diff options
Diffstat (limited to 'fs')
-rw-r--r-- | fs/Makefile | 2 | ||||
-rw-r--r-- | fs/binfmt_flat.c | 30 | ||||
-rw-r--r-- | fs/compat.c | 177 | ||||
-rw-r--r-- | fs/configfs/dir.c | 137 | ||||
-rw-r--r-- | fs/exportfs/expfs.c | 2 | ||||
-rw-r--r-- | fs/inotify.c | 9 | ||||
-rw-r--r-- | fs/nfsd/vfs.c | 7 | ||||
-rw-r--r-- | fs/ocfs2/aops.c | 46 | ||||
-rw-r--r-- | fs/ocfs2/aops.h | 4 | ||||
-rw-r--r-- | fs/ocfs2/extent_map.c | 6 | ||||
-rw-r--r-- | fs/ocfs2/file.c | 86 | ||||
-rw-r--r-- | fs/ocfs2/journal.c | 8 | ||||
-rw-r--r-- | fs/ocfs2/uptodate.c | 4 | ||||
-rw-r--r-- | fs/ocfs2/vote.c | 6 |
14 files changed, 324 insertions, 200 deletions
diff --git a/fs/Makefile b/fs/Makefile index 83bf478e786b..078d3d1191a5 100644 --- a/fs/Makefile +++ b/fs/Makefile | |||
@@ -45,6 +45,7 @@ obj-$(CONFIG_DNOTIFY) += dnotify.o | |||
45 | obj-$(CONFIG_PROC_FS) += proc/ | 45 | obj-$(CONFIG_PROC_FS) += proc/ |
46 | obj-y += partitions/ | 46 | obj-y += partitions/ |
47 | obj-$(CONFIG_SYSFS) += sysfs/ | 47 | obj-$(CONFIG_SYSFS) += sysfs/ |
48 | obj-$(CONFIG_CONFIGFS_FS) += configfs/ | ||
48 | obj-y += devpts/ | 49 | obj-y += devpts/ |
49 | 50 | ||
50 | obj-$(CONFIG_PROFILING) += dcookies.o | 51 | obj-$(CONFIG_PROFILING) += dcookies.o |
@@ -100,5 +101,4 @@ obj-$(CONFIG_BEFS_FS) += befs/ | |||
100 | obj-$(CONFIG_HOSTFS) += hostfs/ | 101 | obj-$(CONFIG_HOSTFS) += hostfs/ |
101 | obj-$(CONFIG_HPPFS) += hppfs/ | 102 | obj-$(CONFIG_HPPFS) += hppfs/ |
102 | obj-$(CONFIG_DEBUG_FS) += debugfs/ | 103 | obj-$(CONFIG_DEBUG_FS) += debugfs/ |
103 | obj-$(CONFIG_CONFIGFS_FS) += configfs/ | ||
104 | obj-$(CONFIG_OCFS2_FS) += ocfs2/ | 104 | obj-$(CONFIG_OCFS2_FS) += ocfs2/ |
diff --git a/fs/binfmt_flat.c b/fs/binfmt_flat.c index 69f44dcdb0b4..b1c902e319c1 100644 --- a/fs/binfmt_flat.c +++ b/fs/binfmt_flat.c | |||
@@ -428,7 +428,6 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
428 | loff_t fpos; | 428 | loff_t fpos; |
429 | unsigned long start_code, end_code; | 429 | unsigned long start_code, end_code; |
430 | int ret; | 430 | int ret; |
431 | int exec_fileno; | ||
432 | 431 | ||
433 | hdr = ((struct flat_hdr *) bprm->buf); /* exec-header */ | 432 | hdr = ((struct flat_hdr *) bprm->buf); /* exec-header */ |
434 | inode = bprm->file->f_dentry->d_inode; | 433 | inode = bprm->file->f_dentry->d_inode; |
@@ -502,21 +501,12 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
502 | goto err; | 501 | goto err; |
503 | } | 502 | } |
504 | 503 | ||
505 | /* check file descriptor */ | ||
506 | exec_fileno = get_unused_fd(); | ||
507 | if (exec_fileno < 0) { | ||
508 | ret = -EMFILE; | ||
509 | goto err; | ||
510 | } | ||
511 | get_file(bprm->file); | ||
512 | fd_install(exec_fileno, bprm->file); | ||
513 | |||
514 | /* Flush all traces of the currently running executable */ | 504 | /* Flush all traces of the currently running executable */ |
515 | if (id == 0) { | 505 | if (id == 0) { |
516 | result = flush_old_exec(bprm); | 506 | result = flush_old_exec(bprm); |
517 | if (result) { | 507 | if (result) { |
518 | ret = result; | 508 | ret = result; |
519 | goto err_close; | 509 | goto err; |
520 | } | 510 | } |
521 | 511 | ||
522 | /* OK, This is the point of no return */ | 512 | /* OK, This is the point of no return */ |
@@ -548,7 +538,7 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
548 | textpos = (unsigned long) -ENOMEM; | 538 | textpos = (unsigned long) -ENOMEM; |
549 | printk("Unable to mmap process text, errno %d\n", (int)-textpos); | 539 | printk("Unable to mmap process text, errno %d\n", (int)-textpos); |
550 | ret = textpos; | 540 | ret = textpos; |
551 | goto err_close; | 541 | goto err; |
552 | } | 542 | } |
553 | 543 | ||
554 | down_write(¤t->mm->mmap_sem); | 544 | down_write(¤t->mm->mmap_sem); |
@@ -564,7 +554,7 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
564 | (int)-datapos); | 554 | (int)-datapos); |
565 | do_munmap(current->mm, textpos, text_len); | 555 | do_munmap(current->mm, textpos, text_len); |
566 | ret = realdatastart; | 556 | ret = realdatastart; |
567 | goto err_close; | 557 | goto err; |
568 | } | 558 | } |
569 | datapos = realdatastart + MAX_SHARED_LIBS * sizeof(unsigned long); | 559 | datapos = realdatastart + MAX_SHARED_LIBS * sizeof(unsigned long); |
570 | 560 | ||
@@ -587,7 +577,7 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
587 | do_munmap(current->mm, textpos, text_len); | 577 | do_munmap(current->mm, textpos, text_len); |
588 | do_munmap(current->mm, realdatastart, data_len + extra); | 578 | do_munmap(current->mm, realdatastart, data_len + extra); |
589 | ret = result; | 579 | ret = result; |
590 | goto err_close; | 580 | goto err; |
591 | } | 581 | } |
592 | 582 | ||
593 | reloc = (unsigned long *) (datapos+(ntohl(hdr->reloc_start)-text_len)); | 583 | reloc = (unsigned long *) (datapos+(ntohl(hdr->reloc_start)-text_len)); |
@@ -606,7 +596,7 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
606 | printk("Unable to allocate RAM for process text/data, errno %d\n", | 596 | printk("Unable to allocate RAM for process text/data, errno %d\n", |
607 | (int)-textpos); | 597 | (int)-textpos); |
608 | ret = textpos; | 598 | ret = textpos; |
609 | goto err_close; | 599 | goto err; |
610 | } | 600 | } |
611 | 601 | ||
612 | realdatastart = textpos + ntohl(hdr->data_start); | 602 | realdatastart = textpos + ntohl(hdr->data_start); |
@@ -652,7 +642,7 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
652 | do_munmap(current->mm, textpos, text_len + data_len + extra + | 642 | do_munmap(current->mm, textpos, text_len + data_len + extra + |
653 | MAX_SHARED_LIBS * sizeof(unsigned long)); | 643 | MAX_SHARED_LIBS * sizeof(unsigned long)); |
654 | ret = result; | 644 | ret = result; |
655 | goto err_close; | 645 | goto err; |
656 | } | 646 | } |
657 | } | 647 | } |
658 | 648 | ||
@@ -717,7 +707,7 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
717 | addr = calc_reloc(*rp, libinfo, id, 0); | 707 | addr = calc_reloc(*rp, libinfo, id, 0); |
718 | if (addr == RELOC_FAILED) { | 708 | if (addr == RELOC_FAILED) { |
719 | ret = -ENOEXEC; | 709 | ret = -ENOEXEC; |
720 | goto err_close; | 710 | goto err; |
721 | } | 711 | } |
722 | *rp = addr; | 712 | *rp = addr; |
723 | } | 713 | } |
@@ -747,7 +737,7 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
747 | rp = (unsigned long *) calc_reloc(addr, libinfo, id, 1); | 737 | rp = (unsigned long *) calc_reloc(addr, libinfo, id, 1); |
748 | if (rp == (unsigned long *)RELOC_FAILED) { | 738 | if (rp == (unsigned long *)RELOC_FAILED) { |
749 | ret = -ENOEXEC; | 739 | ret = -ENOEXEC; |
750 | goto err_close; | 740 | goto err; |
751 | } | 741 | } |
752 | 742 | ||
753 | /* Get the pointer's value. */ | 743 | /* Get the pointer's value. */ |
@@ -762,7 +752,7 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
762 | addr = calc_reloc(addr, libinfo, id, 0); | 752 | addr = calc_reloc(addr, libinfo, id, 0); |
763 | if (addr == RELOC_FAILED) { | 753 | if (addr == RELOC_FAILED) { |
764 | ret = -ENOEXEC; | 754 | ret = -ENOEXEC; |
765 | goto err_close; | 755 | goto err; |
766 | } | 756 | } |
767 | 757 | ||
768 | /* Write back the relocated pointer. */ | 758 | /* Write back the relocated pointer. */ |
@@ -783,8 +773,6 @@ static int load_flat_file(struct linux_binprm * bprm, | |||
783 | stack_len); | 773 | stack_len); |
784 | 774 | ||
785 | return 0; | 775 | return 0; |
786 | err_close: | ||
787 | sys_close(exec_fileno); | ||
788 | err: | 776 | err: |
789 | return ret; | 777 | return ret; |
790 | } | 778 | } |
diff --git a/fs/compat.c b/fs/compat.c index 01f39f87f372..b1f64786a613 100644 --- a/fs/compat.c +++ b/fs/compat.c | |||
@@ -2030,109 +2030,115 @@ union compat_nfsctl_res { | |||
2030 | struct knfsd_fh cr32_getfs; | 2030 | struct knfsd_fh cr32_getfs; |
2031 | }; | 2031 | }; |
2032 | 2032 | ||
2033 | static int compat_nfs_svc_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg) | 2033 | static int compat_nfs_svc_trans(struct nfsctl_arg *karg, |
2034 | { | 2034 | struct compat_nfsctl_arg __user *arg) |
2035 | int err; | 2035 | { |
2036 | 2036 | if (!access_ok(VERIFY_READ, &arg->ca32_svc, sizeof(arg->ca32_svc)) || | |
2037 | err = access_ok(VERIFY_READ, &arg->ca32_svc, sizeof(arg->ca32_svc)); | 2037 | get_user(karg->ca_version, &arg->ca32_version) || |
2038 | err |= get_user(karg->ca_version, &arg->ca32_version); | 2038 | __get_user(karg->ca_svc.svc_port, &arg->ca32_svc.svc32_port) || |
2039 | err |= __get_user(karg->ca_svc.svc_port, &arg->ca32_svc.svc32_port); | 2039 | __get_user(karg->ca_svc.svc_nthreads, |
2040 | err |= __get_user(karg->ca_svc.svc_nthreads, &arg->ca32_svc.svc32_nthreads); | 2040 | &arg->ca32_svc.svc32_nthreads)) |
2041 | return (err) ? -EFAULT : 0; | 2041 | return -EFAULT; |
2042 | return 0; | ||
2042 | } | 2043 | } |
2043 | 2044 | ||
2044 | static int compat_nfs_clnt_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg) | 2045 | static int compat_nfs_clnt_trans(struct nfsctl_arg *karg, |
2045 | { | 2046 | struct compat_nfsctl_arg __user *arg) |
2046 | int err; | 2047 | { |
2047 | 2048 | if (!access_ok(VERIFY_READ, &arg->ca32_client, | |
2048 | err = access_ok(VERIFY_READ, &arg->ca32_client, sizeof(arg->ca32_client)); | 2049 | sizeof(arg->ca32_client)) || |
2049 | err |= get_user(karg->ca_version, &arg->ca32_version); | 2050 | get_user(karg->ca_version, &arg->ca32_version) || |
2050 | err |= __copy_from_user(&karg->ca_client.cl_ident[0], | 2051 | __copy_from_user(&karg->ca_client.cl_ident[0], |
2051 | &arg->ca32_client.cl32_ident[0], | 2052 | &arg->ca32_client.cl32_ident[0], |
2052 | NFSCLNT_IDMAX); | 2053 | NFSCLNT_IDMAX) || |
2053 | err |= __get_user(karg->ca_client.cl_naddr, &arg->ca32_client.cl32_naddr); | 2054 | __get_user(karg->ca_client.cl_naddr, |
2054 | err |= __copy_from_user(&karg->ca_client.cl_addrlist[0], | 2055 | &arg->ca32_client.cl32_naddr) || |
2055 | &arg->ca32_client.cl32_addrlist[0], | 2056 | __copy_from_user(&karg->ca_client.cl_addrlist[0], |
2056 | (sizeof(struct in_addr) * NFSCLNT_ADDRMAX)); | 2057 | &arg->ca32_client.cl32_addrlist[0], |
2057 | err |= __get_user(karg->ca_client.cl_fhkeytype, | 2058 | (sizeof(struct in_addr) * NFSCLNT_ADDRMAX)) || |
2058 | &arg->ca32_client.cl32_fhkeytype); | 2059 | __get_user(karg->ca_client.cl_fhkeytype, |
2059 | err |= __get_user(karg->ca_client.cl_fhkeylen, | 2060 | &arg->ca32_client.cl32_fhkeytype) || |
2060 | &arg->ca32_client.cl32_fhkeylen); | 2061 | __get_user(karg->ca_client.cl_fhkeylen, |
2061 | err |= __copy_from_user(&karg->ca_client.cl_fhkey[0], | 2062 | &arg->ca32_client.cl32_fhkeylen) || |
2062 | &arg->ca32_client.cl32_fhkey[0], | 2063 | __copy_from_user(&karg->ca_client.cl_fhkey[0], |
2063 | NFSCLNT_KEYMAX); | 2064 | &arg->ca32_client.cl32_fhkey[0], |
2065 | NFSCLNT_KEYMAX)) | ||
2066 | return -EFAULT; | ||
2064 | 2067 | ||
2065 | return (err) ? -EFAULT : 0; | 2068 | return 0; |
2066 | } | 2069 | } |
2067 | 2070 | ||
2068 | static int compat_nfs_exp_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg) | 2071 | static int compat_nfs_exp_trans(struct nfsctl_arg *karg, |
2069 | { | 2072 | struct compat_nfsctl_arg __user *arg) |
2070 | int err; | 2073 | { |
2071 | 2074 | if (!access_ok(VERIFY_READ, &arg->ca32_export, | |
2072 | err = access_ok(VERIFY_READ, &arg->ca32_export, sizeof(arg->ca32_export)); | 2075 | sizeof(arg->ca32_export)) || |
2073 | err |= get_user(karg->ca_version, &arg->ca32_version); | 2076 | get_user(karg->ca_version, &arg->ca32_version) || |
2074 | err |= __copy_from_user(&karg->ca_export.ex_client[0], | 2077 | __copy_from_user(&karg->ca_export.ex_client[0], |
2075 | &arg->ca32_export.ex32_client[0], | 2078 | &arg->ca32_export.ex32_client[0], |
2076 | NFSCLNT_IDMAX); | 2079 | NFSCLNT_IDMAX) || |
2077 | err |= __copy_from_user(&karg->ca_export.ex_path[0], | 2080 | __copy_from_user(&karg->ca_export.ex_path[0], |
2078 | &arg->ca32_export.ex32_path[0], | 2081 | &arg->ca32_export.ex32_path[0], |
2079 | NFS_MAXPATHLEN); | 2082 | NFS_MAXPATHLEN) || |
2080 | err |= __get_user(karg->ca_export.ex_dev, | 2083 | __get_user(karg->ca_export.ex_dev, |
2081 | &arg->ca32_export.ex32_dev); | 2084 | &arg->ca32_export.ex32_dev) || |
2082 | err |= __get_user(karg->ca_export.ex_ino, | 2085 | __get_user(karg->ca_export.ex_ino, |
2083 | &arg->ca32_export.ex32_ino); | 2086 | &arg->ca32_export.ex32_ino) || |
2084 | err |= __get_user(karg->ca_export.ex_flags, | 2087 | __get_user(karg->ca_export.ex_flags, |
2085 | &arg->ca32_export.ex32_flags); | 2088 | &arg->ca32_export.ex32_flags) || |
2086 | err |= __get_user(karg->ca_export.ex_anon_uid, | 2089 | __get_user(karg->ca_export.ex_anon_uid, |
2087 | &arg->ca32_export.ex32_anon_uid); | 2090 | &arg->ca32_export.ex32_anon_uid) || |
2088 | err |= __get_user(karg->ca_export.ex_anon_gid, | 2091 | __get_user(karg->ca_export.ex_anon_gid, |
2089 | &arg->ca32_export.ex32_anon_gid); | 2092 | &arg->ca32_export.ex32_anon_gid)) |
2093 | return -EFAULT; | ||
2090 | SET_UID(karg->ca_export.ex_anon_uid, karg->ca_export.ex_anon_uid); | 2094 | SET_UID(karg->ca_export.ex_anon_uid, karg->ca_export.ex_anon_uid); |
2091 | SET_GID(karg->ca_export.ex_anon_gid, karg->ca_export.ex_anon_gid); | 2095 | SET_GID(karg->ca_export.ex_anon_gid, karg->ca_export.ex_anon_gid); |
2092 | 2096 | ||
2093 | return (err) ? -EFAULT : 0; | 2097 | return 0; |
2094 | } | 2098 | } |
2095 | 2099 | ||
2096 | static int compat_nfs_getfd_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg) | 2100 | static int compat_nfs_getfd_trans(struct nfsctl_arg *karg, |
2097 | { | 2101 | struct compat_nfsctl_arg __user *arg) |
2098 | int err; | 2102 | { |
2099 | 2103 | if (!access_ok(VERIFY_READ, &arg->ca32_getfd, | |
2100 | err = access_ok(VERIFY_READ, &arg->ca32_getfd, sizeof(arg->ca32_getfd)); | 2104 | sizeof(arg->ca32_getfd)) || |
2101 | err |= get_user(karg->ca_version, &arg->ca32_version); | 2105 | get_user(karg->ca_version, &arg->ca32_version) || |
2102 | err |= __copy_from_user(&karg->ca_getfd.gd_addr, | 2106 | __copy_from_user(&karg->ca_getfd.gd_addr, |
2103 | &arg->ca32_getfd.gd32_addr, | 2107 | &arg->ca32_getfd.gd32_addr, |
2104 | (sizeof(struct sockaddr))); | 2108 | (sizeof(struct sockaddr))) || |
2105 | err |= __copy_from_user(&karg->ca_getfd.gd_path, | 2109 | __copy_from_user(&karg->ca_getfd.gd_path, |
2106 | &arg->ca32_getfd.gd32_path, | 2110 | &arg->ca32_getfd.gd32_path, |
2107 | (NFS_MAXPATHLEN+1)); | 2111 | (NFS_MAXPATHLEN+1)) || |
2108 | err |= __get_user(karg->ca_getfd.gd_version, | 2112 | __get_user(karg->ca_getfd.gd_version, |
2109 | &arg->ca32_getfd.gd32_version); | 2113 | &arg->ca32_getfd.gd32_version)) |
2114 | return -EFAULT; | ||
2110 | 2115 | ||
2111 | return (err) ? -EFAULT : 0; | 2116 | return 0; |
2112 | } | 2117 | } |
2113 | 2118 | ||
2114 | static int compat_nfs_getfs_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg) | 2119 | static int compat_nfs_getfs_trans(struct nfsctl_arg *karg, |
2120 | struct compat_nfsctl_arg __user *arg) | ||
2115 | { | 2121 | { |
2116 | int err; | 2122 | if (!access_ok(VERIFY_READ,&arg->ca32_getfs,sizeof(arg->ca32_getfs)) || |
2117 | 2123 | get_user(karg->ca_version, &arg->ca32_version) || | |
2118 | err = access_ok(VERIFY_READ, &arg->ca32_getfs, sizeof(arg->ca32_getfs)); | 2124 | __copy_from_user(&karg->ca_getfs.gd_addr, |
2119 | err |= get_user(karg->ca_version, &arg->ca32_version); | 2125 | &arg->ca32_getfs.gd32_addr, |
2120 | err |= __copy_from_user(&karg->ca_getfs.gd_addr, | 2126 | (sizeof(struct sockaddr))) || |
2121 | &arg->ca32_getfs.gd32_addr, | 2127 | __copy_from_user(&karg->ca_getfs.gd_path, |
2122 | (sizeof(struct sockaddr))); | 2128 | &arg->ca32_getfs.gd32_path, |
2123 | err |= __copy_from_user(&karg->ca_getfs.gd_path, | 2129 | (NFS_MAXPATHLEN+1)) || |
2124 | &arg->ca32_getfs.gd32_path, | 2130 | __get_user(karg->ca_getfs.gd_maxlen, |
2125 | (NFS_MAXPATHLEN+1)); | 2131 | &arg->ca32_getfs.gd32_maxlen)) |
2126 | err |= __get_user(karg->ca_getfs.gd_maxlen, | 2132 | return -EFAULT; |
2127 | &arg->ca32_getfs.gd32_maxlen); | ||
2128 | 2133 | ||
2129 | return (err) ? -EFAULT : 0; | 2134 | return 0; |
2130 | } | 2135 | } |
2131 | 2136 | ||
2132 | /* This really doesn't need translations, we are only passing | 2137 | /* This really doesn't need translations, we are only passing |
2133 | * back a union which contains opaque nfs file handle data. | 2138 | * back a union which contains opaque nfs file handle data. |
2134 | */ | 2139 | */ |
2135 | static int compat_nfs_getfh_res_trans(union nfsctl_res *kres, union compat_nfsctl_res __user *res) | 2140 | static int compat_nfs_getfh_res_trans(union nfsctl_res *kres, |
2141 | union compat_nfsctl_res __user *res) | ||
2136 | { | 2142 | { |
2137 | int err; | 2143 | int err; |
2138 | 2144 | ||
@@ -2141,8 +2147,9 @@ static int compat_nfs_getfh_res_trans(union nfsctl_res *kres, union compat_nfsct | |||
2141 | return (err) ? -EFAULT : 0; | 2147 | return (err) ? -EFAULT : 0; |
2142 | } | 2148 | } |
2143 | 2149 | ||
2144 | asmlinkage long compat_sys_nfsservctl(int cmd, struct compat_nfsctl_arg __user *arg, | 2150 | asmlinkage long compat_sys_nfsservctl(int cmd, |
2145 | union compat_nfsctl_res __user *res) | 2151 | struct compat_nfsctl_arg __user *arg, |
2152 | union compat_nfsctl_res __user *res) | ||
2146 | { | 2153 | { |
2147 | struct nfsctl_arg *karg; | 2154 | struct nfsctl_arg *karg; |
2148 | union nfsctl_res *kres; | 2155 | union nfsctl_res *kres; |
diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c index 5638c8f9362f..5f952187fc53 100644 --- a/fs/configfs/dir.c +++ b/fs/configfs/dir.c | |||
@@ -505,13 +505,15 @@ static int populate_groups(struct config_group *group) | |||
505 | int i; | 505 | int i; |
506 | 506 | ||
507 | if (group->default_groups) { | 507 | if (group->default_groups) { |
508 | /* FYI, we're faking mkdir here | 508 | /* |
509 | * FYI, we're faking mkdir here | ||
509 | * I'm not sure we need this semaphore, as we're called | 510 | * I'm not sure we need this semaphore, as we're called |
510 | * from our parent's mkdir. That holds our parent's | 511 | * from our parent's mkdir. That holds our parent's |
511 | * i_mutex, so afaik lookup cannot continue through our | 512 | * i_mutex, so afaik lookup cannot continue through our |
512 | * parent to find us, let alone mess with our tree. | 513 | * parent to find us, let alone mess with our tree. |
513 | * That said, taking our i_mutex is closer to mkdir | 514 | * That said, taking our i_mutex is closer to mkdir |
514 | * emulation, and shouldn't hurt. */ | 515 | * emulation, and shouldn't hurt. |
516 | */ | ||
515 | mutex_lock(&dentry->d_inode->i_mutex); | 517 | mutex_lock(&dentry->d_inode->i_mutex); |
516 | 518 | ||
517 | for (i = 0; group->default_groups[i]; i++) { | 519 | for (i = 0; group->default_groups[i]; i++) { |
@@ -546,20 +548,34 @@ static void unlink_obj(struct config_item *item) | |||
546 | 548 | ||
547 | item->ci_group = NULL; | 549 | item->ci_group = NULL; |
548 | item->ci_parent = NULL; | 550 | item->ci_parent = NULL; |
551 | |||
552 | /* Drop the reference for ci_entry */ | ||
549 | config_item_put(item); | 553 | config_item_put(item); |
550 | 554 | ||
555 | /* Drop the reference for ci_parent */ | ||
551 | config_group_put(group); | 556 | config_group_put(group); |
552 | } | 557 | } |
553 | } | 558 | } |
554 | 559 | ||
555 | static void link_obj(struct config_item *parent_item, struct config_item *item) | 560 | static void link_obj(struct config_item *parent_item, struct config_item *item) |
556 | { | 561 | { |
557 | /* Parent seems redundant with group, but it makes certain | 562 | /* |
558 | * traversals much nicer. */ | 563 | * Parent seems redundant with group, but it makes certain |
564 | * traversals much nicer. | ||
565 | */ | ||
559 | item->ci_parent = parent_item; | 566 | item->ci_parent = parent_item; |
567 | |||
568 | /* | ||
569 | * We hold a reference on the parent for the child's ci_parent | ||
570 | * link. | ||
571 | */ | ||
560 | item->ci_group = config_group_get(to_config_group(parent_item)); | 572 | item->ci_group = config_group_get(to_config_group(parent_item)); |
561 | list_add_tail(&item->ci_entry, &item->ci_group->cg_children); | 573 | list_add_tail(&item->ci_entry, &item->ci_group->cg_children); |
562 | 574 | ||
575 | /* | ||
576 | * We hold a reference on the child for ci_entry on the parent's | ||
577 | * cg_children | ||
578 | */ | ||
563 | config_item_get(item); | 579 | config_item_get(item); |
564 | } | 580 | } |
565 | 581 | ||
@@ -684,6 +700,10 @@ static void client_drop_item(struct config_item *parent_item, | |||
684 | type = parent_item->ci_type; | 700 | type = parent_item->ci_type; |
685 | BUG_ON(!type); | 701 | BUG_ON(!type); |
686 | 702 | ||
703 | /* | ||
704 | * If ->drop_item() exists, it is responsible for the | ||
705 | * config_item_put(). | ||
706 | */ | ||
687 | if (type->ct_group_ops && type->ct_group_ops->drop_item) | 707 | if (type->ct_group_ops && type->ct_group_ops->drop_item) |
688 | type->ct_group_ops->drop_item(to_config_group(parent_item), | 708 | type->ct_group_ops->drop_item(to_config_group(parent_item), |
689 | item); | 709 | item); |
@@ -694,23 +714,28 @@ static void client_drop_item(struct config_item *parent_item, | |||
694 | 714 | ||
695 | static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) | 715 | static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
696 | { | 716 | { |
697 | int ret; | 717 | int ret, module_got = 0; |
698 | struct config_group *group; | 718 | struct config_group *group; |
699 | struct config_item *item; | 719 | struct config_item *item; |
700 | struct config_item *parent_item; | 720 | struct config_item *parent_item; |
701 | struct configfs_subsystem *subsys; | 721 | struct configfs_subsystem *subsys; |
702 | struct configfs_dirent *sd; | 722 | struct configfs_dirent *sd; |
703 | struct config_item_type *type; | 723 | struct config_item_type *type; |
704 | struct module *owner; | 724 | struct module *owner = NULL; |
705 | char *name; | 725 | char *name; |
706 | 726 | ||
707 | if (dentry->d_parent == configfs_sb->s_root) | 727 | if (dentry->d_parent == configfs_sb->s_root) { |
708 | return -EPERM; | 728 | ret = -EPERM; |
729 | goto out; | ||
730 | } | ||
709 | 731 | ||
710 | sd = dentry->d_parent->d_fsdata; | 732 | sd = dentry->d_parent->d_fsdata; |
711 | if (!(sd->s_type & CONFIGFS_USET_DIR)) | 733 | if (!(sd->s_type & CONFIGFS_USET_DIR)) { |
712 | return -EPERM; | 734 | ret = -EPERM; |
735 | goto out; | ||
736 | } | ||
713 | 737 | ||
738 | /* Get a working ref for the duration of this function */ | ||
714 | parent_item = configfs_get_config_item(dentry->d_parent); | 739 | parent_item = configfs_get_config_item(dentry->d_parent); |
715 | type = parent_item->ci_type; | 740 | type = parent_item->ci_type; |
716 | subsys = to_config_group(parent_item)->cg_subsys; | 741 | subsys = to_config_group(parent_item)->cg_subsys; |
@@ -719,15 +744,16 @@ static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) | |||
719 | if (!type || !type->ct_group_ops || | 744 | if (!type || !type->ct_group_ops || |
720 | (!type->ct_group_ops->make_group && | 745 | (!type->ct_group_ops->make_group && |
721 | !type->ct_group_ops->make_item)) { | 746 | !type->ct_group_ops->make_item)) { |
722 | config_item_put(parent_item); | 747 | ret = -EPERM; /* Lack-of-mkdir returns -EPERM */ |
723 | return -EPERM; /* What lack-of-mkdir returns */ | 748 | goto out_put; |
724 | } | 749 | } |
725 | 750 | ||
726 | name = kmalloc(dentry->d_name.len + 1, GFP_KERNEL); | 751 | name = kmalloc(dentry->d_name.len + 1, GFP_KERNEL); |
727 | if (!name) { | 752 | if (!name) { |
728 | config_item_put(parent_item); | 753 | ret = -ENOMEM; |
729 | return -ENOMEM; | 754 | goto out_put; |
730 | } | 755 | } |
756 | |||
731 | snprintf(name, dentry->d_name.len + 1, "%s", dentry->d_name.name); | 757 | snprintf(name, dentry->d_name.len + 1, "%s", dentry->d_name.name); |
732 | 758 | ||
733 | down(&subsys->su_sem); | 759 | down(&subsys->su_sem); |
@@ -748,40 +774,67 @@ static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) | |||
748 | 774 | ||
749 | kfree(name); | 775 | kfree(name); |
750 | if (!item) { | 776 | if (!item) { |
751 | config_item_put(parent_item); | 777 | /* |
752 | return -ENOMEM; | 778 | * If item == NULL, then link_obj() was never called. |
779 | * There are no extra references to clean up. | ||
780 | */ | ||
781 | ret = -ENOMEM; | ||
782 | goto out_put; | ||
753 | } | 783 | } |
754 | 784 | ||
755 | ret = -EINVAL; | 785 | /* |
786 | * link_obj() has been called (via link_group() for groups). | ||
787 | * From here on out, errors must clean that up. | ||
788 | */ | ||
789 | |||
756 | type = item->ci_type; | 790 | type = item->ci_type; |
757 | if (type) { | 791 | if (!type) { |
758 | owner = type->ct_owner; | 792 | ret = -EINVAL; |
759 | if (try_module_get(owner)) { | 793 | goto out_unlink; |
760 | if (group) { | 794 | } |
761 | ret = configfs_attach_group(parent_item, | ||
762 | item, | ||
763 | dentry); | ||
764 | } else { | ||
765 | ret = configfs_attach_item(parent_item, | ||
766 | item, | ||
767 | dentry); | ||
768 | } | ||
769 | 795 | ||
770 | if (ret) { | 796 | owner = type->ct_owner; |
771 | down(&subsys->su_sem); | 797 | if (!try_module_get(owner)) { |
772 | if (group) | 798 | ret = -EINVAL; |
773 | unlink_group(group); | 799 | goto out_unlink; |
774 | else | 800 | } |
775 | unlink_obj(item); | ||
776 | client_drop_item(parent_item, item); | ||
777 | up(&subsys->su_sem); | ||
778 | 801 | ||
779 | config_item_put(parent_item); | 802 | /* |
780 | module_put(owner); | 803 | * I hate doing it this way, but if there is |
781 | } | 804 | * an error, module_put() probably should |
782 | } | 805 | * happen after any cleanup. |
806 | */ | ||
807 | module_got = 1; | ||
808 | |||
809 | if (group) | ||
810 | ret = configfs_attach_group(parent_item, item, dentry); | ||
811 | else | ||
812 | ret = configfs_attach_item(parent_item, item, dentry); | ||
813 | |||
814 | out_unlink: | ||
815 | if (ret) { | ||
816 | /* Tear down everything we built up */ | ||
817 | down(&subsys->su_sem); | ||
818 | if (group) | ||
819 | unlink_group(group); | ||
820 | else | ||
821 | unlink_obj(item); | ||
822 | client_drop_item(parent_item, item); | ||
823 | up(&subsys->su_sem); | ||
824 | |||
825 | if (module_got) | ||
826 | module_put(owner); | ||
783 | } | 827 | } |
784 | 828 | ||
829 | out_put: | ||
830 | /* | ||
831 | * link_obj()/link_group() took a reference from child->parent, | ||
832 | * so the parent is safely pinned. We can drop our working | ||
833 | * reference. | ||
834 | */ | ||
835 | config_item_put(parent_item); | ||
836 | |||
837 | out: | ||
785 | return ret; | 838 | return ret; |
786 | } | 839 | } |
787 | 840 | ||
@@ -801,6 +854,7 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry) | |||
801 | if (sd->s_type & CONFIGFS_USET_DEFAULT) | 854 | if (sd->s_type & CONFIGFS_USET_DEFAULT) |
802 | return -EPERM; | 855 | return -EPERM; |
803 | 856 | ||
857 | /* Get a working ref until we have the child */ | ||
804 | parent_item = configfs_get_config_item(dentry->d_parent); | 858 | parent_item = configfs_get_config_item(dentry->d_parent); |
805 | subsys = to_config_group(parent_item)->cg_subsys; | 859 | subsys = to_config_group(parent_item)->cg_subsys; |
806 | BUG_ON(!subsys); | 860 | BUG_ON(!subsys); |
@@ -817,6 +871,7 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry) | |||
817 | return ret; | 871 | return ret; |
818 | } | 872 | } |
819 | 873 | ||
874 | /* Get a working ref for the duration of this function */ | ||
820 | item = configfs_get_config_item(dentry); | 875 | item = configfs_get_config_item(dentry); |
821 | 876 | ||
822 | /* Drop reference from above, item already holds one. */ | 877 | /* Drop reference from above, item already holds one. */ |
diff --git a/fs/exportfs/expfs.c b/fs/exportfs/expfs.c index b06b54f1bbbb..4c39009350f3 100644 --- a/fs/exportfs/expfs.c +++ b/fs/exportfs/expfs.c | |||
@@ -102,7 +102,7 @@ find_exported_dentry(struct super_block *sb, void *obj, void *parent, | |||
102 | if (acceptable(context, result)) | 102 | if (acceptable(context, result)) |
103 | return result; | 103 | return result; |
104 | if (S_ISDIR(result->d_inode->i_mode)) { | 104 | if (S_ISDIR(result->d_inode->i_mode)) { |
105 | /* there is no other dentry, so fail */ | 105 | err = -EACCES; |
106 | goto err_result; | 106 | goto err_result; |
107 | } | 107 | } |
108 | 108 | ||
diff --git a/fs/inotify.c b/fs/inotify.c index 1f50302849c5..732ec4bd5774 100644 --- a/fs/inotify.c +++ b/fs/inotify.c | |||
@@ -848,7 +848,11 @@ static int inotify_release(struct inode *ignored, struct file *file) | |||
848 | inode = watch->inode; | 848 | inode = watch->inode; |
849 | mutex_lock(&inode->inotify_mutex); | 849 | mutex_lock(&inode->inotify_mutex); |
850 | mutex_lock(&dev->mutex); | 850 | mutex_lock(&dev->mutex); |
851 | remove_watch_no_event(watch, dev); | 851 | |
852 | /* make sure we didn't race with another list removal */ | ||
853 | if (likely(idr_find(&dev->idr, watch->wd))) | ||
854 | remove_watch_no_event(watch, dev); | ||
855 | |||
852 | mutex_unlock(&dev->mutex); | 856 | mutex_unlock(&dev->mutex); |
853 | mutex_unlock(&inode->inotify_mutex); | 857 | mutex_unlock(&inode->inotify_mutex); |
854 | put_inotify_watch(watch); | 858 | put_inotify_watch(watch); |
@@ -890,8 +894,7 @@ static int inotify_ignore(struct inotify_device *dev, s32 wd) | |||
890 | mutex_lock(&dev->mutex); | 894 | mutex_lock(&dev->mutex); |
891 | 895 | ||
892 | /* make sure that we did not race */ | 896 | /* make sure that we did not race */ |
893 | watch = idr_find(&dev->idr, wd); | 897 | if (likely(idr_find(&dev->idr, wd) == watch)) |
894 | if (likely(watch)) | ||
895 | remove_watch(watch, dev); | 898 | remove_watch(watch, dev); |
896 | 899 | ||
897 | mutex_unlock(&dev->mutex); | 900 | mutex_unlock(&dev->mutex); |
diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c index 6aa92d0e6876..1d65f13f458c 100644 --- a/fs/nfsd/vfs.c +++ b/fs/nfsd/vfs.c | |||
@@ -1922,11 +1922,10 @@ nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |||
1922 | value = kmalloc(size, GFP_KERNEL); | 1922 | value = kmalloc(size, GFP_KERNEL); |
1923 | if (!value) | 1923 | if (!value) |
1924 | return -ENOMEM; | 1924 | return -ENOMEM; |
1925 | size = posix_acl_to_xattr(acl, value, size); | 1925 | error = posix_acl_to_xattr(acl, value, size); |
1926 | if (size < 0) { | 1926 | if (error < 0) |
1927 | error = size; | ||
1928 | goto getout; | 1927 | goto getout; |
1929 | } | 1928 | size = error; |
1930 | } else | 1929 | } else |
1931 | size = 0; | 1930 | size = 0; |
1932 | 1931 | ||
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 0d858d0b25be..47152bf9a7f2 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c | |||
@@ -276,13 +276,29 @@ static int ocfs2_writepage(struct page *page, struct writeback_control *wbc) | |||
276 | return ret; | 276 | return ret; |
277 | } | 277 | } |
278 | 278 | ||
279 | /* This can also be called from ocfs2_write_zero_page() which has done | ||
280 | * it's own cluster locking. */ | ||
281 | int ocfs2_prepare_write_nolock(struct inode *inode, struct page *page, | ||
282 | unsigned from, unsigned to) | ||
283 | { | ||
284 | int ret; | ||
285 | |||
286 | down_read(&OCFS2_I(inode)->ip_alloc_sem); | ||
287 | |||
288 | ret = block_prepare_write(page, from, to, ocfs2_get_block); | ||
289 | |||
290 | up_read(&OCFS2_I(inode)->ip_alloc_sem); | ||
291 | |||
292 | return ret; | ||
293 | } | ||
294 | |||
279 | /* | 295 | /* |
280 | * ocfs2_prepare_write() can be an outer-most ocfs2 call when it is called | 296 | * ocfs2_prepare_write() can be an outer-most ocfs2 call when it is called |
281 | * from loopback. It must be able to perform its own locking around | 297 | * from loopback. It must be able to perform its own locking around |
282 | * ocfs2_get_block(). | 298 | * ocfs2_get_block(). |
283 | */ | 299 | */ |
284 | int ocfs2_prepare_write(struct file *file, struct page *page, | 300 | static int ocfs2_prepare_write(struct file *file, struct page *page, |
285 | unsigned from, unsigned to) | 301 | unsigned from, unsigned to) |
286 | { | 302 | { |
287 | struct inode *inode = page->mapping->host; | 303 | struct inode *inode = page->mapping->host; |
288 | int ret; | 304 | int ret; |
@@ -295,11 +311,7 @@ int ocfs2_prepare_write(struct file *file, struct page *page, | |||
295 | goto out; | 311 | goto out; |
296 | } | 312 | } |
297 | 313 | ||
298 | down_read(&OCFS2_I(inode)->ip_alloc_sem); | 314 | ret = ocfs2_prepare_write_nolock(inode, page, from, to); |
299 | |||
300 | ret = block_prepare_write(page, from, to, ocfs2_get_block); | ||
301 | |||
302 | up_read(&OCFS2_I(inode)->ip_alloc_sem); | ||
303 | 315 | ||
304 | ocfs2_meta_unlock(inode, 0); | 316 | ocfs2_meta_unlock(inode, 0); |
305 | out: | 317 | out: |
@@ -625,11 +637,31 @@ static ssize_t ocfs2_direct_IO(int rw, | |||
625 | int ret; | 637 | int ret; |
626 | 638 | ||
627 | mlog_entry_void(); | 639 | mlog_entry_void(); |
640 | |||
641 | /* | ||
642 | * We get PR data locks even for O_DIRECT. This allows | ||
643 | * concurrent O_DIRECT I/O but doesn't let O_DIRECT with | ||
644 | * extending and buffered zeroing writes race. If they did | ||
645 | * race then the buffered zeroing could be written back after | ||
646 | * the O_DIRECT I/O. It's one thing to tell people not to mix | ||
647 | * buffered and O_DIRECT writes, but expecting them to | ||
648 | * understand that file extension is also an implicit buffered | ||
649 | * write is too much. By getting the PR we force writeback of | ||
650 | * the buffered zeroing before proceeding. | ||
651 | */ | ||
652 | ret = ocfs2_data_lock(inode, 0); | ||
653 | if (ret < 0) { | ||
654 | mlog_errno(ret); | ||
655 | goto out; | ||
656 | } | ||
657 | ocfs2_data_unlock(inode, 0); | ||
658 | |||
628 | ret = blockdev_direct_IO_no_locking(rw, iocb, inode, | 659 | ret = blockdev_direct_IO_no_locking(rw, iocb, inode, |
629 | inode->i_sb->s_bdev, iov, offset, | 660 | inode->i_sb->s_bdev, iov, offset, |
630 | nr_segs, | 661 | nr_segs, |
631 | ocfs2_direct_IO_get_blocks, | 662 | ocfs2_direct_IO_get_blocks, |
632 | ocfs2_dio_end_io); | 663 | ocfs2_dio_end_io); |
664 | out: | ||
633 | mlog_exit(ret); | 665 | mlog_exit(ret); |
634 | return ret; | 666 | return ret; |
635 | } | 667 | } |
diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h index d40456d509a0..e88c3f0b8fa9 100644 --- a/fs/ocfs2/aops.h +++ b/fs/ocfs2/aops.h | |||
@@ -22,8 +22,8 @@ | |||
22 | #ifndef OCFS2_AOPS_H | 22 | #ifndef OCFS2_AOPS_H |
23 | #define OCFS2_AOPS_H | 23 | #define OCFS2_AOPS_H |
24 | 24 | ||
25 | int ocfs2_prepare_write(struct file *file, struct page *page, | 25 | int ocfs2_prepare_write_nolock(struct inode *inode, struct page *page, |
26 | unsigned from, unsigned to); | 26 | unsigned from, unsigned to); |
27 | 27 | ||
28 | struct ocfs2_journal_handle *ocfs2_start_walk_page_trans(struct inode *inode, | 28 | struct ocfs2_journal_handle *ocfs2_start_walk_page_trans(struct inode *inode, |
29 | struct page *page, | 29 | struct page *page, |
diff --git a/fs/ocfs2/extent_map.c b/fs/ocfs2/extent_map.c index 4601fc256f11..1a5c69071df6 100644 --- a/fs/ocfs2/extent_map.c +++ b/fs/ocfs2/extent_map.c | |||
@@ -569,7 +569,7 @@ static int ocfs2_extent_map_insert(struct inode *inode, | |||
569 | 569 | ||
570 | ret = -ENOMEM; | 570 | ret = -ENOMEM; |
571 | ctxt.new_ent = kmem_cache_alloc(ocfs2_em_ent_cachep, | 571 | ctxt.new_ent = kmem_cache_alloc(ocfs2_em_ent_cachep, |
572 | GFP_KERNEL); | 572 | GFP_NOFS); |
573 | if (!ctxt.new_ent) { | 573 | if (!ctxt.new_ent) { |
574 | mlog_errno(ret); | 574 | mlog_errno(ret); |
575 | return ret; | 575 | return ret; |
@@ -583,14 +583,14 @@ static int ocfs2_extent_map_insert(struct inode *inode, | |||
583 | if (ctxt.need_left && !ctxt.left_ent) { | 583 | if (ctxt.need_left && !ctxt.left_ent) { |
584 | ctxt.left_ent = | 584 | ctxt.left_ent = |
585 | kmem_cache_alloc(ocfs2_em_ent_cachep, | 585 | kmem_cache_alloc(ocfs2_em_ent_cachep, |
586 | GFP_KERNEL); | 586 | GFP_NOFS); |
587 | if (!ctxt.left_ent) | 587 | if (!ctxt.left_ent) |
588 | break; | 588 | break; |
589 | } | 589 | } |
590 | if (ctxt.need_right && !ctxt.right_ent) { | 590 | if (ctxt.need_right && !ctxt.right_ent) { |
591 | ctxt.right_ent = | 591 | ctxt.right_ent = |
592 | kmem_cache_alloc(ocfs2_em_ent_cachep, | 592 | kmem_cache_alloc(ocfs2_em_ent_cachep, |
593 | GFP_KERNEL); | 593 | GFP_NOFS); |
594 | if (!ctxt.right_ent) | 594 | if (!ctxt.right_ent) |
595 | break; | 595 | break; |
596 | } | 596 | } |
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index 581eb451a41a..a9559c874530 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c | |||
@@ -613,7 +613,8 @@ leave: | |||
613 | 613 | ||
614 | /* Some parts of this taken from generic_cont_expand, which turned out | 614 | /* Some parts of this taken from generic_cont_expand, which turned out |
615 | * to be too fragile to do exactly what we need without us having to | 615 | * to be too fragile to do exactly what we need without us having to |
616 | * worry about recursive locking in ->commit_write(). */ | 616 | * worry about recursive locking in ->prepare_write() and |
617 | * ->commit_write(). */ | ||
617 | static int ocfs2_write_zero_page(struct inode *inode, | 618 | static int ocfs2_write_zero_page(struct inode *inode, |
618 | u64 size) | 619 | u64 size) |
619 | { | 620 | { |
@@ -641,7 +642,7 @@ static int ocfs2_write_zero_page(struct inode *inode, | |||
641 | goto out; | 642 | goto out; |
642 | } | 643 | } |
643 | 644 | ||
644 | ret = ocfs2_prepare_write(NULL, page, offset, offset); | 645 | ret = ocfs2_prepare_write_nolock(inode, page, offset, offset); |
645 | if (ret < 0) { | 646 | if (ret < 0) { |
646 | mlog_errno(ret); | 647 | mlog_errno(ret); |
647 | goto out_unlock; | 648 | goto out_unlock; |
@@ -695,13 +696,26 @@ out: | |||
695 | return ret; | 696 | return ret; |
696 | } | 697 | } |
697 | 698 | ||
699 | /* | ||
700 | * A tail_to_skip value > 0 indicates that we're being called from | ||
701 | * ocfs2_file_aio_write(). This has the following implications: | ||
702 | * | ||
703 | * - we don't want to update i_size | ||
704 | * - di_bh will be NULL, which is fine because it's only used in the | ||
705 | * case where we want to update i_size. | ||
706 | * - ocfs2_zero_extend() will then only be filling the hole created | ||
707 | * between i_size and the start of the write. | ||
708 | */ | ||
698 | static int ocfs2_extend_file(struct inode *inode, | 709 | static int ocfs2_extend_file(struct inode *inode, |
699 | struct buffer_head *di_bh, | 710 | struct buffer_head *di_bh, |
700 | u64 new_i_size) | 711 | u64 new_i_size, |
712 | size_t tail_to_skip) | ||
701 | { | 713 | { |
702 | int ret = 0; | 714 | int ret = 0; |
703 | u32 clusters_to_add; | 715 | u32 clusters_to_add; |
704 | 716 | ||
717 | BUG_ON(!tail_to_skip && !di_bh); | ||
718 | |||
705 | /* setattr sometimes calls us like this. */ | 719 | /* setattr sometimes calls us like this. */ |
706 | if (new_i_size == 0) | 720 | if (new_i_size == 0) |
707 | goto out; | 721 | goto out; |
@@ -714,27 +728,44 @@ static int ocfs2_extend_file(struct inode *inode, | |||
714 | OCFS2_I(inode)->ip_clusters; | 728 | OCFS2_I(inode)->ip_clusters; |
715 | 729 | ||
716 | if (clusters_to_add) { | 730 | if (clusters_to_add) { |
717 | ret = ocfs2_extend_allocation(inode, clusters_to_add); | 731 | /* |
732 | * protect the pages that ocfs2_zero_extend is going to | ||
733 | * be pulling into the page cache.. we do this before the | ||
734 | * metadata extend so that we don't get into the situation | ||
735 | * where we've extended the metadata but can't get the data | ||
736 | * lock to zero. | ||
737 | */ | ||
738 | ret = ocfs2_data_lock(inode, 1); | ||
718 | if (ret < 0) { | 739 | if (ret < 0) { |
719 | mlog_errno(ret); | 740 | mlog_errno(ret); |
720 | goto out; | 741 | goto out; |
721 | } | 742 | } |
722 | 743 | ||
723 | ret = ocfs2_zero_extend(inode, new_i_size); | 744 | ret = ocfs2_extend_allocation(inode, clusters_to_add); |
724 | if (ret < 0) { | 745 | if (ret < 0) { |
725 | mlog_errno(ret); | 746 | mlog_errno(ret); |
726 | goto out; | 747 | goto out_unlock; |
727 | } | 748 | } |
728 | } | ||
729 | 749 | ||
730 | /* No allocation required, we just use this helper to | 750 | ret = ocfs2_zero_extend(inode, (u64)new_i_size - tail_to_skip); |
731 | * do a trivial update of i_size. */ | 751 | if (ret < 0) { |
732 | ret = ocfs2_simple_size_update(inode, di_bh, new_i_size); | 752 | mlog_errno(ret); |
733 | if (ret < 0) { | 753 | goto out_unlock; |
734 | mlog_errno(ret); | 754 | } |
735 | goto out; | 755 | } |
756 | |||
757 | if (!tail_to_skip) { | ||
758 | /* We're being called from ocfs2_setattr() which wants | ||
759 | * us to update i_size */ | ||
760 | ret = ocfs2_simple_size_update(inode, di_bh, new_i_size); | ||
761 | if (ret < 0) | ||
762 | mlog_errno(ret); | ||
736 | } | 763 | } |
737 | 764 | ||
765 | out_unlock: | ||
766 | if (clusters_to_add) /* this is the only case in which we lock */ | ||
767 | ocfs2_data_unlock(inode, 1); | ||
768 | |||
738 | out: | 769 | out: |
739 | return ret; | 770 | return ret; |
740 | } | 771 | } |
@@ -793,7 +824,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr) | |||
793 | if (i_size_read(inode) > attr->ia_size) | 824 | if (i_size_read(inode) > attr->ia_size) |
794 | status = ocfs2_truncate_file(inode, bh, attr->ia_size); | 825 | status = ocfs2_truncate_file(inode, bh, attr->ia_size); |
795 | else | 826 | else |
796 | status = ocfs2_extend_file(inode, bh, attr->ia_size); | 827 | status = ocfs2_extend_file(inode, bh, attr->ia_size, 0); |
797 | if (status < 0) { | 828 | if (status < 0) { |
798 | if (status != -ENOSPC) | 829 | if (status != -ENOSPC) |
799 | mlog_errno(status); | 830 | mlog_errno(status); |
@@ -1049,21 +1080,12 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb, | |||
1049 | if (!clusters) | 1080 | if (!clusters) |
1050 | break; | 1081 | break; |
1051 | 1082 | ||
1052 | ret = ocfs2_extend_allocation(inode, clusters); | 1083 | ret = ocfs2_extend_file(inode, NULL, newsize, count); |
1053 | if (ret < 0) { | 1084 | if (ret < 0) { |
1054 | if (ret != -ENOSPC) | 1085 | if (ret != -ENOSPC) |
1055 | mlog_errno(ret); | 1086 | mlog_errno(ret); |
1056 | goto out; | 1087 | goto out; |
1057 | } | 1088 | } |
1058 | |||
1059 | /* Fill any holes which would've been created by this | ||
1060 | * write. If we're O_APPEND, this will wind up | ||
1061 | * (correctly) being a noop. */ | ||
1062 | ret = ocfs2_zero_extend(inode, (u64) newsize - count); | ||
1063 | if (ret < 0) { | ||
1064 | mlog_errno(ret); | ||
1065 | goto out; | ||
1066 | } | ||
1067 | break; | 1089 | break; |
1068 | } | 1090 | } |
1069 | 1091 | ||
@@ -1146,6 +1168,22 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb, | |||
1146 | ocfs2_iocb_set_rw_locked(iocb); | 1168 | ocfs2_iocb_set_rw_locked(iocb); |
1147 | } | 1169 | } |
1148 | 1170 | ||
1171 | /* | ||
1172 | * We're fine letting folks race truncates and extending | ||
1173 | * writes with read across the cluster, just like they can | ||
1174 | * locally. Hence no rw_lock during read. | ||
1175 | * | ||
1176 | * Take and drop the meta data lock to update inode fields | ||
1177 | * like i_size. This allows the checks down below | ||
1178 | * generic_file_aio_read() a chance of actually working. | ||
1179 | */ | ||
1180 | ret = ocfs2_meta_lock(inode, NULL, NULL, 0); | ||
1181 | if (ret < 0) { | ||
1182 | mlog_errno(ret); | ||
1183 | goto bail; | ||
1184 | } | ||
1185 | ocfs2_meta_unlock(inode, 0); | ||
1186 | |||
1149 | ret = generic_file_aio_read(iocb, buf, count, iocb->ki_pos); | 1187 | ret = generic_file_aio_read(iocb, buf, count, iocb->ki_pos); |
1150 | if (ret == -EINVAL) | 1188 | if (ret == -EINVAL) |
1151 | mlog(ML_ERROR, "generic_file_aio_read returned -EINVAL\n"); | 1189 | mlog(ML_ERROR, "generic_file_aio_read returned -EINVAL\n"); |
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c index 6a610ae53583..eebc3cfa6be8 100644 --- a/fs/ocfs2/journal.c +++ b/fs/ocfs2/journal.c | |||
@@ -117,7 +117,7 @@ struct ocfs2_journal_handle *ocfs2_alloc_handle(struct ocfs2_super *osb) | |||
117 | { | 117 | { |
118 | struct ocfs2_journal_handle *retval = NULL; | 118 | struct ocfs2_journal_handle *retval = NULL; |
119 | 119 | ||
120 | retval = kcalloc(1, sizeof(*retval), GFP_KERNEL); | 120 | retval = kcalloc(1, sizeof(*retval), GFP_NOFS); |
121 | if (!retval) { | 121 | if (!retval) { |
122 | mlog(ML_ERROR, "Failed to allocate memory for journal " | 122 | mlog(ML_ERROR, "Failed to allocate memory for journal " |
123 | "handle!\n"); | 123 | "handle!\n"); |
@@ -870,9 +870,11 @@ static int ocfs2_force_read_journal(struct inode *inode) | |||
870 | if (p_blocks > CONCURRENT_JOURNAL_FILL) | 870 | if (p_blocks > CONCURRENT_JOURNAL_FILL) |
871 | p_blocks = CONCURRENT_JOURNAL_FILL; | 871 | p_blocks = CONCURRENT_JOURNAL_FILL; |
872 | 872 | ||
873 | /* We are reading journal data which should not | ||
874 | * be put in the uptodate cache */ | ||
873 | status = ocfs2_read_blocks(OCFS2_SB(inode->i_sb), | 875 | status = ocfs2_read_blocks(OCFS2_SB(inode->i_sb), |
874 | p_blkno, p_blocks, bhs, 0, | 876 | p_blkno, p_blocks, bhs, 0, |
875 | inode); | 877 | NULL); |
876 | if (status < 0) { | 878 | if (status < 0) { |
877 | mlog_errno(status); | 879 | mlog_errno(status); |
878 | goto bail; | 880 | goto bail; |
@@ -982,7 +984,7 @@ static void ocfs2_queue_recovery_completion(struct ocfs2_journal *journal, | |||
982 | { | 984 | { |
983 | struct ocfs2_la_recovery_item *item; | 985 | struct ocfs2_la_recovery_item *item; |
984 | 986 | ||
985 | item = kmalloc(sizeof(struct ocfs2_la_recovery_item), GFP_KERNEL); | 987 | item = kmalloc(sizeof(struct ocfs2_la_recovery_item), GFP_NOFS); |
986 | if (!item) { | 988 | if (!item) { |
987 | /* Though we wish to avoid it, we are in fact safe in | 989 | /* Though we wish to avoid it, we are in fact safe in |
988 | * skipping local alloc cleanup as fsck.ocfs2 is more | 990 | * skipping local alloc cleanup as fsck.ocfs2 is more |
diff --git a/fs/ocfs2/uptodate.c b/fs/ocfs2/uptodate.c index 04a684dfdd96..b8a00a793326 100644 --- a/fs/ocfs2/uptodate.c +++ b/fs/ocfs2/uptodate.c | |||
@@ -337,7 +337,7 @@ static void __ocfs2_set_buffer_uptodate(struct ocfs2_inode_info *oi, | |||
337 | (unsigned long long)oi->ip_blkno, | 337 | (unsigned long long)oi->ip_blkno, |
338 | (unsigned long long)block, expand_tree); | 338 | (unsigned long long)block, expand_tree); |
339 | 339 | ||
340 | new = kmem_cache_alloc(ocfs2_uptodate_cachep, GFP_KERNEL); | 340 | new = kmem_cache_alloc(ocfs2_uptodate_cachep, GFP_NOFS); |
341 | if (!new) { | 341 | if (!new) { |
342 | mlog_errno(-ENOMEM); | 342 | mlog_errno(-ENOMEM); |
343 | return; | 343 | return; |
@@ -349,7 +349,7 @@ static void __ocfs2_set_buffer_uptodate(struct ocfs2_inode_info *oi, | |||
349 | * has no way of tracking that. */ | 349 | * has no way of tracking that. */ |
350 | for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) { | 350 | for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) { |
351 | tree[i] = kmem_cache_alloc(ocfs2_uptodate_cachep, | 351 | tree[i] = kmem_cache_alloc(ocfs2_uptodate_cachep, |
352 | GFP_KERNEL); | 352 | GFP_NOFS); |
353 | if (!tree[i]) { | 353 | if (!tree[i]) { |
354 | mlog_errno(-ENOMEM); | 354 | mlog_errno(-ENOMEM); |
355 | goto out_free; | 355 | goto out_free; |
diff --git a/fs/ocfs2/vote.c b/fs/ocfs2/vote.c index 53049a204197..ee42765a8553 100644 --- a/fs/ocfs2/vote.c +++ b/fs/ocfs2/vote.c | |||
@@ -586,7 +586,7 @@ static struct ocfs2_net_wait_ctxt *ocfs2_new_net_wait_ctxt(unsigned int response | |||
586 | { | 586 | { |
587 | struct ocfs2_net_wait_ctxt *w; | 587 | struct ocfs2_net_wait_ctxt *w; |
588 | 588 | ||
589 | w = kcalloc(1, sizeof(*w), GFP_KERNEL); | 589 | w = kcalloc(1, sizeof(*w), GFP_NOFS); |
590 | if (!w) { | 590 | if (!w) { |
591 | mlog_errno(-ENOMEM); | 591 | mlog_errno(-ENOMEM); |
592 | goto bail; | 592 | goto bail; |
@@ -749,7 +749,7 @@ static struct ocfs2_vote_msg * ocfs2_new_vote_request(struct ocfs2_super *osb, | |||
749 | 749 | ||
750 | BUG_ON(!ocfs2_is_valid_vote_request(type)); | 750 | BUG_ON(!ocfs2_is_valid_vote_request(type)); |
751 | 751 | ||
752 | request = kcalloc(1, sizeof(*request), GFP_KERNEL); | 752 | request = kcalloc(1, sizeof(*request), GFP_NOFS); |
753 | if (!request) { | 753 | if (!request) { |
754 | mlog_errno(-ENOMEM); | 754 | mlog_errno(-ENOMEM); |
755 | } else { | 755 | } else { |
@@ -1129,7 +1129,7 @@ static int ocfs2_handle_vote_message(struct o2net_msg *msg, | |||
1129 | struct ocfs2_super *osb = data; | 1129 | struct ocfs2_super *osb = data; |
1130 | struct ocfs2_vote_work *work; | 1130 | struct ocfs2_vote_work *work; |
1131 | 1131 | ||
1132 | work = kmalloc(sizeof(struct ocfs2_vote_work), GFP_KERNEL); | 1132 | work = kmalloc(sizeof(struct ocfs2_vote_work), GFP_NOFS); |
1133 | if (!work) { | 1133 | if (!work) { |
1134 | status = -ENOMEM; | 1134 | status = -ENOMEM; |
1135 | mlog_errno(status); | 1135 | mlog_errno(status); |