projects / pve-kernel.git / blame
| Commit | Line | Data |
|---|---|---|
| 47f3b899 SI |
1 | From 5506202b83e65b844309093e712b5b507eb1e403 Mon Sep 17 00:00:00 2001 |
| 2 | From: Jann Horn <jannh@google.com> | |
| 3 | Date: Tue, 13 Nov 2018 07:42:38 +0000 | |
| 4 | Subject: [PATCH 09/11] userns: also map extents in the reverse map to kernel | |
| 5 | IDs | |
| 6 | ||
| 7 | BugLink: https://launchpad.net/bugs/1801924 | |
| 8 | ||
| 9 | The current logic first clones the extent array and sorts both copies, then | |
| 10 | maps the lower IDs of the forward mapping into the lower namespace, but | |
| 11 | doesn't map the lower IDs of the reverse mapping. | |
| 12 | ||
| 13 | This means that code in a nested user namespace with >5 extents will see | |
| 14 | incorrect IDs. It also breaks some access checks, like | |
| 15 | inode_owner_or_capable() and privileged_wrt_inode_uidgid(), so a process | |
| 16 | can incorrectly appear to be capable relative to an inode. | |
| 17 | ||
| 18 | To fix it, we have to make sure that the "lower_first" members of extents | |
| 19 | in both arrays are translated; and we have to make sure that the reverse | |
| 20 | map is sorted *after* the translation (since otherwise the translation can | |
| 21 | break the sorting). | |
| 22 | ||
| 23 | This is CVE-2018-18955. | |
| 24 | ||
| 25 | Fixes: 6397fac4915a ("userns: bump idmap limits to 340") | |
| 26 | Cc: stable@vger.kernel.org | |
| 27 | Signed-off-by: Jann Horn <jannh@google.com> | |
| 28 | Tested-by: Eric W. Biederman <ebiederm@xmission.com> | |
| 29 | Reviewed-by: Eric W. Biederman <ebiederm@xmission.com> | |
| 30 | Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> | |
| 31 | ||
| 32 | CVE-2018-18955 | |
| 33 | ||
| 34 | (cherry picked from commit d2f007dbe7e4c9583eea6eb04d60001e85c6f1bd) | |
| 35 | Signed-off-by: Tyler Hicks <tyhicks@canonical.com> | |
| 36 | Acked-by: Colin King <colin.king@canonical.com> | |
| 37 | Acked-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com> | |
| 38 | Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com> | |
| 39 | --- | |
| 40 | kernel/user_namespace.c | 12 ++++++++---- | |
| 41 | 1 file changed, 8 insertions(+), 4 deletions(-) | |
| 42 | ||
| 43 | diff --git a/kernel/user_namespace.c b/kernel/user_namespace.c | |
| 44 | index 08d638386b83..12de8c144db9 100644 | |
| 45 | --- a/kernel/user_namespace.c | |
| 46 | +++ b/kernel/user_namespace.c | |
| 47 | @@ -983,10 +983,6 @@ static ssize_t map_write(struct file *file, const char __user *buf, | |
| 48 | if (!new_idmap_permitted(file, ns, cap_setid, &new_map)) | |
| 49 | goto out; | |
| 50 | ||
| 51 | - ret = sort_idmaps(&new_map); | |
| 52 | - if (ret < 0) | |
| 53 | - goto out; | |
| 54 | - | |
| 55 | ret = -EPERM; | |
| 56 | /* Map the lower ids from the parent user namespace to the | |
| 57 | * kernel global id space. | |
| 58 | @@ -1013,6 +1009,14 @@ static ssize_t map_write(struct file *file, const char __user *buf, | |
| 59 | e->lower_first = lower_first; | |
| 60 | } | |
| 61 | ||
| 62 | + /* | |
| 63 | + * If we want to use binary search for lookup, this clones the extent | |
| 64 | + * array and sorts both copies. | |
| 65 | + */ | |
| 66 | + ret = sort_idmaps(&new_map); | |
| 67 | + if (ret < 0) | |
| 68 | + goto out; | |
| 69 | + | |
| 70 | /* Install the map */ | |
| 71 | if (new_map.nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS) { | |
| 72 | memcpy(map->extent, new_map.extent, | |
| 73 | -- | |
| 74 | 2.11.0 | |
| 75 |