From patchwork Fri Apr 22 17:09:36 2011 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Shirish Pargaonkar X-Patchwork-Id: 727661 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by demeter1.kernel.org (8.14.4/8.14.3) with ESMTP id p3MH6MC3022494 for ; Fri, 22 Apr 2011 17:06:22 GMT Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756216Ab1DVRGW (ORCPT ); Fri, 22 Apr 2011 13:06:22 -0400 Received: from mail-gy0-f174.google.com ([209.85.160.174]:60326 "EHLO mail-gy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756052Ab1DVRGV (ORCPT ); Fri, 22 Apr 2011 13:06:21 -0400 Received: by gyd10 with SMTP id 10so190633gyd.19 for ; Fri, 22 Apr 2011 10:06:20 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:from:to:cc:subject:date:message-id:x-mailer; bh=Q9UpgyE0xkBPyHFirt3kfJGKWoHS1F7XuL9NsWAPPQo=; b=Tz44L+fGXXPAkot5KekxITJ6QEzIDvgYkn0v191BPRGaqXz8vZkJB+lLwolBu7xvh6 VTRsA3d4xg2gghMFNME5ijxw6JTJuWjOQp2T5/nW8THnnw5Fry9wE34/b1W348EPQnto qBVfaOWuyRnEQDPem+3bkaJNeOm3YnseIhWqA= DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=from:to:cc:subject:date:message-id:x-mailer; b=J2zgsNbnYnina0oyIt4TfFlTE8VRv+4TtSx2i9TkQ34uFEhyWSvcndbH0Ey4OfYQnF vWkgjHV1+0Gd5hFXeB8yhV2dWuA5q1C6yBGg3Z9RDYbkbv/K70tYPpDdmCcZjnPuwSA2 RSwlCehMnAS/vRP1K1TXuATi2KiwOKgWw1ZxQ= Received: by 10.90.16.34 with SMTP id 34mr1451216agp.28.1303491980676; Fri, 22 Apr 2011 10:06:20 -0700 (PDT) Received: from localhost ([32.97.110.58]) by mx.google.com with ESMTPS id d20sm2958056ang.17.2011.04.22.10.06.18 (version=TLSv1/SSLv3 cipher=OTHER); Fri, 22 Apr 2011 10:06:19 -0700 (PDT) From: shirishpargaonkar@gmail.com To: smfrench@gmail.com Cc: linux-cifs@vger.kernel.org, Shirish Pargaonkar Subject: [PATCH 2/2] cifs: Invoke id mapping functions (try #17 repost) Date: Fri, 22 Apr 2011 12:09:36 -0500 Message-Id: <1303492176-11679-1-git-send-email-shirishpargaonkar@gmail.com> X-Mailer: git-send-email 1.6.0.2 Sender: linux-cifs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-cifs@vger.kernel.org X-Greylist: IP, sender and recipient auto-whitelisted, not delayed by milter-greylist-4.2.6 (demeter1.kernel.org [140.211.167.41]); Fri, 22 Apr 2011 17:06:22 +0000 (UTC) From: Shirish Pargaonkar rb tree search and insertion routines. A SID which needs to be mapped, is looked up in one of the rb trees depending on whether SID is either owner or group SID. If found in the tree, a (mapped) id from that node is assigned to uid or gid as appropriate. If unmapped, an upcall is attempted to map the SID to an id. If upcall is successful, node is marked as mapped. If upcall fails, node stays marked as unmapped and a mapping is attempted again only after an arbitrary time period has passed. To map a SID, which can be either a Owner SID or a Group SID, key description starts with the string "os" or "gs" followed by SID converted to a string. Without "os" or "gs", cifs.upcall does not know whether SID needs to be mapped to either an uid or a gid. Nodes in rb tree have fields to prevent multiple upcalls for a SID. Searching, adding, and removing nodes is done within global locks. Whenever a node is either found or inserted in a tree, a reference is taken on that node. Shrinker routine prunes a node if it has expired but does not prune an expired node if its refcount is not zero (i.e. sid/id of that node is_being/will_be accessed). Thus a node, if its SID needs to be mapped by making an upcall, can safely stay and its fields accessed without shrinker pruning it. A reference (refcount) is put on the node without holding the spinlock but a reference is get on the node by holding the spinlock. Every time an existing mapped node is accessed or mapping is attempted, its timestamp is updated to prevent it from getting erased or a to prevent multiple unnecessary repeat mapping retries respectively. For now, cifs.upcall is only used to map a SID to an id (uid or gid) but it would be used to obtain an SID for an id. Signed-off-by: Shirish Pargaonkar --- fs/cifs/cifsacl.c | 325 +++++++++++++++++++++++++++++++++++++++++++++++++---- fs/cifs/cifsacl.h | 24 ++++ 2 files changed, 325 insertions(+), 24 deletions(-) diff --git a/fs/cifs/cifsacl.c b/fs/cifs/cifsacl.c index 061fc3a..a4aa0f0 100644 --- a/fs/cifs/cifsacl.c +++ b/fs/cifs/cifsacl.c @@ -54,7 +54,33 @@ static const struct cifs_sid sid_authusers = { /* group users */ static const struct cifs_sid sid_user = {1, 2 , {0, 0, 0, 0, 0, 5}, {} }; -static const struct cred *root_cred; +const struct cred *root_cred; + +static void +shrink_idmap_tree(struct rb_root *root, int nr_to_scan, int *nr_rem, + int *nr_del) +{ + struct rb_node *node; + struct rb_node *tmp; + struct cifs_sid_id *psidid; + + node = rb_first(root); + while (node) { + tmp = node; + node = rb_next(tmp); + psidid = rb_entry(tmp, struct cifs_sid_id, rbnode); + if (nr_to_scan == 0 || *nr_del == nr_to_scan) + ++(*nr_rem); + else { + if (time_after(jiffies, psidid->time + SID_MAP_EXPIRE) + && psidid->refcount == 0) { + rb_erase(tmp, root); + ++(*nr_del); + } else + ++(*nr_rem); + } + } +} /* * Run idmap cache shrinker. @@ -62,9 +88,21 @@ static const struct cred *root_cred; static int cifs_idmap_shrinker(struct shrinker *shrink, int nr_to_scan, gfp_t gfp_mask) { - /* Use a pruning scheme in a subsequent patch instead */ - cifs_destroy_idmaptrees(); - return 0; + int nr_del = 0; + int nr_rem = 0; + struct rb_root *root; + + root = &uidtree; + spin_lock(&siduidlock); + shrink_idmap_tree(root, nr_to_scan, &nr_rem, &nr_del); + spin_unlock(&siduidlock); + + root = &gidtree; + spin_lock(&sidgidlock); + shrink_idmap_tree(root, nr_to_scan, &nr_rem, &nr_del); + spin_unlock(&sidgidlock); + + return nr_rem; } static struct shrinker cifs_shrinker = { @@ -92,7 +130,6 @@ cifs_idmap_key_destroy(struct key *key) kfree(key->payload.data); } -static struct key_type cifs_idmap_key_type = { .name = "cifs.cifs_idmap", .instantiate = cifs_idmap_key_instantiate, @@ -101,6 +138,220 @@ struct key_type cifs_idmap_key_type = { .match = user_match, }; +static void +sid_to_str(struct cifs_sid *sidptr, char *sidstr) +{ + int i; + unsigned long saval; + char *strptr; + + strptr = sidstr; + + sprintf(strptr, "%s", "S"); + strptr = sidstr + strlen(sidstr); + + sprintf(strptr, "-%d", sidptr->revision); + strptr = sidstr + strlen(sidstr); + + for (i = 0; i < 6; ++i) { + if (sidptr->authority[i]) { + sprintf(strptr, "-%d", sidptr->authority[i]); + strptr = sidstr + strlen(sidstr); + } + } + + for (i = 0; i < sidptr->num_subauth; ++i) { + saval = le32_to_cpu(sidptr->sub_auth[i]); + sprintf(strptr, "-%ld", saval); + strptr = sidstr + strlen(sidstr); + } +} + +static void +id_rb_insert(struct rb_root *root, struct cifs_sid *sidptr, + struct cifs_sid_id **psidid, char *typestr) +{ + int rc; + char *strptr; + struct rb_node *node = root->rb_node; + struct rb_node *parent = NULL; + struct rb_node **linkto = &(root->rb_node); + struct cifs_sid_id *lsidid; + + while (node) { + lsidid = rb_entry(node, struct cifs_sid_id, rbnode); + parent = node; + rc = compare_sids(sidptr, &((lsidid)->sid)); + if (rc > 0) { + linkto = &(node->rb_left); + node = node->rb_left; + } else if (rc < 0) { + linkto = &(node->rb_right); + node = node->rb_right; + } + } + + memcpy(&(*psidid)->sid, sidptr, sizeof(struct cifs_sid)); + (*psidid)->time = jiffies - (SID_MAP_RETRY + 1); + (*psidid)->refcount = 0; + + sprintf((*psidid)->sidstr, "%s", typestr); + strptr = (*psidid)->sidstr + strlen((*psidid)->sidstr); + sid_to_str(&(*psidid)->sid, strptr); + + clear_bit(SID_ID_PENDING, &(*psidid)->state); + clear_bit(SID_ID_MAPPED, &(*psidid)->state); + + rb_link_node(&(*psidid)->rbnode, parent, linkto); + rb_insert_color(&(*psidid)->rbnode, root); +} + +static struct cifs_sid_id * +id_rb_search(struct rb_root *root, struct cifs_sid *sidptr) +{ + int rc; + struct rb_node *node = root->rb_node; + struct rb_node *parent = NULL; + struct rb_node **linkto = &(root->rb_node); + struct cifs_sid_id *lsidid; + + while (node) { + lsidid = rb_entry(node, struct cifs_sid_id, rbnode); + parent = node; + rc = compare_sids(sidptr, &((lsidid)->sid)); + if (rc > 0) { + linkto = &(node->rb_left); + node = node->rb_left; + } else if (rc < 0) { + linkto = &(node->rb_right); + node = node->rb_right; + } else /* node found */ + return lsidid; + } + + return NULL; +} + +static int +sidid_pending_wait(void *unused) +{ + schedule(); + return signal_pending(current) ? -ERESTARTSYS : 0; +} + +static int +sid_to_id(struct cifs_sb_info *cifs_sb, struct cifs_sid *psid, + struct cifs_fattr *fattr, uint sidtype) +{ + int rc; + unsigned long cid; + struct key *idkey; + const struct cred *saved_cred; + struct cifs_sid_id *psidid, *npsidid; + struct rb_root *cidtree; + spinlock_t *cidlock; + + if (sidtype == SIDOWNER) { + cid = cifs_sb->mnt_uid; /* default uid, in case upcall fails */ + cidlock = &siduidlock; + cidtree = &uidtree; + } else if (sidtype == SIDGROUP) { + cid = cifs_sb->mnt_gid; /* default gid, in case upcall fails */ + cidlock = &sidgidlock; + cidtree = &gidtree; + } else + return -ENOENT; + + spin_lock(cidlock); + psidid = id_rb_search(cidtree, psid); + + if (!psidid) { /* node does not exist, allocate one & attempt adding */ + spin_unlock(cidlock); + npsidid = kzalloc(sizeof(struct cifs_sid_id), GFP_KERNEL); + if (!npsidid) + return -ENOMEM; + + npsidid->sidstr = kmalloc(SIDLEN, GFP_KERNEL); + if (!npsidid->sidstr) { + kfree(npsidid); + return -ENOMEM; + } + + spin_lock(cidlock); + psidid = id_rb_search(cidtree, psid); + if (psidid) { /* node happened to get inserted meanwhile */ + ++psidid->refcount; + spin_unlock(cidlock); + kfree(npsidid->sidstr); + kfree(npsidid); + } else { + psidid = npsidid; + id_rb_insert(cidtree, psid, &psidid, + sidtype == SIDOWNER ? "os:" : "gs:"); + ++psidid->refcount; + spin_unlock(cidlock); + } + } else { + ++psidid->refcount; + spin_unlock(cidlock); + } + + /* + * If we are here, it is safe to access psidid and its fields + * since a reference was taken earlier while holding the spinlock. + * A reference on the node is put without holding the spinlock + * and it is OK to do so in this case, shrinker will not erase + * this node until all references are put and we do not access + * any fields of the node after a reference is put . + */ + if (test_bit(SID_ID_MAPPED, &psidid->state)) { + cid = psidid->id; + psidid->time = jiffies; /* update ts for accessing */ + goto sid_to_id_out; + } + + if (time_after(psidid->time + SID_MAP_RETRY, jiffies)) + goto sid_to_id_out; + + if (!test_and_set_bit(SID_ID_PENDING, &psidid->state)) { + saved_cred = override_creds(root_cred); + idkey = request_key(&cifs_idmap_key_type, psidid->sidstr, ""); + if (IS_ERR(idkey)) + cFYI(1, "%s: Can't map SID to an id", __func__); + else { + cid = *(unsigned long *)idkey->payload.value; + psidid->id = cid; + set_bit(SID_ID_MAPPED, &psidid->state); + key_put(idkey); + kfree(psidid->sidstr); + } + revert_creds(saved_cred); + psidid->time = jiffies; /* update ts for accessing */ + clear_bit(SID_ID_PENDING, &psidid->state); + wake_up_bit(&psidid->state, SID_ID_PENDING); + } else { + rc = wait_on_bit(&psidid->state, SID_ID_PENDING, + sidid_pending_wait, TASK_INTERRUPTIBLE); + if (rc) { + cFYI(1, "%s: sidid_pending_wait interrupted %d", + __func__, rc); + --psidid->refcount; /* decremented without spinlock */ + return rc; + } + if (test_bit(SID_ID_MAPPED, &psidid->state)) + cid = psidid->id; + } + +sid_to_id_out: + --psidid->refcount; /* decremented without spinlock */ + if (sidtype == SIDOWNER) + fattr->cf_uid = cid; + else + fattr->cf_gid = cid; + + return 0; +} + int init_cifs_idmap(void) { @@ -242,16 +493,24 @@ int compare_sids(const struct cifs_sid *ctsid, const struct cifs_sid *cwsid) int num_subauth, num_sat, num_saw; if ((!ctsid) || (!cwsid)) - return 0; + return 1; /* compare the revision */ - if (ctsid->revision != cwsid->revision) - return 0; + if (ctsid->revision != cwsid->revision) { + if (ctsid->revision > cwsid->revision) + return 1; + else + return -1; + } /* compare all of the six auth values */ for (i = 0; i < 6; ++i) { - if (ctsid->authority[i] != cwsid->authority[i]) - return 0; + if (ctsid->authority[i] != cwsid->authority[i]) { + if (ctsid->authority[i] > cwsid->authority[i]) + return 1; + else + return -1; + } } /* compare all of the subauth values if any */ @@ -260,12 +519,16 @@ int compare_sids(const struct cifs_sid *ctsid, const struct cifs_sid *cwsid) num_subauth = num_sat < num_saw ? num_sat : num_saw; if (num_subauth) { for (i = 0; i < num_subauth; ++i) { - if (ctsid->sub_auth[i] != cwsid->sub_auth[i]) - return 0; + if (ctsid->sub_auth[i] != cwsid->sub_auth[i]) { + if (ctsid->sub_auth[i] > cwsid->sub_auth[i]) + return 1; + else + return -1; + } } } - return 1; /* sids compare/match */ + return 0; /* sids compare/match */ } @@ -520,22 +783,22 @@ static void parse_dacl(struct cifs_acl *pdacl, char *end_of_acl, #ifdef CONFIG_CIFS_DEBUG2 dump_ace(ppace[i], end_of_acl); #endif - if (compare_sids(&(ppace[i]->sid), pownersid)) + if (compare_sids(&(ppace[i]->sid), pownersid) == 0) access_flags_to_mode(ppace[i]->access_req, ppace[i]->type, &fattr->cf_mode, &user_mask); - if (compare_sids(&(ppace[i]->sid), pgrpsid)) + if (compare_sids(&(ppace[i]->sid), pgrpsid) == 0) access_flags_to_mode(ppace[i]->access_req, ppace[i]->type, &fattr->cf_mode, &group_mask); - if (compare_sids(&(ppace[i]->sid), &sid_everyone)) + if (compare_sids(&(ppace[i]->sid), &sid_everyone) == 0) access_flags_to_mode(ppace[i]->access_req, ppace[i]->type, &fattr->cf_mode, &other_mask); - if (compare_sids(&(ppace[i]->sid), &sid_authusers)) + if (compare_sids(&(ppace[i]->sid), &sid_authusers) == 0) access_flags_to_mode(ppace[i]->access_req, ppace[i]->type, &fattr->cf_mode, @@ -613,10 +876,10 @@ static int parse_sid(struct cifs_sid *psid, char *end_of_acl) /* Convert CIFS ACL to POSIX form */ -static int parse_sec_desc(struct cifs_ntsd *pntsd, int acl_len, - struct cifs_fattr *fattr) +static int parse_sec_desc(struct cifs_sb_info *cifs_sb, + struct cifs_ntsd *pntsd, int acl_len, struct cifs_fattr *fattr) { - int rc; + int rc = 0; struct cifs_sid *owner_sid_ptr, *group_sid_ptr; struct cifs_acl *dacl_ptr; /* no need for SACL ptr */ char *end_of_acl = ((char *)pntsd) + acl_len; @@ -638,12 +901,26 @@ static int parse_sec_desc(struct cifs_ntsd *pntsd, int acl_len, le32_to_cpu(pntsd->sacloffset), dacloffset); /* cifs_dump_mem("owner_sid: ", owner_sid_ptr, 64); */ rc = parse_sid(owner_sid_ptr, end_of_acl); - if (rc) + if (rc) { + cFYI(1, "%s: Error %d parsing Owner SID", __func__, rc); return rc; + } + rc = sid_to_id(cifs_sb, owner_sid_ptr, fattr, SIDOWNER); + if (rc) { + cFYI(1, "%s: Error %d mapping Owner SID to uid", __func__, rc); + return rc; + } rc = parse_sid(group_sid_ptr, end_of_acl); - if (rc) + if (rc) { + cFYI(1, "%s: Error %d mapping Owner SID to gid", __func__, rc); return rc; + } + rc = sid_to_id(cifs_sb, group_sid_ptr, fattr, SIDGROUP); + if (rc) { + cFYI(1, "%s: Error %d mapping Group SID to gid", __func__, rc); + return rc; + } if (dacloffset) parse_dacl(dacl_ptr, end_of_acl, owner_sid_ptr, @@ -658,7 +935,7 @@ static int parse_sec_desc(struct cifs_ntsd *pntsd, int acl_len, memcpy((void *)(&(cifscred->gsid)), (void *)group_sid_ptr, sizeof(struct cifs_sid)); */ - return 0; + return rc; } @@ -865,7 +1142,7 @@ cifs_acl_to_fattr(struct cifs_sb_info *cifs_sb, struct cifs_fattr *fattr, rc = PTR_ERR(pntsd); cERROR(1, "%s: error %d getting sec desc", __func__, rc); } else { - rc = parse_sec_desc(pntsd, acllen, fattr); + rc = parse_sec_desc(cifs_sb, pntsd, acllen, fattr); kfree(pntsd); if (rc) cERROR(1, "parse sec desc failed rc = %d", rc); diff --git a/fs/cifs/cifsacl.h b/fs/cifs/cifsacl.h index c4ae7d0..757cf5a 100644 --- a/fs/cifs/cifsacl.h +++ b/fs/cifs/cifsacl.h @@ -39,6 +39,15 @@ #define ACCESS_ALLOWED 0 #define ACCESS_DENIED 1 +#define SIDOWNER 1 +#define SIDGROUP 2 +#define SIDLEN 150 /* S- 1 revision- 6 authorities- max 5 sub authorities */ + +#define SID_ID_MAPPED 0 +#define SID_ID_PENDING 1 +#define SID_MAP_EXPIRE (3600 * HZ) /* map entry expires after one hour */ +#define SID_MAP_RETRY (300 * HZ) /* wait 5 minutes for next attempt to map */ + struct cifs_ntsd { __le16 revision; /* revision level */ __le16 type; @@ -74,6 +83,21 @@ struct cifs_wksid { char sidname[SIDNAMELENGTH]; } __attribute__((packed)); +struct cifs_sid_id { + unsigned int refcount; /* increment with spinlock, decrement without */ + unsigned long id; + unsigned long time; + unsigned long state; + char *sidstr; + struct rb_node rbnode; + struct cifs_sid sid; +}; + +#ifdef __KERNEL__ +extern struct key_type cifs_idmap_key_type; +extern const struct cred *root_cred; +#endif /* KERNEL */ + extern int match_sid(struct cifs_sid *); extern int compare_sids(const struct cifs_sid *, const struct cifs_sid *);