Linux-NFS Archive on lore.kernel.org
 help / color / Atom feed
From: Alexey Dobriyan <adobriyan@gmail.com>
To: akpm@linux-foundation.org
Cc: linux-kernel@vger.kernel.org, segoon@openwall.com,
	linux-nfs@vger.kernel.org
Subject: [PATCH] cred: simpler, 1D supplementary groups
Date: Wed, 17 Aug 2016 23:19:27 +0300
Message-ID: <20160817201927.GA2096@p183.telecom.by> (raw)

Current supplementary groups code can massively overallocate memory and
is implemented in a way so that access to individual gid is done
via 2D array.

If number of gids is <= 32, memory allocation is more or less tolerable
(140/148 bytes). But if it is not, code allocates full page (!) regardless
and, what's even more fun, doesn't reuse small 32-entry array.

2D array means dependent shifts, loads and LEAs without possibility
to optimize them (gid is never known at compile time).

All of the above is unnecessary. Switch to the usual trailing-zero-len-array
scheme. Memory is allocated with kmalloc/vmalloc() and only as much as needed.
Accesses become simpler (LEA 8(gi,idx,4) or even without displacement).

Maximum number of gids is 65536 which translates to 256KB+8 bytes.
I think kernel can handle such allocation.

On my usual desktop system with whole 9 (nine) aux groups,
struct group_info shrinks from 148 bytes to 44 bytes, yay!

Nice side effects:
* "gi->gid[i]" is shorter than "GROUP_AT(gi, i)", less typing,

* fix little mess in net/ipv4/ping.c
  should have been using GROUP_AT macro but this point becomes moot,

* aux group allocation is persistent and should be accounted as such.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
---

 arch/s390/kernel/compat_linux.c            |    4 -
 drivers/staging/lustre/lustre/ptlrpc/sec.c |    2 
 fs/nfsd/auth.c                             |    6 +-
 fs/nfsd/nfs4state.c                        |    2 
 fs/proc/array.c                            |    2 
 include/linux/cred.h                       |   11 ----
 kernel/groups.c                            |   67 +++++++++--------------------
 kernel/uid16.c                             |    4 -
 net/ipv4/ping.c                            |   15 ++----
 net/sunrpc/auth_generic.c                  |    4 -
 net/sunrpc/auth_gss/gss_rpc_xdr.c          |    2 
 net/sunrpc/auth_gss/svcauth_gss.c          |    2 
 net/sunrpc/auth_unix.c                     |    4 -
 net/sunrpc/svcauth_unix.c                  |    6 +-
 14 files changed, 46 insertions(+), 85 deletions(-)

--- a/arch/s390/kernel/compat_linux.c
+++ b/arch/s390/kernel/compat_linux.c
@@ -189,7 +189,7 @@ static int groups16_to_user(u16 __user *grouplist, struct group_info *group_info
 	kgid_t kgid;
 
 	for (i = 0; i < group_info->ngroups; i++) {
-		kgid = GROUP_AT(group_info, i);
+		kgid = group_info->gid[i];
 		group = (u16)from_kgid_munged(user_ns, kgid);
 		if (put_user(group, grouplist+i))
 			return -EFAULT;
@@ -213,7 +213,7 @@ static int groups16_from_user(struct group_info *group_info, u16 __user *groupli
 		if (!gid_valid(kgid))
 			return -EINVAL;
 
-		GROUP_AT(group_info, i) = kgid;
+		group_info->gid[i] = kgid;
 	}
 
 	return 0;
--- a/drivers/staging/lustre/lustre/ptlrpc/sec.c
+++ b/drivers/staging/lustre/lustre/ptlrpc/sec.c
@@ -2204,7 +2204,7 @@ int sptlrpc_pack_user_desc(struct lustre_msg *msg, int offset)
 	task_lock(current);
 	if (pud->pud_ngroups > current_ngroups)
 		pud->pud_ngroups = current_ngroups;
-	memcpy(pud->pud_groups, current_cred()->group_info->blocks[0],
+	memcpy(pud->pud_groups, current_cred()->group_info->gid,
 	       pud->pud_ngroups * sizeof(__u32));
 	task_unlock(current);
 
--- a/fs/nfsd/auth.c
+++ b/fs/nfsd/auth.c
@@ -55,10 +55,10 @@ int nfsd_setuser(struct svc_rqst *rqstp, struct svc_export *exp)
 			goto oom;
 
 		for (i = 0; i < rqgi->ngroups; i++) {
-			if (gid_eq(GLOBAL_ROOT_GID, GROUP_AT(rqgi, i)))
-				GROUP_AT(gi, i) = exp->ex_anon_gid;
+			if (gid_eq(GLOBAL_ROOT_GID, rqgi->gid[i]))
+				gi->gid[i] = exp->ex_anon_gid;
 			else
-				GROUP_AT(gi, i) = GROUP_AT(rqgi, i);
+				gi->gid[i] = rqgi->gid[i];
 		}
 	} else {
 		gi = get_group_info(rqgi);
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -1903,7 +1903,7 @@ static bool groups_equal(struct group_info *g1, struct group_info *g2)
 	if (g1->ngroups != g2->ngroups)
 		return false;
 	for (i=0; i<g1->ngroups; i++)
-		if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
+		if (!gid_eq(g1->gid[i], g2->gid[i]))
 			return false;
 	return true;
 }
--- a/fs/proc/array.c
+++ b/fs/proc/array.c
@@ -211,7 +211,7 @@ static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
 	group_info = cred->group_info;
 	for (g = 0; g < group_info->ngroups; g++)
 		seq_printf(m, "%d ",
-			   from_kgid_munged(user_ns, GROUP_AT(group_info, g)));
+			   from_kgid_munged(user_ns, group_info->gid[g]));
 	put_cred(cred);
 
 #ifdef CONFIG_PID_NS
--- a/include/linux/cred.h
+++ b/include/linux/cred.h
@@ -26,15 +26,10 @@ struct inode;
 /*
  * COW Supplementary groups list
  */
-#define NGROUPS_SMALL		32
-#define NGROUPS_PER_BLOCK	((unsigned int)(PAGE_SIZE / sizeof(kgid_t)))
-
 struct group_info {
 	atomic_t	usage;
 	int		ngroups;
-	int		nblocks;
-	kgid_t		small_block[NGROUPS_SMALL];
-	kgid_t		*blocks[0];
+	kgid_t		gid[0];
 };
 
 /**
@@ -88,10 +83,6 @@ extern void set_groups(struct cred *, struct group_info *);
 extern int groups_search(const struct group_info *, kgid_t);
 extern bool may_setgroups(void);
 
-/* access the groups "array" with this macro */
-#define GROUP_AT(gi, i) \
-	((gi)->blocks[(i) / NGROUPS_PER_BLOCK][(i) % NGROUPS_PER_BLOCK])
-
 /*
  * The security context of a task
  *
--- a/kernel/groups.c
+++ b/kernel/groups.c
@@ -7,55 +7,31 @@
 #include <linux/security.h>
 #include <linux/syscalls.h>
 #include <linux/user_namespace.h>
+#include <linux/vmalloc.h>
 #include <asm/uaccess.h>
 
 struct group_info *groups_alloc(int gidsetsize)
 {
-	struct group_info *group_info;
-	int nblocks;
-	int i;
-
-	nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
-	/* Make sure we always allocate at least one indirect block pointer */
-	nblocks = nblocks ? : 1;
-	group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
-	if (!group_info)
+	struct group_info *gi;
+	unsigned int len;
+
+	len = sizeof(struct group_info) + sizeof(kgid_t) * gidsetsize;
+	gi = kmalloc(len, GFP_KERNEL_ACCOUNT|__GFP_NOWARN|__GFP_NORETRY);
+	if (!gi)
+		gi = __vmalloc(len, GFP_KERNEL_ACCOUNT|__GFP_HIGHMEM, PAGE_KERNEL);
+	if (!gi)
 		return NULL;
-	group_info->ngroups = gidsetsize;
-	group_info->nblocks = nblocks;
-	atomic_set(&group_info->usage, 1);
-
-	if (gidsetsize <= NGROUPS_SMALL)
-		group_info->blocks[0] = group_info->small_block;
-	else {
-		for (i = 0; i < nblocks; i++) {
-			kgid_t *b;
-			b = (void *)__get_free_page(GFP_USER);
-			if (!b)
-				goto out_undo_partial_alloc;
-			group_info->blocks[i] = b;
-		}
-	}
-	return group_info;
 
-out_undo_partial_alloc:
-	while (--i >= 0) {
-		free_page((unsigned long)group_info->blocks[i]);
-	}
-	kfree(group_info);
-	return NULL;
+	atomic_set(&gi->usage, 1);
+	gi->ngroups = gidsetsize;
+	return gi;
 }
 
 EXPORT_SYMBOL(groups_alloc);
 
 void groups_free(struct group_info *group_info)
 {
-	if (group_info->blocks[0] != group_info->small_block) {
-		int i;
-		for (i = 0; i < group_info->nblocks; i++)
-			free_page((unsigned long)group_info->blocks[i]);
-	}
-	kfree(group_info);
+	kvfree(group_info);
 }
 
 EXPORT_SYMBOL(groups_free);
@@ -70,7 +46,7 @@ static int groups_to_user(gid_t __user *grouplist,
 
 	for (i = 0; i < count; i++) {
 		gid_t gid;
-		gid = from_kgid_munged(user_ns, GROUP_AT(group_info, i));
+		gid = from_kgid_munged(user_ns, group_info->gid[i]);
 		if (put_user(gid, grouplist+i))
 			return -EFAULT;
 	}
@@ -95,7 +71,7 @@ static int groups_from_user(struct group_info *group_info,
 		if (!gid_valid(kgid))
 			return -EINVAL;
 
-		GROUP_AT(group_info, i) = kgid;
+		group_info->gid[i] = kgid;
 	}
 	return 0;
 }
@@ -115,15 +91,14 @@ static void groups_sort(struct group_info *group_info)
 		for (base = 0; base < max; base++) {
 			int left = base;
 			int right = left + stride;
-			kgid_t tmp = GROUP_AT(group_info, right);
+			kgid_t tmp = group_info->gid[right];
 
-			while (left >= 0 && gid_gt(GROUP_AT(group_info, left), tmp)) {
-				GROUP_AT(group_info, right) =
-				    GROUP_AT(group_info, left);
+			while (left >= 0 && gid_gt(group_info->gid[left], tmp)) {
+				group_info->gid[right] = group_info->gid[left];
 				right = left;
 				left -= stride;
 			}
-			GROUP_AT(group_info, right) = tmp;
+			group_info->gid[right] = tmp;
 		}
 		stride /= 3;
 	}
@@ -141,9 +116,9 @@ int groups_search(const struct group_info *group_info, kgid_t grp)
 	right = group_info->ngroups;
 	while (left < right) {
 		unsigned int mid = (left+right)/2;
-		if (gid_gt(grp, GROUP_AT(group_info, mid)))
+		if (gid_gt(grp, group_info->gid[mid]))
 			left = mid + 1;
-		else if (gid_lt(grp, GROUP_AT(group_info, mid)))
+		else if (gid_lt(grp, group_info->gid[mid]))
 			right = mid;
 		else
 			return 1;
--- a/kernel/uid16.c
+++ b/kernel/uid16.c
@@ -117,7 +117,7 @@ static int groups16_to_user(old_gid_t __user *grouplist,
 	kgid_t kgid;
 
 	for (i = 0; i < group_info->ngroups; i++) {
-		kgid = GROUP_AT(group_info, i);
+		kgid = group_info->gid[i];
 		group = high2lowgid(from_kgid_munged(user_ns, kgid));
 		if (put_user(group, grouplist+i))
 			return -EFAULT;
@@ -142,7 +142,7 @@ static int groups16_from_user(struct group_info *group_info,
 		if (!gid_valid(kgid))
 			return -EINVAL;
 
-		GROUP_AT(group_info, i) = kgid;
+		group_info->gid[i] = kgid;
 	}
 
 	return 0;
--- a/net/ipv4/ping.c
+++ b/net/ipv4/ping.c
@@ -258,7 +258,7 @@ int ping_init_sock(struct sock *sk)
 	struct net *net = sock_net(sk);
 	kgid_t group = current_egid();
 	struct group_info *group_info;
-	int i, j, count;
+	int i;
 	kgid_t low, high;
 	int ret = 0;
 
@@ -270,16 +270,11 @@ int ping_init_sock(struct sock *sk)
 		return 0;
 
 	group_info = get_current_groups();
-	count = group_info->ngroups;
-	for (i = 0; i < group_info->nblocks; i++) {
-		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
-		for (j = 0; j < cp_count; j++) {
-			kgid_t gid = group_info->blocks[i][j];
-			if (gid_lte(low, gid) && gid_lte(gid, high))
-				goto out_release_group;
-		}
+	for (i = 0; i < group_info->ngroups; i++) {
+		kgid_t gid = group_info->gid[i];
 
-		count -= cp_count;
+		if (gid_lte(low, gid) && gid_lte(gid, high))
+			goto out_release_group;
 	}
 
 	ret = -EACCES;
--- a/net/sunrpc/auth_generic.c
+++ b/net/sunrpc/auth_generic.c
@@ -176,8 +176,8 @@ generic_match(struct auth_cred *acred, struct rpc_cred *cred, int flags)
 	if (gcred->acred.group_info->ngroups != acred->group_info->ngroups)
 		goto out_nomatch;
 	for (i = 0; i < gcred->acred.group_info->ngroups; i++) {
-		if (!gid_eq(GROUP_AT(gcred->acred.group_info, i),
-				GROUP_AT(acred->group_info, i)))
+		if (!gid_eq(gcred->acred.group_info->gid[i],
+				acred->group_info->gid[i]))
 			goto out_nomatch;
 	}
 out_match:
--- a/net/sunrpc/auth_gss/gss_rpc_xdr.c
+++ b/net/sunrpc/auth_gss/gss_rpc_xdr.c
@@ -229,7 +229,7 @@ static int gssx_dec_linux_creds(struct xdr_stream *xdr,
 		kgid = make_kgid(&init_user_ns, tmp);
 		if (!gid_valid(kgid))
 			goto out_free_groups;
-		GROUP_AT(creds->cr_group_info, i) = kgid;
+		creds->cr_group_info->gid[i] = kgid;
 	}
 
 	return 0;
--- a/net/sunrpc/auth_gss/svcauth_gss.c
+++ b/net/sunrpc/auth_gss/svcauth_gss.c
@@ -479,7 +479,7 @@ static int rsc_parse(struct cache_detail *cd,
 			kgid = make_kgid(&init_user_ns, id);
 			if (!gid_valid(kgid))
 				goto out;
-			GROUP_AT(rsci.cred.cr_group_info, i) = kgid;
+			rsci.cred.cr_group_info->gid[i] = kgid;
 		}
 
 		/* mech name */
--- a/net/sunrpc/auth_unix.c
+++ b/net/sunrpc/auth_unix.c
@@ -79,7 +79,7 @@ unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags, gfp_t
 
 	cred->uc_gid = acred->gid;
 	for (i = 0; i < groups; i++)
-		cred->uc_gids[i] = GROUP_AT(acred->group_info, i);
+		cred->uc_gids[i] = acred->group_info->gid[i];
 	if (i < NFS_NGROUPS)
 		cred->uc_gids[i] = INVALID_GID;
 
@@ -127,7 +127,7 @@ unx_match(struct auth_cred *acred, struct rpc_cred *rcred, int flags)
 	if (groups > NFS_NGROUPS)
 		groups = NFS_NGROUPS;
 	for (i = 0; i < groups ; i++)
-		if (!gid_eq(cred->uc_gids[i], GROUP_AT(acred->group_info, i)))
+		if (!gid_eq(cred->uc_gids[i], acred->group_info->gid[i]))
 			return 0;
 	if (groups < NFS_NGROUPS && gid_valid(cred->uc_gids[groups]))
 		return 0;
--- a/net/sunrpc/svcauth_unix.c
+++ b/net/sunrpc/svcauth_unix.c
@@ -517,7 +517,7 @@ static int unix_gid_parse(struct cache_detail *cd,
 		kgid = make_kgid(&init_user_ns, gid);
 		if (!gid_valid(kgid))
 			goto out;
-		GROUP_AT(ug.gi, i) = kgid;
+		ug.gi->gid[i] = kgid;
 	}
 
 	ugp = unix_gid_lookup(cd, uid);
@@ -564,7 +564,7 @@ static int unix_gid_show(struct seq_file *m,
 
 	seq_printf(m, "%u %d:", from_kuid_munged(user_ns, ug->uid), glen);
 	for (i = 0; i < glen; i++)
-		seq_printf(m, " %d", from_kgid_munged(user_ns, GROUP_AT(ug->gi, i)));
+		seq_printf(m, " %d", from_kgid_munged(user_ns, ug->gi->gid[i]));
 	seq_printf(m, "\n");
 	return 0;
 }
@@ -817,7 +817,7 @@ svcauth_unix_accept(struct svc_rqst *rqstp, __be32 *authp)
 		return SVC_CLOSE;
 	for (i = 0; i < slen; i++) {
 		kgid_t kgid = make_kgid(&init_user_ns, svc_getnl(argv));
-		GROUP_AT(cred->cr_group_info, i) = kgid;
+		cred->cr_group_info->gid[i] = kgid;
 	}
 	if (svc_getu32(argv) != htonl(RPC_AUTH_NULL) || svc_getu32(argv) != 0) {
 		*authp = rpc_autherr_badverf;

                 reply index

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20160817201927.GA2096@p183.telecom.by \
    --to=adobriyan@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-nfs@vger.kernel.org \
    --cc=segoon@openwall.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-NFS Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-nfs/0 linux-nfs/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-nfs linux-nfs/ https://lore.kernel.org/linux-nfs \
		linux-nfs@vger.kernel.org
	public-inbox-index linux-nfs

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-nfs


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git