Linux-borne) programs that wish to view the native - * system through the Linux /proc model; the other -- this one -- is to - * support Linux binaries via the LX brand. These two implementations differ - * greatly in their aspirations (and their willingness to bend the truth - * of the system to accommodate those aspirations); they should not be unified. */ #include <sys/cpupart.h> @@ -136,6 +129,7 @@ static vnode_t *lxpr_lookup_procdir(vnode_t *, char *); static vnode_t *lxpr_lookup_piddir(vnode_t *, char *); static vnode_t *lxpr_lookup_not_a_dir(vnode_t *, char *); static vnode_t *lxpr_lookup_fddir(vnode_t *, char *); +static vnode_t *lxpr_lookup_fdinfodir(vnode_t *, char *); static vnode_t *lxpr_lookup_netdir(vnode_t *, char *); static vnode_t *lxpr_lookup_sysdir(vnode_t *, char *); static vnode_t *lxpr_lookup_sys_fsdir(vnode_t *, char *); @@ -153,6 +147,7 @@ static int lxpr_readdir_procdir(lxpr_node_t *, uio_t *, int *); static int lxpr_readdir_piddir(lxpr_node_t *, uio_t *, int *); static int lxpr_readdir_not_a_dir(lxpr_node_t *, uio_t *, int *); static int lxpr_readdir_fddir(lxpr_node_t *, uio_t *, int *); +static int lxpr_readdir_fdinfodir(lxpr_node_t *, uio_t *, int *); static int lxpr_readdir_netdir(lxpr_node_t *, uio_t *, int *); static int lxpr_readdir_sysdir(lxpr_node_t *, uio_t *, int *); static int lxpr_readdir_sys_fsdir(lxpr_node_t *, uio_t *, int *); @@ -175,6 +170,7 @@ static void lxpr_read_devices(lxpr_node_t *, lxpr_uiobuf_t *); static void lxpr_read_diskstats(lxpr_node_t *, lxpr_uiobuf_t *); static void lxpr_read_isdir(lxpr_node_t *, lxpr_uiobuf_t *); static void lxpr_read_fd(lxpr_node_t *, lxpr_uiobuf_t *); +static void lxpr_read_fdinfo(lxpr_node_t *, lxpr_uiobuf_t *); static void lxpr_read_filesystems(lxpr_node_t *, lxpr_uiobuf_t *); static void lxpr_read_kmsg(lxpr_node_t *, lxpr_uiobuf_t *, ldi_handle_t); static void lxpr_read_loadavg(lxpr_node_t *, lxpr_uiobuf_t *); @@ -425,6 +421,7 @@ static lxpr_dirent_t piddir[] = { { LXPR_PID_STATUS, "status" }, { LXPR_PID_TASKDIR, "task" }, { LXPR_PID_FDDIR, "fd" }, + { LXPR_PID_FDINFODIR, "fdinfo" }, { LXPR_PID_UIDMAP, "uid_map" } }; @@ -455,6 +452,7 @@ static lxpr_dirent_t tiddir[] = { { LXPR_PID_STATM, "statm" }, { LXPR_PID_TID_STATUS, "status" }, { LXPR_PID_FDDIR, "fd" }, + { LXPR_PID_FDINFODIR, "fdinfo" }, { LXPR_PID_UIDMAP, "uid_map" } }; @@ -793,7 +791,6 @@ lxpr_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct) /* * lxpr_close(): Vnode operation for VOP_CLOSE() */ -/* ARGSUSED */ static int lxpr_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr, caller_context_t *ct) @@ -844,6 +841,8 @@ static void (*lxpr_read_function[])() = { lxpr_read_isdir, /* /proc/<pid>/task/nn */ lxpr_read_isdir, /* /proc/<pid>/fd */ lxpr_read_fd, /* /proc/<pid>/fd/nn */ + lxpr_read_isdir, /* /proc/<pid>/fdinfo */ + lxpr_read_fdinfo, /* /proc/<pid>/fdinfo/nn */ lxpr_read_pid_id_map, /* /proc/<pid>/uid_map */ lxpr_read_pid_auxv, /* /proc/<pid>/task/<tid>/auxv */ lxpr_read_pid_cgroup, /* /proc/<pid>/task/<tid>/cgroup */ @@ -867,6 +866,8 @@ static void (*lxpr_read_function[])() = { lxpr_read_pid_tid_status, /* /proc/<pid>/task/<tid>/status */ lxpr_read_isdir, /* /proc/<pid>/task/<tid>/fd */ lxpr_read_fd, /* /proc/<pid>/task/<tid>/fd/nn */ + lxpr_read_isdir, /* /proc/<pid>/task/<tid>/fdinfo */ + lxpr_read_fdinfo, /* /proc/<pid>/task/<tid>/fdinfo/nn */ lxpr_read_pid_id_map, /* /proc/<pid>/task/<tid>/uid_map */ lxpr_read_cgroups, /* /proc/cgroups */ lxpr_read_cmdline, /* /proc/cmdline */ @@ -1014,6 +1015,8 @@ static vnode_t *(*lxpr_lookup_function[])() = { lxpr_lookup_task_tid_dir, /* /proc/<pid>/task/nn */ lxpr_lookup_fddir, /* /proc/<pid>/fd */ lxpr_lookup_not_a_dir, /* /proc/<pid>/fd/nn */ + lxpr_lookup_fdinfodir, /* /proc/<pid>/fdinfo */ + lxpr_lookup_not_a_dir, /* /proc/<pid>/fdinfo/nn */ lxpr_lookup_not_a_dir, /* /proc/<pid>/uid_map */ lxpr_lookup_not_a_dir, /* /proc/<pid>/task/<tid>/auxv */ lxpr_lookup_not_a_dir, /* /proc/<pid>/task/<tid>/cgroup */ @@ -1037,6 +1040,8 @@ static vnode_t *(*lxpr_lookup_function[])() = { lxpr_lookup_not_a_dir, /* /proc/<pid>/task/<tid>/status */ lxpr_lookup_fddir, /* /proc/<pid>/task/<tid>/fd */ lxpr_lookup_not_a_dir, /* /proc/<pid>/task/<tid>/fd/nn */ + lxpr_lookup_fdinfodir, /* /proc/<pid>/task/<tid>/fdinfo */ + lxpr_lookup_not_a_dir, /* /proc/<pid>/task/<tid>/fdinfo/nn */ lxpr_lookup_not_a_dir, /* /proc/<pid>/task/<tid>/uid_map */ lxpr_lookup_not_a_dir, /* /proc/cgroups */ lxpr_lookup_not_a_dir, /* /proc/cmdline */ @@ -1184,6 +1189,8 @@ static int (*lxpr_readdir_function[])() = { lxpr_readdir_task_tid_dir, /* /proc/<pid>/task/nn */ lxpr_readdir_fddir, /* /proc/<pid>/fd */ lxpr_readdir_not_a_dir, /* /proc/<pid>/fd/nn */ + lxpr_readdir_fdinfodir, /* /proc/<pid>/fdinfo */ + lxpr_readdir_not_a_dir, /* /proc/<pid>/fdinfo/nn */ lxpr_readdir_not_a_dir, /* /proc/<pid>/uid_map */ lxpr_readdir_not_a_dir, /* /proc/<pid>/task/<tid>/auxv */ lxpr_readdir_not_a_dir, /* /proc/<pid>/task/<tid>/cgroup */ @@ -1207,6 +1214,8 @@ static int (*lxpr_readdir_function[])() = { lxpr_readdir_not_a_dir, /* /proc/<pid>/task/<tid>/status */ lxpr_readdir_fddir, /* /proc/<pid>/task/<tid>/fd */ lxpr_readdir_not_a_dir, /* /proc/<pid>/task/<tid>/fd/nn */ + lxpr_readdir_fdinfodir, /* /proc/<pid>/task/<tid>/fdinfo */ + lxpr_readdir_not_a_dir, /* /proc/<pid>/task/<tid>/fdinfo/nn */ lxpr_readdir_not_a_dir, /* /proc/<pid>/task/<tid>/uid_map */ lxpr_readdir_not_a_dir, /* /proc/cgroups */ lxpr_readdir_not_a_dir, /* /proc/cmdline */ @@ -1331,7 +1340,6 @@ CTASSERT(ARRAY_SIZE(lxpr_readdir_function) == LXPR_NFILES); * (at least in general, and certainly the difference is unlikely to be enough * to justify have different routines for 32 and 64 bit reads */ -/* ARGSUSED */ static int lxpr_read(vnode_t *vp, uio_t *uiop, int ioflag, cred_t *cr, caller_context_t *ct) @@ -1393,21 +1401,18 @@ lxpr_read(vnode_t *vp, uio_t *uiop, int ioflag, cred_t *cr, * - empty file * - wait to be able to read a file that will never have anything to read */ -/* ARGSUSED */ static void lxpr_read_isdir(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { lxpr_uiobuf_seterr(uiobuf, EISDIR); } -/* ARGSUSED */ static void lxpr_read_invalid(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { lxpr_uiobuf_seterr(uiobuf, EINVAL); } -/* ARGSUSED */ static void lxpr_read_empty(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -1589,7 +1594,6 @@ lxpr_read_pid_tid_comm(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%s\n", buf); } -/* ARGSUSED */ static int lxpr_write_pid_tid_comm(lxpr_node_t *lxpnp, struct uio *uio, struct cred *cr, caller_context_t *ct) @@ -2238,6 +2242,58 @@ lxpr_enumerate_mounts(zone_t *zone) return (result); } +static uint_t +lxpr_get_mountid(zone_t *zone, vfs_t *match_vfsp) +{ + lx_zone_data_t *lxzd = ztolxzd(zone); + vfs_t *vfsp, *vfslist; + uint_t mount_id; + + if (match_vfsp == NULL) + return (0); + + /* Mount IDs start at 15 for the root, see lxpr_enumerate_mounts() */ + mount_id = 15; + + ASSERT(zone != global_zone); + ASSERT(lxzd != NULL); + ASSERT(lxzd->lxzd_vdisks != NULL); + + if (zone->zone_rootvp->v_vfsp == match_vfsp) + return (mount_id); + + vfs_list_read_lock(); + + vfsp = vfslist = zone->zone_vfslist; + + do { + if (vfsp == zone->zone_rootvp->v_vfsp) + continue; + + if (vfsp == NULL) + break; + + /* Skip mounts we shouldn't show */ + if ((vfsp->vfs_flag & VFS_NOMNTTAB) != 0) { + vfsp = vfsp->vfs_zone_next; + continue; + } + + mount_id++; + + if (vfsp == match_vfsp) { + vfs_list_unlock(); + return (mount_id); + } + + vfsp = vfsp->vfs_zone_next; + } while (vfsp != vfslist); + + vfs_list_unlock(); + + return (0); +} + /* * lxpr_read_pid_mountinfo(): information about process mount points. */ @@ -2860,7 +2916,6 @@ lxpr_read_pid_tid_stat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) cpu /* 39 */); } -/* ARGSUSED */ static void lxpr_read_net_arp(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -3042,7 +3097,6 @@ lxpr_read_net_dev(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) kmem_free(ksr, sidx); } -/* ARGSUSED */ static void lxpr_read_net_dev_mcast(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -3061,7 +3115,6 @@ lxpr_inet6_out(const in6_addr_t *addr, char buf[33]) buf[32] = '\0'; } -/* ARGSUSED */ static void lxpr_read_net_if_inet6(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -3102,19 +3155,16 @@ lxpr_read_net_if_inet6(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) netstack_rele(ns); } -/* ARGSUSED */ static void lxpr_read_net_igmp(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { } -/* ARGSUSED */ static void lxpr_read_net_ip_mr_cache(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { } -/* ARGSUSED */ static void lxpr_read_net_ip_mr_vif(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -3159,7 +3209,6 @@ lxpr_format_route_ipv6(ire_t *ire, lxpr_uiobuf_t *uiobuf) name); } -/* ARGSUSED */ static void lxpr_read_net_ipv6_route(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -3180,19 +3229,16 @@ lxpr_read_net_ipv6_route(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) netstack_rele(ns); } -/* ARGSUSED */ static void lxpr_read_net_mcfilter(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { } -/* ARGSUSED */ static void lxpr_read_net_netstat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { } -/* ARGSUSED */ static void lxpr_read_net_raw(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -3222,7 +3268,7 @@ lxpr_format_route_ipv4(ire_t *ire, lxpr_uiobuf_t *uiobuf) /* * Search for a suitable IRE for naming purposes. * On Linux, the default route is typically associated with the - * interface used to access gateway. The default IRE on Illumos + * interface used to access gateway. The default IRE on illumos * typically lacks an ill reference but its parent might have one. */ nire = ire; @@ -3267,7 +3313,6 @@ lxpr_format_route_ipv4(ire_t *ire, lxpr_uiobuf_t *uiobuf) ire->ire_metrics.iulp_rtt); } -/* ARGSUSED */ static void lxpr_read_net_route(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -3291,19 +3336,16 @@ lxpr_read_net_route(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) netstack_rele(ns); } -/* ARGSUSED */ static void lxpr_read_net_rpc(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { } -/* ARGSUSED */ static void lxpr_read_net_rt_cache(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { } -/* ARGSUSED */ static void lxpr_read_net_sockstat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -3453,7 +3495,6 @@ lxpr_read_net_snmp(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) kmem_free(ksr, sidx); } -/* ARGSUSED */ static void lxpr_read_net_stat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -3861,7 +3902,6 @@ lxpr_read_net_unix(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) #define LX_KMSG_PRI "<0>" -/* ARGSUSED */ static void lxpr_read_kmsg(lxpr_node_t *lxpnp, struct lxpr_uiobuf *uiobuf, ldi_handle_t lh) { @@ -4052,7 +4092,6 @@ lxpr_read_meminfo(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * Note: we currently also use this for /proc/{pid}/mounts since we don't * yet support mount namespaces. */ -/* ARGSUSED */ static void lxpr_read_mounts(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4130,7 +4169,6 @@ nextp: * /proc/diskstats, and also because "fdisk -l" and a few other things look * here to find all disks on the system. */ -/* ARGSUSED */ static void lxpr_read_partitions(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4163,7 +4201,6 @@ lxpr_read_partitions(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * the expected Linux major numbers. See lx devfs where some of the major * numbers have no defined constants. */ -/* ARGSUSED */ static void lxpr_read_devices(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4186,7 +4223,6 @@ lxpr_read_devices(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * See the block comment above the per-device output-generating line for the * details of the format. */ -/* ARGSUSED */ static void lxpr_read_diskstats(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4315,7 +4351,6 @@ lxpr_read_diskstats(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) /* * lxpr_read_version(): read the contents of the "version" file. */ -/* ARGSUSED */ static void lxpr_read_version(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4350,7 +4385,6 @@ lxpr_read_version(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) version); } -/* ARGSUSED */ static void lxpr_read_vmstat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4405,7 +4439,6 @@ lxpr_read_vmstat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * lxpr_read_stat(): read the contents of the "stat" file. * */ -/* ARGSUSED */ static void lxpr_read_stat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4603,7 +4636,6 @@ lxpr_read_stat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * It is important to use formatting identical to the Linux implementation * so that consumers do not break. See swap_show() in mm/swapfile.c. */ -/* ARGSUSED */ static void lxpr_read_swaps(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4636,7 +4668,6 @@ lxpr_read_swaps(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) } } -/* ARGSUSED */ static void lxpr_read_sys_fs_aiomax(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4644,7 +4675,6 @@ lxpr_read_sys_fs_aiomax(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%llu\n", LX_AIO_MAX_NR); } -/* ARGSUSED */ static void lxpr_read_sys_fs_aionr(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4670,7 +4700,6 @@ lxpr_read_sys_fs_aionr(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * (zone's proc limit) * (process.max-file-descriptor rctl privileged limit). * The privileged rctl limit is the same as rlim_fd_max. */ -/* ARGSUSED */ static void lxpr_read_sys_fs_filemax(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4694,7 +4723,6 @@ lxpr_read_sys_fs_filemax(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * number of files in use within a zone, so we approximate that value by * looking at the current "fi_nfiles" value for each process in the zone. */ -/* ARGSUSED */ static void lxpr_read_sys_fs_filenr(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4742,7 +4770,6 @@ extern int inotify_maxevents; extern int inotify_maxinstances; extern int inotify_maxwatches; -/* ARGSUSED */ static void lxpr_read_sys_fs_inotify_max_queued_events(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) @@ -4751,7 +4778,6 @@ lxpr_read_sys_fs_inotify_max_queued_events(lxpr_node_t *lxpnp, lxpr_uiobuf_printf(uiobuf, "%d\n", inotify_maxevents); } -/* ARGSUSED */ static void lxpr_read_sys_fs_inotify_max_user_instances(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) @@ -4760,7 +4786,6 @@ lxpr_read_sys_fs_inotify_max_user_instances(lxpr_node_t *lxpnp, lxpr_uiobuf_printf(uiobuf, "%d\n", inotify_maxinstances); } -/* ARGSUSED */ static void lxpr_read_sys_fs_inotify_max_user_watches(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) @@ -4769,7 +4794,6 @@ lxpr_read_sys_fs_inotify_max_user_watches(lxpr_node_t *lxpnp, lxpr_uiobuf_printf(uiobuf, "%d\n", inotify_maxwatches); } -/* ARGSUSED */ static void lxpr_read_sys_fs_pipe_max(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4786,7 +4810,6 @@ lxpr_read_sys_fs_pipe_max(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%u\n", pipe_max); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_caplcap(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4794,7 +4817,6 @@ lxpr_read_sys_kernel_caplcap(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%d\n", LX_CAP_MAX_VALID); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_corepatt(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4837,7 +4859,6 @@ lxpr_read_sys_kernel_corepatt(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%s\n", tr); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_hostname(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4845,7 +4866,6 @@ lxpr_read_sys_kernel_hostname(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%s\n", uts_nodename()); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_msgmax(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4862,7 +4882,6 @@ lxpr_read_sys_kernel_msgmax(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%u\n", val); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_msgmnb(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4878,7 +4897,6 @@ lxpr_read_sys_kernel_msgmnb(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%u\n", (uint_t)val); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_msgmni(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4894,7 +4912,6 @@ lxpr_read_sys_kernel_msgmni(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%u\n", (uint_t)val); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_ngroups_max(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4902,7 +4919,6 @@ lxpr_read_sys_kernel_ngroups_max(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%d\n", ngroups_max); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_osrel(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4920,7 +4936,6 @@ lxpr_read_sys_kernel_osrel(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%s\n", version); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_pid_max(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -4983,7 +4998,6 @@ lxpr_read_sys_kernel_rand_bootid(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) /* * The amount of entropy available (in bits). */ -/* ARGSUSED */ static void lxpr_read_sys_kernel_rand_entavl(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5011,7 +5025,6 @@ lxpr_read_sys_kernel_rand_uuid(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%s\n", uuid); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_sem(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5041,7 +5054,6 @@ lxpr_read_sys_kernel_sem(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) vmsl, vmns, vopm, vmni); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_shmall(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5058,7 +5070,6 @@ lxpr_read_sys_kernel_shmall(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%u\n", (uint_t)btop(val)); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_shmmax(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5077,7 +5088,6 @@ lxpr_read_sys_kernel_shmmax(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%u\n", (uint_t)val); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_shmmni(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5096,7 +5106,6 @@ lxpr_read_sys_kernel_shmmni(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%u\n", (uint_t)val); } -/* ARGSUSED */ static void lxpr_read_sys_kernel_threads_max(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5104,7 +5113,6 @@ lxpr_read_sys_kernel_threads_max(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%d\n", LXPTOZ(lxpnp)->zone_nlwps_ctl); } -/* ARGSUSED */ static void lxpr_read_sys_net_core_somaxc(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5135,7 +5143,6 @@ lxpr_read_sys_net_core_somaxc(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * * Note that the Linux setting is the inverse of the illumos value. */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_icmp_eib(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5165,7 +5172,6 @@ lxpr_read_sys_net_ipv4_icmp_eib(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * in lx at this time, thus we do not support Linux-ABI methods for * enabling/disabling forwarding, and this is always 0. */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_ip_forward(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5182,7 +5188,6 @@ lxpr_read_sys_net_ipv4_ip_forward(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * illumos: tcp_smallest_anon_port & tcp_largest_anon_port * Not in tcp(7p) man page. */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_ip_lport_range(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5259,7 +5264,6 @@ lxpr_read_sys_net_ipv4_tcp_cc_curr(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * in the tcp_input_data() function on the use of tcp_fin_wait_2_flush_interval. * The value is in milliseconds. */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_tcp_fin_to(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5293,7 +5297,6 @@ lxpr_read_sys_net_ipv4_tcp_fin_to(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * 9 times (giving a total of 11.25 minutes) so we emulate this by dividing out * tcps_keepalive_abort_interval by 9. */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_tcp_ka_int(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5324,7 +5327,6 @@ lxpr_read_sys_net_ipv4_tcp_ka_int(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * The interval for sending out the first probe in milliseconds. The default is * two hours. */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_tcp_ka_tim(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5353,7 +5355,6 @@ lxpr_read_sys_net_ipv4_tcp_ka_tim(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * * illumos: tcp_conn_req_max_q0 */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_tcp_max_syn_bl(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5396,7 +5397,6 @@ lxpr_read_sys_net_ipv4_tcp_max_syn_bl(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * The interval_max value is the maximum RTO in ms. * The extra value is an extra time (in ms) to add in to the RTO. */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_tcp_retry2(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5436,7 +5436,6 @@ lxpr_read_sys_net_ipv4_tcp_retry2(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * tcp_xmit_hiwat is the default TCP send window size * tcp_max_buf is the maximum TCP send and receive buffer size */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_tcp_rwmem(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5476,7 +5475,6 @@ lxpr_read_sys_net_ipv4_tcp_rwmem(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * tcp_sack_permitted 0 == disabled, 1 == no initiate but accept, * 2 == initiate and accept. default is 2. */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_tcp_sack(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5509,7 +5507,6 @@ lxpr_read_sys_net_ipv4_tcp_sack(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * window scale option will be set only if the user has requested a send or * receive window larger than 64K. The default value of is 1. */ -/* ARGSUSED */ static void lxpr_read_sys_net_ipv4_tcp_winscale(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5573,7 +5570,6 @@ lxpr_read_sys_vm_dirty(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%u\n", val); } -/* ARGSUSED */ static void lxpr_read_sys_vm_max_map_cnt(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5582,7 +5578,6 @@ lxpr_read_sys_vm_max_map_cnt(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%d\n", 16777215); } -/* ARGSUSED */ static void lxpr_read_sys_vm_minfr_kb(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5590,7 +5585,6 @@ lxpr_read_sys_vm_minfr_kb(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%d\n", 0); } -/* ARGSUSED */ static void lxpr_read_sys_vm_nhpages(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5598,7 +5592,6 @@ lxpr_read_sys_vm_nhpages(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%d\n", 0); } -/* ARGSUSED */ static void lxpr_read_sys_vm_overcommit_mem(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5606,7 +5599,6 @@ lxpr_read_sys_vm_overcommit_mem(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_printf(uiobuf, "%d\n", 0); } -/* ARGSUSED */ static void lxpr_read_sys_vm_swappiness(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5620,7 +5612,6 @@ lxpr_read_sys_vm_swappiness(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * format is: "%.2lf, %.2lf",uptime_secs, idle_secs * Use fixed point arithmetic to get 2 decimal places */ -/* ARGSUSED */ static void lxpr_read_uptime(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -5685,7 +5676,6 @@ lxpr_read_uptime(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) * cgroup subsystems as being installed. The commented example below shows * how to print a subsystem entry. */ -/* ARGSUSED */ static void lxpr_read_cgroups(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -6000,7 +5990,6 @@ lx_cpuinfo_mapping_t lx_cpuinfo_mappings[] = { #define LX_CPUINFO_MAPPING_MAX \ (sizeof (lx_cpuinfo_mappings) / sizeof (lx_cpuinfo_mappings[0])) -/* ARGSUSED */ static void lxpr_read_cpuinfo(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -6131,7 +6120,6 @@ lxpr_read_cpuinfo(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) mutex_exit(&cpu_lock); } -/* ARGSUSED */ static void lxpr_read_fd(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -6139,12 +6127,71 @@ lxpr_read_fd(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) lxpr_uiobuf_seterr(uiobuf, EFAULT); } +static void +lxpr_read_fdinfo(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) +{ + zone_t *zone = LXPTOZ(lxpnp); + proc_t *p; + file_t *fp; + vnode_t *vp; + offset_t off; + short uf_flag; + int fd; + + ASSERT(lxpnp->lxpr_type == LXPR_PID_FDINFO_FD); + + p = lxpr_lock(lxpnp, NO_ZOMB); + if (p == NULL) { + lxpr_uiobuf_seterr(uiobuf, EINVAL); + return; + } + + if ((p->p_flag & SSYS) || p->p_as == &kas) { + lxpr_uiobuf_seterr(uiobuf, EFAULT); + lxpr_unlock(p); + return; + } + + fd = lxpnp->lxpr_desc; + + fp = lxpr_getf(p, fd, &uf_flag); + if (fp == NULL) { + lxpr_uiobuf_seterr(uiobuf, ENOENT); + lxpr_unlock(p); + return; + } + vp = fp->f_vnode; + + /* + * Check that the offset value in the underlying file_t is plausible + * and reset to 0 if not. + */ + if (fp->f_offset == -1) { + off = 0; + } else { + off = fp->f_offset; + if (VOP_SEEK(vp, 0, &off, NULL) != 0) + off = 0; + } + + lxpr_uiobuf_printf(uiobuf, "pos:\t%ld\n", off); + lxpr_uiobuf_printf(uiobuf, "flags:\t0%o\n", + lxpr_open_flags_convert(uf_flag, + fp->f_flag2 << 16 | fp->f_flag)); + lxpr_uiobuf_printf(uiobuf, "mnt_id:\t%u\n", + lxpr_get_mountid(zone, vp->v_vfsp)); + + /* Could show additional fields based on vp->v_type */ + + lxpr_releasef(p, fd); + lxpr_unlock(p); +} + /* * Report a list of file systems loaded in the kernel. We only report the ones * which we support and which may be checked by various components to see if * they are loaded. */ -/* ARGSUSED */ static void lxpr_read_filesystems(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf) { @@ -6269,6 +6316,10 @@ lxpr_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, */ if ((flags & FOLLOW) == 0) vap->va_type = VLNK; + case LXPR_PID_FDINFO_FD: + case LXPR_PID_TID_FDINFO_FD: + /* Linux leaves the file size for these as 0 */ + break; default: break; } @@ -6401,7 +6452,6 @@ lxpr_doaccess(lxpr_node_t *lxpnp, boolean_t shallow, int mode, int flags, return (EACCES); } -/* ARGSUSED */ static vnode_t * lxpr_lookup_not_a_dir(vnode_t *dp, char *comp) { @@ -6411,7 +6461,6 @@ lxpr_lookup_not_a_dir(vnode_t *dp, char *comp) /* * lxpr_lookup(): Vnode operation for VOP_LOOKUP() */ -/* ARGSUSED */ static int lxpr_lookup(vnode_t *dp, char *comp, vnode_t **vpp, pathname_t *pathp, int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct, @@ -6649,6 +6698,15 @@ lxpr_lookup_fddir(vnode_t *dp, char *comp) } static vnode_t * +lxpr_lookup_fdinfodir(vnode_t *dp, char *comp) +{ + ASSERT(VTOLXP(dp)->lxpr_type == LXPR_PID_FDINFODIR || + VTOLXP(dp)->lxpr_type == LXPR_PID_TID_FDINFODIR); + + return (lxpr_lookup_fdinfonode(dp, comp)); +} + +static vnode_t * lxpr_lookup_netdir(vnode_t *dp, char *comp) { ASSERT(VTOLXP(dp)->lxpr_type == LXPR_NETDIR); @@ -6790,7 +6848,6 @@ lxpr_lookup_sys_fs_inotifydir(vnode_t *dp, char *comp) /* * lxpr_readdir(): Vnode operation for VOP_READDIR() */ -/* ARGSUSED */ static int lxpr_readdir(vnode_t *dp, uio_t *uiop, cred_t *cr, int *eofp, caller_context_t *ct, int flags) @@ -6832,7 +6889,6 @@ lxpr_readdir(vnode_t *dp, uio_t *uiop, cred_t *cr, int *eofp, return (lxpr_readdir_function[type](lxpnp, uiop, eofp)); } -/* ARGSUSED */ static int lxpr_readdir_not_a_dir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp) { @@ -7267,7 +7323,8 @@ lxpr_readdir_task_tid_dir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp) } static int -lxpr_readdir_fddir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp) +lxpr_readdir_fdlist(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp, + lxpr_nodetype_t inodetype) { /* bp holds one dirent64 structure */ longlong_t bp[DIRENT64_RECLEN(LXPNSIZ) / sizeof (longlong_t)]; @@ -7279,8 +7336,13 @@ lxpr_readdir_fddir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp) proc_t *p; uf_info_t *fip; - ASSERT(lxpnp->lxpr_type == LXPR_PID_FDDIR || - lxpnp->lxpr_type == LXPR_PID_TID_FDDIR); + ASSERT( + (inodetype == LXPR_PID_FD_FD && ( + lxpnp->lxpr_type == LXPR_PID_FDDIR || + lxpnp->lxpr_type == LXPR_PID_TID_FDDIR)) || + (inodetype == LXPR_PID_FDINFO_FD && ( + lxpnp->lxpr_type == LXPR_PID_FDINFODIR || + lxpnp->lxpr_type == LXPR_PID_TID_FDINFODIR))); oresid = uiop->uio_resid; @@ -7345,7 +7407,7 @@ lxpr_readdir_fddir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp) if (fip->fi_list[fd].uf_file == NULL) continue; - dirent->d_ino = lxpr_inode(LXPR_PID_FD_FD, p->p_pid, fd); + dirent->d_ino = lxpr_inode(inodetype, p->p_pid, fd); len = snprintf(dirent->d_name, LXPNSIZ, "%d", fd); ASSERT(len < LXPNSIZ); reclen = DIRENT64_RECLEN(len); @@ -7380,6 +7442,18 @@ out: } static int +lxpr_readdir_fddir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp) +{ + return (lxpr_readdir_fdlist(lxpnp, uiop, eofp, LXPR_PID_FD_FD)); +} + +static int +lxpr_readdir_fdinfodir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp) +{ + return (lxpr_readdir_fdlist(lxpnp, uiop, eofp, LXPR_PID_FDINFO_FD)); +} + +static int lxpr_readdir_sysdir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp) { ASSERT(lxpnp->lxpr_type == LXPR_SYSDIR); @@ -7491,7 +7565,6 @@ lxpr_tokenize_num(char *str, long *pv, char **ep) return (0); } -/* ARGSUSED */ static int lxpr_write_tcp_property(lxpr_node_t *lxpnp, struct uio *uio, struct cred *cr, caller_context_t *ct, char *prop, @@ -7644,7 +7717,6 @@ lxpr_xlate_sack(char *val, int size) * We're updating a property on the ip stack so we can't reuse * lxpr_write_tcp_property. */ -/* ARGSUSED */ static int lxpr_write_sys_net_ipv4_icmp_eib(lxpr_node_t *lxpnp, struct uio *uio, struct cred *cr, caller_context_t *ct) @@ -7702,7 +7774,6 @@ lxpr_write_sys_net_ipv4_icmp_eib(lxpr_node_t *lxpnp, struct uio *uio, * set two properties on the netstack_tcp, so we can't reuse * lxpr_write_tcp_property. */ -/* ARGSUSED */ static int lxpr_write_sys_net_ipv4_ip_lport_range(lxpr_node_t *lxpnp, struct uio *uio, struct cred *cr, caller_context_t *ct) @@ -7785,7 +7856,6 @@ lxpr_write_sys_net_ipv4_ip_lport_range(lxpr_node_t *lxpnp, struct uio *uio, * * See the Linux tcp(7) man page. */ -/* ARGSUSED */ static int lxpr_write_sys_net_ipv4_tcp_rwmem(lxpr_node_t *lxpnp, struct uio *uio, struct cred *cr, caller_context_t *ct) @@ -7952,7 +8022,6 @@ lxpr_write_sys_net_ipv4_tcp_winscale(lxpr_node_t *lxpnp, struct uio *uio, NULL)); } -/* ARGSUSED */ static int lxpr_write_sys_fs_pipe_max(lxpr_node_t *lxpnp, struct uio *uio, struct cred *cr, caller_context_t *ct) @@ -8005,7 +8074,6 @@ lxpr_write_sys_fs_pipe_max(lxpr_node_t *lxpnp, struct uio *uio, return (0); } -/* ARGSUSED */ static int lxpr_write_sys_kernel_corepatt(lxpr_node_t *lxpnp, struct uio *uio, struct cred *cr, caller_context_t *ct) @@ -8066,7 +8134,6 @@ lxpr_write_sys_kernel_corepatt(lxpr_node_t *lxpnp, struct uio *uio, return (0); } -/* ARGSUSED */ static int lxpr_write_pid_loginuid(lxpr_node_t *lxpnp, struct uio *uio, struct cred *cr, caller_context_t *ct) @@ -8167,12 +8234,11 @@ lxpr_readlink_exe(lxpr_node_t *lxpnp, char *buf, size_t size, cred_t *cr) /* * lxpr_readlink(): Vnode operation for VOP_READLINK() */ -/* ARGSUSED */ static int lxpr_readlink(vnode_t *vp, uio_t *uiop, cred_t *cr, caller_context_t *ct) { - char bp[MAXPATHLEN + 1]; - size_t buflen = sizeof (bp); + char *bp; + size_t buflen, klen; lxpr_node_t *lxpnp = VTOLXP(vp); vnode_t *rvp = lxpnp->lxpr_realvp; pid_t pid; @@ -8199,11 +8265,14 @@ lxpr_readlink(vnode_t *vp, uio_t *uiop, cred_t *cr, caller_context_t *ct) if (vp->v_type != VLNK && lxpnp->lxpr_type != LXPR_PID_FD_FD) return (EINVAL); + buflen = klen = MAXPATHLEN + 1; + bp = kmem_alloc(klen, KM_SLEEP); + /* Try to produce a symlink name for anything that has a realvp */ if (rvp != NULL) { error = lxpr_doaccess(lxpnp, B_TRUE, VREAD, 0, cr, ct); if (error != 0) - return (error); + goto out; error = vnodetopath(NULL, rvp, bp, buflen, cr); @@ -8221,7 +8290,7 @@ lxpr_readlink(vnode_t *vp, uio_t *uiop, cred_t *cr, caller_context_t *ct) */ if (lxpnp->lxpr_type != LXPR_PID_FD_FD || lxpr_readlink_fdnode(lxpnp, bp, buflen) != 0) { - return (error); + goto out; } } } else { @@ -8239,18 +8308,24 @@ lxpr_readlink(vnode_t *vp, uio_t *uiop, cred_t *cr, caller_context_t *ct) case LXPR_PID_CURDIR: case LXPR_PID_ROOTDIR: case LXPR_PID_EXE: - return (EACCES); + error = EACCES; + goto out; default: /* * Need to return error so that nothing thinks * that the symlink is empty and hence "." */ - return (EINVAL); + error = EINVAL; + goto out; } } /* copy the link data to user space */ - return (uiomove(bp, strlen(bp), UIO_READ, uiop)); + error = uiomove(bp, strlen(bp), UIO_READ, uiop); + +out: + kmem_free(bp, klen); + return (error); } @@ -8259,7 +8334,6 @@ lxpr_readlink(vnode_t *vp, uio_t *uiop, cred_t *cr, caller_context_t *ct) * Vnode is no longer referenced, deallocate the file * and all its resources. */ -/* ARGSUSED */ static void lxpr_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct) { @@ -8322,7 +8396,6 @@ lxpr_realvp(vnode_t *vp, vnode_t **vpp, caller_context_t *ct) /* Pollhead for fake POLLET support below */ static struct pollhead lxpr_pollhead; -/* ARGSUSED */ static int lxpr_poll(vnode_t *vp, short ev, int anyyet, short *reventsp, pollhead_t **phpp, caller_context_t *ct) @@ -8358,7 +8431,6 @@ lxpr_poll(vnode_t *vp, short ev, int anyyet, short *reventsp, return (0); } -/* ARGSUSED */ static int lxpr_write(vnode_t *vp, uio_t *uiop, int ioflag, cred_t *cr, caller_context_t *ct) @@ -8383,7 +8455,6 @@ lxpr_write(vnode_t *vp, uio_t *uiop, int ioflag, cred_t *cr, } /* Needed for writable files which are first "truncated" */ -/* ARGSUSED */ static int lxpr_space(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset, cred_t *cred, caller_context_t *ct) @@ -8402,7 +8473,6 @@ lxpr_space(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset, * Needed for writable files which are first "truncated". We only support * truncation. */ -/* ARGSUSED */ static int lxpr_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, caller_context_t *ct) @@ -8420,7 +8490,6 @@ lxpr_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, /* * We need to allow open with O_CREAT for the writable files. */ -/* ARGSUSED */ static int lxpr_create(vnode_t *dvp, char *nm, vattr_t *vap, enum vcexcl exclusive, int mode, vnode_t **vpp, cred_t *cr, int flag, caller_context_t *ct, diff --git a/usr/src/uts/common/brand/lx/syscall/lx_mount.c b/usr/src/uts/common/brand/lx/syscall/lx_mount.c index 2524e9044a..fff6c81339 100644 --- a/usr/src/uts/common/brand/lx/syscall/lx_mount.c +++ b/usr/src/uts/common/brand/lx/syscall/lx_mount.c @@ -118,6 +118,7 @@ static mount_opt_t lx_tmpfs_options[] = { { "mode", MOUNT_OPT_UINT }, { "uid", MOUNT_OPT_UINT }, { "gid", MOUNT_OPT_UINT }, + { "nr_inodes", MOUNT_OPT_PASSTHRU }, { NULL, MOUNT_OPT_INVALID } }; @@ -312,7 +313,8 @@ lx_mnt_opt_rm(char *opts, char *rmopt, char *retstr, int retlen) ASSERT((opts != NULL) && (rmopt != NULL)); - retstr[0] = '\0'; + if (retstr != NULL) + retstr[0] = '\0'; /* If no options were specified, there's no problem. */ if (opts_len == 0) @@ -324,12 +326,13 @@ lx_mnt_opt_rm(char *opts, char *rmopt, char *retstr, int retlen) for (optend = optstart; *optend != ',' && *optend != '\0'; optend++) ; - /*LINTED*/ optlen = optend - optstart; - if (optlen >= retlen) - return (-1); - (void) strncpy(retstr, optstart, optlen); - retstr[optlen] = '\0'; + if (retstr != NULL) { + if (optlen >= retlen) + return (-1); + (void) strncpy(retstr, optstart, optlen); + retstr[optlen] = '\0'; + } if (*optend == ',') optend++; @@ -414,9 +417,8 @@ lx_mount(const char *sourcep, const char *targetp, const char *fstypep, uint_t flags, const void *datap) { char fstype[16]; - char source[MAXPATHLEN]; - char target[MAXPATHLEN]; - char options[MAX_MNTOPT_STR]; + char *source, *target, *options; + size_t sourcel, targetl, optionsl; int sflags, rv; struct mounta ma, *map = &ma; vfs_t *vfsp; @@ -452,15 +454,25 @@ lx_mount(const char *sourcep, const char *targetp, const char *fstypep, return (0); } + /* Make sure we support the requested mount flags. */ + if ((flags & ~LX_MS_SUPPORTED) != 0) + return (set_errno(ENOTSUP)); + + sourcel = targetl = MAXPATHLEN; + optionsl = MAX_MNTOPT_STR; + source = kmem_alloc(sourcel, KM_SLEEP); + target = kmem_alloc(targetl, KM_SLEEP); + options = kmem_alloc(optionsl, KM_SLEEP); + sflags = MS_SYSSPACE | MS_OPTIONSTR; options[0] = '\0'; /* Copy in parameters that are always present. */ - if ((rv = lx_mnt_copyin_arg(sourcep, source, sizeof (source))) != 0) - return (set_errno(rv)); + if ((rv = lx_mnt_copyin_arg(sourcep, source, sourcel)) != 0) + goto out; - if ((rv = lx_mnt_copyin_arg(targetp, target, sizeof (target))) != 0) - return (set_errno(rv)); + if ((rv = lx_mnt_copyin_arg(targetp, target, targetl)) != 0) + goto out; /* * While SunOS is picky about mount(2) target paths being absolute, @@ -468,26 +480,62 @@ lx_mount(const char *sourcep, const char *targetp, const char *fstypep, * requirement we must lookup the full path. */ if (target[0] != '/') { - vnode_t *vp; - - if ((rv = lookupnameatcred(target, UIO_SYSSPACE, FOLLOW, - NULLVPP, &vp, NULL, CRED())) != 0) - return (set_errno(rv)); + if ((rv = lookupname(target, UIO_SYSSPACE, FOLLOW, + NULLVPP, &vp)) != 0) { + goto out; + } - rv = vnodetopath(NULL, vp, target, MAXPATHLEN, CRED()); + rv = vnodetopath(NULL, vp, target, targetl, CRED()); VN_RELE(vp); if (rv != 0) - return (set_errno(rv)); + goto out; } - /* Make sure we support the requested mount flags. */ - if ((flags & ~LX_MS_SUPPORTED) != 0) - return (set_errno(ENOTSUP)); + /* + * Following commits* in September 2020, systemd mounts most + * filesystems via an open file descriptor in order to avoid + * following mount point symlinks. + * It does this by opening the mount point with O_NOFOLLOW|O_PATH + * and then performing the mount on /proc/self/fd/<fd>. + * In illumos, this results in a mount attempt on the lx_proc vnode + * instead of the intended mount point, which fails since /proc files + * are generally marked as unmountable (and we wouldn't want this + * anyway). + * As a workaround, the /proc vnode's link target is retrieved and + * used for mount. This lookup causes procfs to return the path of + * the vnode's realvp which is the path of the underlying directory. + * Additonally, since systemd is holding an open file descriptor for + * the mount point, overlay mounts also need to be allowed. + * + * * https://github.com/systemd/systemd/commit/28126409b20bca9aa6f + * * https://github.com/systemd/systemd/commit/21935150a0c42b91a32 + */ + if (strncmp(target, "/proc/self/fd/", 14) == 0 && + lookupname(target, UIO_SYSSPACE, NO_FOLLOW, NULLVPP, &vp) == 0) { + struct iovec iov = {0}; + struct uio uio = {0}; + + iov.iov_base = target; + iov.iov_len = targetl; + uio.uio_iov = &iov; + uio.uio_iovcnt = 1; + uio.uio_resid = targetl; + uio.uio_segflg = UIO_SYSSPACE; + uio.uio_llimit = MAXOFFSET_T; + + rv = VOP_READLINK(vp, &uio, CRED(), NULL); + VN_RELE(vp); + if (rv != 0) + goto out; + target[targetl - uio.uio_resid] = '\0'; + sflags |= MS_OVERLAY; + } /* Copy in Linux mount options. */ if (datap != NULL && - (rv = lx_mnt_copyin_arg(datap, options, sizeof (options))) != 0) - return (set_errno(rv)); + (rv = lx_mnt_copyin_arg(datap, options, optionsl)) != 0) { + goto out; + } /* Do filesystem specific mount work. */ if (flags & LX_MS_BIND) { @@ -496,44 +544,56 @@ lx_mount(const char *sourcep, const char *targetp, const char *fstypep, /* Verify Linux mount options. */ if ((rv = lx_mnt_opt_verify(options, lofs_options)) != 0) - return (set_errno(rv)); + goto out; } else if (strcmp(fstype, "tmpfs") == 0) { char idstr[64]; /* Verify Linux mount options. */ if ((rv = lx_mnt_opt_verify(options, lx_tmpfs_options)) != 0) - return (set_errno(rv)); + goto out; /* * Linux defaults to mode=1777 for tmpfs mounts. */ if (strstr(options, "mode=") == NULL) { if (options[0] != '\0') - (void) strlcat(options, ",", sizeof (options)); - (void) strlcat(options, "mode=1777", sizeof (options)); + (void) strlcat(options, ",", optionsl); + (void) strlcat(options, "mode=1777", optionsl); } + /* + * Linux supports "nr_inodes=<val>" for tmpfs. There is no + * analogue in illumos - drop the option. + */ + (void) lx_mnt_opt_rm(options, "nr_inodes=", NULL, 0); + switch (lx_mnt_opt_rm(options, "uid=", idstr, sizeof (idstr))) { case 0: uid = -1; break; case 1: - if (lx_mnt_opt_val(idstr, &uid) < 0) - return (set_errno(EINVAL)); + if (lx_mnt_opt_val(idstr, &uid) < 0) { + rv = EINVAL; + goto out; + } break; default: - return (set_errno(E2BIG)); + rv = E2BIG; + goto out; } switch (lx_mnt_opt_rm(options, "gid=", idstr, sizeof (idstr))) { case 0: gid = -1; break; case 1: - if (lx_mnt_opt_val(idstr, &gid) < 0) - return (set_errno(EINVAL)); + if (lx_mnt_opt_val(idstr, &gid) < 0) { + rv = EINVAL; + goto out; + } break; default: - return (set_errno(E2BIG)); + rv = E2BIG; + goto out; } /* @@ -550,14 +610,14 @@ lx_mount(const char *sourcep, const char *targetp, const char *fstypep, /* Verify Linux mount options. */ if ((rv = lx_mnt_opt_verify(options, lx_proc_options)) != 0) - return (set_errno(rv)); + goto out; } else if (strcmp(fstype, "sysfs") == 0) { /* Translate sysfs mount requests to lx_sysfs requests. */ (void) strcpy(fstype, "lx_sysfs"); /* Verify Linux mount options. */ if ((rv = lx_mnt_opt_verify(options, lx_sysfs_options)) != 0) - return (set_errno(rv)); + goto out; } else if (strcmp(fstype, "cgroup") == 0) { /* Translate cgroup mount requests to lx_cgroup requests. */ (void) strcpy(fstype, "lx_cgroup"); @@ -572,7 +632,7 @@ lx_mount(const char *sourcep, const char *targetp, const char *fstypep, /* Verify Linux mount options. */ if ((rv = lx_mnt_opt_verify(options, lx_autofs_options)) != 0) - return (set_errno(rv)); + goto out; /* Linux seems to always allow overlay mounts */ sflags |= MS_OVERLAY; @@ -599,23 +659,27 @@ lx_mount(const char *sourcep, const char *targetp, const char *fstypep, flags &= ~(LX_MS_RELATIME | LX_MS_NOATIME); } if ((flags & LX_MS_NODEV) && - (rv = lx_mnt_add_opt("nodev", options, sizeof (options))) != 0) - return (set_errno(rv)); + (rv = lx_mnt_add_opt("nodev", options, optionsl)) != 0) { + goto out; + } if ((flags & LX_MS_NOEXEC) && - (rv = lx_mnt_add_opt("noexec", options, sizeof (options))) != 0) - return (set_errno(rv)); + (rv = lx_mnt_add_opt("noexec", options, optionsl)) != 0) { + goto out; + } if ((flags & LX_MS_NOATIME) && - (rv = lx_mnt_add_opt("noatime", options, sizeof (options))) != 0) - return (set_errno(rv)); + (rv = lx_mnt_add_opt("noatime", options, optionsl)) != 0) { + goto out; + } if ((rv = lookupname(target, UIO_SYSSPACE, FOLLOW, NULLVPP, &vp)) != 0) - return (set_errno(rv)); + goto out; /* If mounting proc over itself, just return ok */ if (strcmp(fstype, "lx_proc") == 0 && strcmp("lx_proc", vfssw[vp->v_vfsp->vfs_fstype].vsw_name) == 0) { VN_RELE(vp); - return (0); + rv = 0; + goto out; } map->spec = source; @@ -625,12 +689,12 @@ lx_mount(const char *sourcep, const char *targetp, const char *fstypep, map->dataptr = NULL; map->datalen = 0; map->optptr = options; - map->optlen = sizeof (options); + map->optlen = optionsl; rv = domount(NULL, map, vp, CRED(), &vfsp); VN_RELE(vp); if (rv != 0) - return (set_errno(rv)); + goto out; VFS_RELE(vfsp); if (strcmp(fstype, "tmpfs") == 0 && (uid != -1 || gid != -1)) { @@ -642,7 +706,12 @@ lx_mount(const char *sourcep, const char *targetp, const char *fstypep, } } - return (0); +out: + kmem_free(source, sourcel); + kmem_free(target, targetl); + kmem_free(options, optionsl); + + return (rv == 0 ? rv : set_errno(rv)); } /* diff --git a/usr/src/uts/common/brand/lx/syscall/lx_stat.c b/usr/src/uts/common/brand/lx/syscall/lx_stat.c index 9af0080138..0f5460816b 100644 --- a/usr/src/uts/common/brand/lx/syscall/lx_stat.c +++ b/usr/src/uts/common/brand/lx/syscall/lx_stat.c @@ -393,7 +393,7 @@ lx_fstatat64(int fd, char *name, void *outp, int flag) vnode_t *vp = NULL; cred_t *cr = NULL; model_t model = get_udatamodel(); - enum symfollow follow = FOLLOW; + int follow = FOLLOW; int error; char c; @@ -403,18 +403,13 @@ lx_fstatat64(int fd, char *name, void *outp, int flag) if ((flag & ~LX_FSTATAT_ALLOWED) != 0) { return (set_errno(EINVAL)); } - if ((flag & LX_AT_NO_AUTOMOUNT) != 0) { - /* - * While AT_NO_AUTOMOUNT is a legal flag for fstatat64, it is - * not yet supported by lx_autofs. - */ - lx_unsupported("fstatat(AT_NO_AUTOMOUNT)"); - return (set_errno(EINVAL)); - } if ((flag & LX_AT_SYMLINK_NOFOLLOW) != 0) { follow = NO_FOLLOW; } + if ((flag & LX_AT_NO_AUTOMOUNT) != 0) + follow |= __FLXNOAUTO; + if (copyin(name, &c, sizeof (c)) != 0) { return (set_errno(EFAULT)); } diff --git a/usr/src/uts/common/fs/lookup.c b/usr/src/uts/common/fs/lookup.c index 093db5a4b4..32a3d19708 100644 --- a/usr/src/uts/common/fs/lookup.c +++ b/usr/src/uts/common/fs/lookup.c @@ -246,6 +246,9 @@ lookuppnvp( pp = &presrvd; } + if (flags & __FLXNOAUTO) + lookup_flags |= __FLXNOAUTO; + if (auditing) audit_anchorpath(pnp, vp == rootvp); diff --git a/usr/src/uts/common/sys/file.h b/usr/src/uts/common/sys/file.h index be96a68d46..66620ab7b9 100644 --- a/usr/src/uts/common/sys/file.h +++ b/usr/src/uts/common/sys/file.h @@ -125,6 +125,11 @@ typedef struct fpollinfo { * Private interface for lx O_PATH|O_NOFOLLOW emulation for symlinks. */ #define __FLXPATH 0x80000000 +/* + * Private interface for lx fstatat(AT_NO_AUTOMOUNT) emulation. + * Since usage is disjoint, the __FLXPATH bit is re-used. + */ +#define __FLXNOAUTO 0x80000000 #if defined(_KERNEL) || defined(_FAKE_KERNEL) |