Merge pull request #6262 from sudo-sturbia:msgqueue/syscalls3

PiperOrigin-RevId: 391416650

8 files changed, 208 insertions, 45 deletions

diff --git a/pkg/abi/linux/msgqueue.go b/pkg/abi/linux/msgqueue.go
index e1e8d0357..0612a8214 100644
--- a/pkg/abi/linux/msgqueue.go
+++ b/pkg/abi/linux/msgqueue.go
@@ -47,7 +47,7 @@ const (
 	MSGSSZ  = 16
 
 	// MSGSEG is simplified due to the inexistance of a ternary operator.
-	MSGSEG = (MSGPOOL * 1024) / MSGSSZ
+	MSGSEG = 0xffff
 )
 
 // MsqidDS is equivelant to struct msqid64_ds. Source:
diff --git a/pkg/sentry/kernel/ipc/object.go b/pkg/sentry/kernel/ipc/object.go
index 387b35e7e..facd157c7 100644
--- a/pkg/sentry/kernel/ipc/object.go
+++ b/pkg/sentry/kernel/ipc/object.go
@@ -19,6 +19,8 @@ package ipc
 
 import (
 	"gvisor.dev/gvisor/pkg/abi/linux"
+	"gvisor.dev/gvisor/pkg/context"
+	"gvisor.dev/gvisor/pkg/errors/linuxerr"
 	"gvisor.dev/gvisor/pkg/sentry/fs"
 	"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
 )
@@ -113,3 +115,36 @@ func (o *Object) CheckPermissions(creds *auth.Credentials, req fs.PermMask) bool
 	}
 	return creds.HasCapabilityIn(linux.CAP_IPC_OWNER, o.UserNS)
 }
+
+// Set modifies attributes for an IPC object. See *ctl(IPC_SET).
+//
+// Precondition: Mechanism.mu must be held.
+func (o *Object) Set(ctx context.Context, perm *linux.IPCPerm) error {
+	creds := auth.CredentialsFromContext(ctx)
+	uid := creds.UserNamespace.MapToKUID(auth.UID(perm.UID))
+	gid := creds.UserNamespace.MapToKGID(auth.GID(perm.GID))
+	if !uid.Ok() || !gid.Ok() {
+		// The man pages don't specify an errno for invalid uid/gid, but EINVAL
+		// is generally used for invalid arguments.
+		return linuxerr.EINVAL
+	}
+
+	if !o.CheckOwnership(creds) {
+		// "The argument cmd has the value IPC_SET or IPC_RMID, but the
+		//  effective user ID of the calling process is not the creator (as
+		//  found in msg_perm.cuid) or the owner (as found in msg_perm.uid)
+		//  of the message queue, and the caller is not privileged (Linux:
+		//  does not have the CAP_SYS_ADMIN capability)."
+		return linuxerr.EPERM
+	}
+
+	// User may only modify the lower 9 bits of the mode. All the other bits are
+	// always 0 for the underlying inode.
+	mode := linux.FileMode(perm.Mode & 0x1ff)
+
+	o.Perms = fs.FilePermsFromMode(mode)
+	o.Owner.UID = uid
+	o.Owner.GID = gid
+
+	return nil
+}
diff --git a/pkg/sentry/kernel/msgqueue/msgqueue.go b/pkg/sentry/kernel/msgqueue/msgqueue.go
index fab396d7c..7c459d076 100644
--- a/pkg/sentry/kernel/msgqueue/msgqueue.go
+++ b/pkg/sentry/kernel/msgqueue/msgqueue.go
@@ -206,6 +206,48 @@ func (r *Registry) FindByID(id ipc.ID) (*Queue, error) {
 	return mech.(*Queue), nil
 }
 
+// IPCInfo reports global parameters for message queues. See msgctl(IPC_INFO).
+func (r *Registry) IPCInfo(ctx context.Context) *linux.MsgInfo {
+	return &linux.MsgInfo{
+		MsgPool: linux.MSGPOOL,
+		MsgMap:  linux.MSGMAP,
+		MsgMax:  linux.MSGMAX,
+		MsgMnb:  linux.MSGMNB,
+		MsgMni:  linux.MSGMNI,
+		MsgSsz:  linux.MSGSSZ,
+		MsgTql:  linux.MSGTQL,
+		MsgSeg:  linux.MSGSEG,
+	}
+}
+
+// MsgInfo reports global parameters for message queues. See msgctl(MSG_INFO).
+func (r *Registry) MsgInfo(ctx context.Context) *linux.MsgInfo {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+
+	var messages, bytes uint64
+	r.reg.ForAllObjects(
+		func(o ipc.Mechanism) {
+			q := o.(*Queue)
+			q.mu.Lock()
+			messages += q.messageCount
+			bytes += q.byteCount
+			q.mu.Unlock()
+		},
+	)
+
+	return &linux.MsgInfo{
+		MsgPool: int32(r.reg.ObjectCount()),
+		MsgMap:  int32(messages),
+		MsgTql:  int32(bytes),
+		MsgMax:  linux.MSGMAX,
+		MsgMnb:  linux.MSGMNB,
+		MsgMni:  linux.MSGMNI,
+		MsgSsz:  linux.MSGSSZ,
+		MsgSeg:  linux.MSGSEG,
+	}
+}
+
 // Send appends a message to the message queue, and returns an error if sending
 // fails. See msgsnd(2).
 func (q *Queue) Send(ctx context.Context, m Message, b Blocker, wait bool, pid int32) error {
@@ -465,6 +507,73 @@ func (q *Queue) msgAtIndex(mType int64) *Message {
 	return msg
 }
 
+// Set modifies some values of the queue. See msgctl(IPC_SET).
+func (q *Queue) Set(ctx context.Context, ds *linux.MsqidDS) error {
+	q.mu.Lock()
+	defer q.mu.Unlock()
+
+	creds := auth.CredentialsFromContext(ctx)
+	if ds.MsgQbytes > maxQueueBytes && !creds.HasCapabilityIn(linux.CAP_SYS_RESOURCE, q.obj.UserNS) {
+		// "An attempt (IPC_SET) was made to increase msg_qbytes beyond the
+		// system parameter MSGMNB, but the caller is not privileged (Linux:
+		// does not have the CAP_SYS_RESOURCE capability)."
+		return linuxerr.EPERM
+	}
+
+	if err := q.obj.Set(ctx, &ds.MsgPerm); err != nil {
+		return err
+	}
+
+	q.maxBytes = ds.MsgQbytes
+	q.changeTime = ktime.NowFromContext(ctx)
+	return nil
+}
+
+// Stat returns a MsqidDS object filled with information about the queue. See
+// msgctl(IPC_STAT) and msgctl(MSG_STAT).
+func (q *Queue) Stat(ctx context.Context) (*linux.MsqidDS, error) {
+	return q.stat(ctx, fs.PermMask{Read: true})
+}
+
+// StatAny is similar to Queue.Stat, but doesn't require read permission. See
+// msgctl(MSG_STAT_ANY).
+func (q *Queue) StatAny(ctx context.Context) (*linux.MsqidDS, error) {
+	return q.stat(ctx, fs.PermMask{})
+}
+
+// stat returns a MsqidDS object filled with information about the queue. An
+// error is returned if the user doesn't have the specified permissions.
+func (q *Queue) stat(ctx context.Context, mask fs.PermMask) (*linux.MsqidDS, error) {
+	q.mu.Lock()
+	defer q.mu.Unlock()
+
+	creds := auth.CredentialsFromContext(ctx)
+	if !q.obj.CheckPermissions(creds, mask) {
+		// "The caller must have read permission on the message queue."
+		return nil, linuxerr.EACCES
+	}
+
+	return &linux.MsqidDS{
+		MsgPerm: linux.IPCPerm{
+			Key:  uint32(q.obj.Key),
+			UID:  uint32(creds.UserNamespace.MapFromKUID(q.obj.Owner.UID)),
+			GID:  uint32(creds.UserNamespace.MapFromKGID(q.obj.Owner.GID)),
+			CUID: uint32(creds.UserNamespace.MapFromKUID(q.obj.Creator.UID)),
+			CGID: uint32(creds.UserNamespace.MapFromKGID(q.obj.Creator.GID)),
+			Mode: uint16(q.obj.Perms.LinuxMode()),
+			Seq:  0, // IPC sequences not supported.
+		},
+		MsgStime:  q.sendTime.TimeT(),
+		MsgRtime:  q.receiveTime.TimeT(),
+		MsgCtime:  q.changeTime.TimeT(),
+		MsgCbytes: q.byteCount,
+		MsgQnum:   q.messageCount,
+		MsgQbytes: q.maxBytes,
+		MsgLspid:  q.sendPID,
+		MsgLrpid:  q.receivePID,
+	}, nil
+}
+
 // Lock implements ipc.Mechanism.Lock.
 func (q *Queue) Lock() {
 	q.mu.Lock()
diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go
index 8a5c81a68..28e466948 100644
--- a/pkg/sentry/kernel/semaphore/semaphore.go
+++ b/pkg/sentry/kernel/semaphore/semaphore.go
@@ -336,19 +336,15 @@ func (s *Set) Size() int {
 	return len(s.sems)
 }
 
-// Change changes some fields from the set atomically.
-func (s *Set) Change(ctx context.Context, creds *auth.Credentials, owner fs.FileOwner, perms fs.FilePermissions) error {
+// Set modifies attributes for a semaphore set. See semctl(IPC_SET).
+func (s *Set) Set(ctx context.Context, ds *linux.SemidDS) error {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
-	// "The effective UID of the calling process must match the owner or creator
-	// of the semaphore set, or the caller must be privileged."
-	if !s.obj.CheckOwnership(creds) {
-		return linuxerr.EACCES
+	if err := s.obj.Set(ctx, &ds.SemPerm); err != nil {
+		return err
 	}
 
-	s.obj.Owner = owner
-	s.obj.Perms = perms
 	s.changeTime = ktime.NowFromContext(ctx)
 	return nil
 }
diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go
index b8da0c76c..ab938fa3c 100644
--- a/pkg/sentry/kernel/shm/shm.go
+++ b/pkg/sentry/kernel/shm/shm.go
@@ -618,25 +618,10 @@ func (s *Shm) Set(ctx context.Context, ds *linux.ShmidDS) error {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
-	creds := auth.CredentialsFromContext(ctx)
-	if !s.obj.CheckOwnership(creds) {
-		return linuxerr.EPERM
-	}
-
-	uid := creds.UserNamespace.MapToKUID(auth.UID(ds.ShmPerm.UID))
-	gid := creds.UserNamespace.MapToKGID(auth.GID(ds.ShmPerm.GID))
-	if !uid.Ok() || !gid.Ok() {
-		return linuxerr.EINVAL
+	if err := s.obj.Set(ctx, &ds.ShmPerm); err != nil {
+		return err
 	}
 
-	// User may only modify the lower 9 bits of the mode. All the other bits are
-	// always 0 for the underlying inode.
-	mode := linux.FileMode(ds.ShmPerm.Mode & 0x1ff)
-	s.obj.Perms = fs.FilePermsFromMode(mode)
-
-	s.obj.Owner.UID = uid
-	s.obj.Owner.GID = gid
-
 	s.changeTime = ktime.NowFromContext(ctx)
 	return nil
 }
diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go
index 56f90d952..2046a48b9 100644
--- a/pkg/sentry/syscalls/linux/linux64.go
+++ b/pkg/sentry/syscalls/linux/linux64.go
@@ -123,7 +123,7 @@ var AMD64 = &kernel.SyscallTable{
 		68:  syscalls.Supported("msgget", Msgget),
 		69:  syscalls.Supported("msgsnd", Msgsnd),
 		70:  syscalls.Supported("msgrcv", Msgrcv),
-		71:  syscalls.PartiallySupported("msgctl", Msgctl, "Only supports IPC_RMID option.", []string{"gvisor.dev/issue/135"}),
+		71:  syscalls.Supported("msgctl", Msgctl),
 		72:  syscalls.PartiallySupported("fcntl", Fcntl, "Not all options are supported.", nil),
 		73:  syscalls.PartiallySupported("flock", Flock, "Locks are held within the sandbox only.", nil),
 		74:  syscalls.PartiallySupported("fsync", Fsync, "Full data flush is not guaranteed at this time.", nil),
@@ -616,7 +616,7 @@ var ARM64 = &kernel.SyscallTable{
 		184: syscalls.ErrorWithEvent("mq_notify", linuxerr.ENOSYS, "", []string{"gvisor.dev/issue/136"}),       // TODO(b/29354921)
 		185: syscalls.ErrorWithEvent("mq_getsetattr", linuxerr.ENOSYS, "", []string{"gvisor.dev/issue/136"}),   // TODO(b/29354921)
 		186: syscalls.Supported("msgget", Msgget),
-		187: syscalls.PartiallySupported("msgctl", Msgctl, "Only supports IPC_RMID option.", []string{"gvisor.dev/issue/135"}),
+		187: syscalls.Supported("msgctl", Msgctl),
 		188: syscalls.Supported("msgrcv", Msgrcv),
 		189: syscalls.Supported("msgsnd", Msgsnd),
 		190: syscalls.Supported("semget", Semget),
diff --git a/pkg/sentry/syscalls/linux/sys_msgqueue.go b/pkg/sentry/syscalls/linux/sys_msgqueue.go
index 5259ade90..60b989ee7 100644
--- a/pkg/sentry/syscalls/linux/sys_msgqueue.go
+++ b/pkg/sentry/syscalls/linux/sys_msgqueue.go
@@ -130,12 +130,63 @@ func receive(t *kernel.Task, id ipc.ID, mType int64, maxSize int64, msgCopy, wai
 func Msgctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
 	id := ipc.ID(args[0].Int())
 	cmd := args[1].Int()
+	buf := args[2].Pointer()
 
 	creds := auth.CredentialsFromContext(t)
 
+	r := t.IPCNamespace().MsgqueueRegistry()
+
 	switch cmd {
+	case linux.IPC_INFO:
+		info := r.IPCInfo(t)
+		_, err := info.CopyOut(t, buf)
+		return 0, nil, err
+	case linux.MSG_INFO:
+		msgInfo := r.MsgInfo(t)
+		_, err := msgInfo.CopyOut(t, buf)
+		return 0, nil, err
 	case linux.IPC_RMID:
-		return 0, nil, t.IPCNamespace().MsgqueueRegistry().Remove(id, creds)
+		return 0, nil, r.Remove(id, creds)
+	}
+
+	// Remaining commands use a queue.
+	queue, err := r.FindByID(id)
+	if err != nil {
+		return 0, nil, err
+	}
+
+	switch cmd {
+	case linux.MSG_STAT:
+		// Technically, we should be treating id as "an index into the kernel's
+		// internal array that maintains information about all shared memory
+		// segments on the system". Since we don't track segments in an array,
+		// we'll just pretend the msqid is the index and do the same thing as
+		// IPC_STAT. Linux also uses the index as the msqid.
+		fallthrough
+	case linux.IPC_STAT:
+		stat, err := queue.Stat(t)
+		if err != nil {
+			return 0, nil, err
+		}
+		_, err = stat.CopyOut(t, buf)
+		return 0, nil, err
+
+	case linux.MSG_STAT_ANY:
+		stat, err := queue.StatAny(t)
+		if err != nil {
+			return 0, nil, err
+		}
+		_, err = stat.CopyOut(t, buf)
+		return 0, nil, err
+
+	case linux.IPC_SET:
+		var ds linux.MsqidDS
+		if _, err := ds.CopyIn(t, buf); err != nil {
+			return 0, nil, linuxerr.EINVAL
+		}
+		err := queue.Set(t, &ds)
+		return 0, nil, err
+
 	default:
 		return 0, nil, linuxerr.EINVAL
 	}
diff --git a/pkg/sentry/syscalls/linux/sys_sem.go b/pkg/sentry/syscalls/linux/sys_sem.go
index f61cc466c..5a119b21c 100644
--- a/pkg/sentry/syscalls/linux/sys_sem.go
+++ b/pkg/sentry/syscalls/linux/sys_sem.go
@@ -23,7 +23,6 @@ import (
 	"gvisor.dev/gvisor/pkg/hostarch"
 	"gvisor.dev/gvisor/pkg/marshal/primitive"
 	"gvisor.dev/gvisor/pkg/sentry/arch"
-	"gvisor.dev/gvisor/pkg/sentry/fs"
 	"gvisor.dev/gvisor/pkg/sentry/kernel"
 	"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
 	"gvisor.dev/gvisor/pkg/sentry/kernel/ipc"
@@ -166,8 +165,7 @@ func Semctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal
 			return 0, nil, err
 		}
 
-		perms := fs.FilePermsFromMode(linux.FileMode(s.SemPerm.Mode & 0777))
-		return 0, nil, ipcSet(t, id, auth.UID(s.SemPerm.UID), auth.GID(s.SemPerm.GID), perms)
+		return 0, nil, ipcSet(t, id, &s)
 
 	case linux.GETPID:
 		v, err := getPID(t, id, num)
@@ -243,24 +241,13 @@ func remove(t *kernel.Task, id ipc.ID) error {
 	return r.Remove(id, creds)
 }
 
-func ipcSet(t *kernel.Task, id ipc.ID, uid auth.UID, gid auth.GID, perms fs.FilePermissions) error {
+func ipcSet(t *kernel.Task, id ipc.ID, ds *linux.SemidDS) error {
 	r := t.IPCNamespace().SemaphoreRegistry()
 	set := r.FindByID(id)
 	if set == nil {
 		return linuxerr.EINVAL
 	}
-
-	creds := auth.CredentialsFromContext(t)
-	kuid := creds.UserNamespace.MapToKUID(uid)
-	if !kuid.Ok() {
-		return linuxerr.EINVAL
-	}
-	kgid := creds.UserNamespace.MapToKGID(gid)
-	if !kgid.Ok() {
-		return linuxerr.EINVAL
-	}
-	owner := fs.FileOwner{UID: kuid, GID: kgid}
-	return set.Change(t, creds, owner, perms)
+	return set.Set(t, ds)
 }
 
 func ipcStat(t *kernel.Task, id ipc.ID) (*linux.SemidDS, error) {