summaryrefslogtreecommitdiffhomepage
path: root/pkg/sentry/vfs
diff options
context:
space:
mode:
Diffstat (limited to 'pkg/sentry/vfs')
-rw-r--r--pkg/sentry/vfs/BUILD18
-rw-r--r--pkg/sentry/vfs/context.go2
-rw-r--r--pkg/sentry/vfs/device.go31
-rw-r--r--pkg/sentry/vfs/file_description.go70
-rw-r--r--pkg/sentry/vfs/file_description_impl_util.go4
-rw-r--r--pkg/sentry/vfs/file_description_impl_util_test.go6
-rw-r--r--pkg/sentry/vfs/filesystem.go2
-rw-r--r--pkg/sentry/vfs/filesystem_type.go2
-rw-r--r--pkg/sentry/vfs/mount.go2
-rw-r--r--pkg/sentry/vfs/pathname.go2
-rw-r--r--pkg/sentry/vfs/permissions.go29
-rw-r--r--pkg/sentry/vfs/testutil.go2
-rw-r--r--pkg/sentry/vfs/vfs.go20
13 files changed, 152 insertions, 38 deletions
diff --git a/pkg/sentry/vfs/BUILD b/pkg/sentry/vfs/BUILD
index 35c7be259..6b1009328 100644
--- a/pkg/sentry/vfs/BUILD
+++ b/pkg/sentry/vfs/BUILD
@@ -1,7 +1,6 @@
-load("//tools/go_stateify:defs.bzl", "go_library")
-load("@io_bazel_rules_go//go:def.bzl", "go_test")
+load("//tools:defs.bzl", "go_library", "go_test")
-package(licenses = ["notice"])
+licenses(["notice"])
go_library(
name = "vfs",
@@ -24,18 +23,17 @@ go_library(
"testutil.go",
"vfs.go",
],
- importpath = "gvisor.dev/gvisor/pkg/sentry/vfs",
visibility = ["//pkg/sentry:internal"],
deps = [
"//pkg/abi/linux",
+ "//pkg/context",
"//pkg/fspath",
"//pkg/sentry/arch",
- "//pkg/sentry/context",
"//pkg/sentry/kernel/auth",
"//pkg/sentry/memmap",
- "//pkg/sentry/usermem",
"//pkg/sync",
"//pkg/syserror",
+ "//pkg/usermem",
"//pkg/waiter",
],
)
@@ -47,14 +45,14 @@ go_test(
"file_description_impl_util_test.go",
"mount_test.go",
],
- embed = [":vfs"],
+ library = ":vfs",
deps = [
"//pkg/abi/linux",
- "//pkg/sentry/context",
- "//pkg/sentry/context/contexttest",
+ "//pkg/context",
+ "//pkg/sentry/contexttest",
"//pkg/sentry/kernel/auth",
- "//pkg/sentry/usermem",
"//pkg/sync",
"//pkg/syserror",
+ "//pkg/usermem",
],
)
diff --git a/pkg/sentry/vfs/context.go b/pkg/sentry/vfs/context.go
index 705194ebc..d97362b9a 100644
--- a/pkg/sentry/vfs/context.go
+++ b/pkg/sentry/vfs/context.go
@@ -15,7 +15,7 @@
package vfs
import (
- "gvisor.dev/gvisor/pkg/sentry/context"
+ "gvisor.dev/gvisor/pkg/context"
)
// contextID is this package's type for context.Context.Value keys.
diff --git a/pkg/sentry/vfs/device.go b/pkg/sentry/vfs/device.go
index cb672e36f..3af2aa58d 100644
--- a/pkg/sentry/vfs/device.go
+++ b/pkg/sentry/vfs/device.go
@@ -17,7 +17,7 @@ package vfs
import (
"fmt"
- "gvisor.dev/gvisor/pkg/sentry/context"
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -98,3 +98,32 @@ func (vfs *VirtualFilesystem) OpenDeviceSpecialFile(ctx context.Context, mnt *Mo
}
return rd.dev.Open(ctx, mnt, d, *opts)
}
+
+// GetAnonBlockDevMinor allocates and returns an unused minor device number for
+// an "anonymous" block device with major number 0.
+func (vfs *VirtualFilesystem) GetAnonBlockDevMinor() (uint32, error) {
+ vfs.anonBlockDevMinorMu.Lock()
+ defer vfs.anonBlockDevMinorMu.Unlock()
+ minor := vfs.anonBlockDevMinorNext
+ const maxDevMinor = (1 << 20) - 1
+ for minor < maxDevMinor {
+ if _, ok := vfs.anonBlockDevMinor[minor]; !ok {
+ vfs.anonBlockDevMinor[minor] = struct{}{}
+ vfs.anonBlockDevMinorNext = minor + 1
+ return minor, nil
+ }
+ minor++
+ }
+ return 0, syserror.EMFILE
+}
+
+// PutAnonBlockDevMinor deallocates a minor device number returned by a
+// previous call to GetAnonBlockDevMinor.
+func (vfs *VirtualFilesystem) PutAnonBlockDevMinor(minor uint32) {
+ vfs.anonBlockDevMinorMu.Lock()
+ defer vfs.anonBlockDevMinorMu.Unlock()
+ delete(vfs.anonBlockDevMinor, minor)
+ if minor < vfs.anonBlockDevMinorNext {
+ vfs.anonBlockDevMinorNext = minor
+ }
+}
diff --git a/pkg/sentry/vfs/file_description.go b/pkg/sentry/vfs/file_description.go
index 6afe280bc..225024463 100644
--- a/pkg/sentry/vfs/file_description.go
+++ b/pkg/sentry/vfs/file_description.go
@@ -18,12 +18,12 @@ import (
"sync/atomic"
"gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/sentry/arch"
- "gvisor.dev/gvisor/pkg/sentry/context"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/memmap"
- "gvisor.dev/gvisor/pkg/sentry/usermem"
"gvisor.dev/gvisor/pkg/syserror"
+ "gvisor.dev/gvisor/pkg/usermem"
"gvisor.dev/gvisor/pkg/waiter"
)
@@ -49,8 +49,23 @@ type FileDescription struct {
// A reference is held on vd. vd is immutable.
vd VirtualDentry
+ // opts contains options passed to FileDescription.Init(). opts is
+ // immutable.
opts FileDescriptionOptions
+ // readable is MayReadFileWithOpenFlags(statusFlags). readable is
+ // immutable.
+ //
+ // readable is analogous to Linux's FMODE_READ.
+ readable bool
+
+ // writable is MayWriteFileWithOpenFlags(statusFlags). If writable is true,
+ // the FileDescription holds a write count on vd.mount. writable is
+ // immutable.
+ //
+ // writable is analogous to Linux's FMODE_WRITE.
+ writable bool
+
// impl is the FileDescriptionImpl associated with this Filesystem. impl is
// immutable. This should be the last field in FileDescription.
impl FileDescriptionImpl
@@ -77,10 +92,17 @@ type FileDescriptionOptions struct {
UseDentryMetadata bool
}
-// Init must be called before first use of fd. It takes references on mnt and
-// d. statusFlags is the initial file description status flags, which is
-// usually the full set of flags passed to open(2).
-func (fd *FileDescription) Init(impl FileDescriptionImpl, statusFlags uint32, mnt *Mount, d *Dentry, opts *FileDescriptionOptions) {
+// Init must be called before first use of fd. If it succeeds, it takes
+// references on mnt and d. statusFlags is the initial file description status
+// flags, which is usually the full set of flags passed to open(2).
+func (fd *FileDescription) Init(impl FileDescriptionImpl, statusFlags uint32, mnt *Mount, d *Dentry, opts *FileDescriptionOptions) error {
+ writable := MayWriteFileWithOpenFlags(statusFlags)
+ if writable {
+ if err := mnt.CheckBeginWrite(); err != nil {
+ return err
+ }
+ }
+
fd.refs = 1
fd.statusFlags = statusFlags | linux.O_LARGEFILE
fd.vd = VirtualDentry{
@@ -89,7 +111,10 @@ func (fd *FileDescription) Init(impl FileDescriptionImpl, statusFlags uint32, mn
}
fd.vd.IncRef()
fd.opts = *opts
+ fd.readable = MayReadFileWithOpenFlags(statusFlags)
+ fd.writable = writable
fd.impl = impl
+ return nil
}
// IncRef increments fd's reference count.
@@ -117,6 +142,9 @@ func (fd *FileDescription) TryIncRef() bool {
func (fd *FileDescription) DecRef() {
if refs := atomic.AddInt64(&fd.refs, -1); refs == 0 {
fd.impl.Release()
+ if fd.writable {
+ fd.vd.mount.EndWrite()
+ }
fd.vd.DecRef()
} else if refs < 0 {
panic("FileDescription.DecRef() called without holding a reference")
@@ -194,6 +222,16 @@ func (fd *FileDescription) SetStatusFlags(ctx context.Context, creds *auth.Crede
return nil
}
+// IsReadable returns true if fd was opened for reading.
+func (fd *FileDescription) IsReadable() bool {
+ return fd.readable
+}
+
+// IsWritable returns true if fd was opened for writing.
+func (fd *FileDescription) IsWritable() bool {
+ return fd.writable
+}
+
// Impl returns the FileDescriptionImpl associated with fd.
func (fd *FileDescription) Impl() FileDescriptionImpl {
return fd.impl
@@ -241,6 +279,8 @@ type FileDescriptionImpl interface {
// Errors:
//
// - If opts.Flags specifies unsupported options, PRead returns EOPNOTSUPP.
+ //
+ // Preconditions: The FileDescription was opened for reading.
PRead(ctx context.Context, dst usermem.IOSequence, offset int64, opts ReadOptions) (int64, error)
// Read is similar to PRead, but does not specify an offset.
@@ -254,6 +294,8 @@ type FileDescriptionImpl interface {
// Errors:
//
// - If opts.Flags specifies unsupported options, Read returns EOPNOTSUPP.
+ //
+ // Preconditions: The FileDescription was opened for reading.
Read(ctx context.Context, dst usermem.IOSequence, opts ReadOptions) (int64, error)
// PWrite writes src to the file, starting at the given offset, and returns
@@ -268,6 +310,8 @@ type FileDescriptionImpl interface {
//
// - If opts.Flags specifies unsupported options, PWrite returns
// EOPNOTSUPP.
+ //
+ // Preconditions: The FileDescription was opened for writing.
PWrite(ctx context.Context, src usermem.IOSequence, offset int64, opts WriteOptions) (int64, error)
// Write is similar to PWrite, but does not specify an offset, which is
@@ -281,6 +325,8 @@ type FileDescriptionImpl interface {
// Errors:
//
// - If opts.Flags specifies unsupported options, Write returns EOPNOTSUPP.
+ //
+ // Preconditions: The FileDescription was opened for writing.
Write(ctx context.Context, src usermem.IOSequence, opts WriteOptions) (int64, error)
// IterDirents invokes cb on each entry in the directory represented by the
@@ -411,11 +457,17 @@ func (fd *FileDescription) StatFS(ctx context.Context) (linux.Statfs, error) {
// offset, and returns the number of bytes read. PRead is permitted to return
// partial reads with a nil error.
func (fd *FileDescription) PRead(ctx context.Context, dst usermem.IOSequence, offset int64, opts ReadOptions) (int64, error) {
+ if !fd.readable {
+ return 0, syserror.EBADF
+ }
return fd.impl.PRead(ctx, dst, offset, opts)
}
// Read is similar to PRead, but does not specify an offset.
func (fd *FileDescription) Read(ctx context.Context, dst usermem.IOSequence, opts ReadOptions) (int64, error) {
+ if !fd.readable {
+ return 0, syserror.EBADF
+ }
return fd.impl.Read(ctx, dst, opts)
}
@@ -423,11 +475,17 @@ func (fd *FileDescription) Read(ctx context.Context, dst usermem.IOSequence, opt
// offset, and returns the number of bytes written. PWrite is permitted to
// return partial writes with a nil error.
func (fd *FileDescription) PWrite(ctx context.Context, src usermem.IOSequence, offset int64, opts WriteOptions) (int64, error) {
+ if !fd.writable {
+ return 0, syserror.EBADF
+ }
return fd.impl.PWrite(ctx, src, offset, opts)
}
// Write is similar to PWrite, but does not specify an offset.
func (fd *FileDescription) Write(ctx context.Context, src usermem.IOSequence, opts WriteOptions) (int64, error) {
+ if !fd.writable {
+ return 0, syserror.EBADF
+ }
return fd.impl.Write(ctx, src, opts)
}
diff --git a/pkg/sentry/vfs/file_description_impl_util.go b/pkg/sentry/vfs/file_description_impl_util.go
index c00b3c84b..fb9b87fdc 100644
--- a/pkg/sentry/vfs/file_description_impl_util.go
+++ b/pkg/sentry/vfs/file_description_impl_util.go
@@ -19,12 +19,12 @@ import (
"io"
"gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/sentry/arch"
- "gvisor.dev/gvisor/pkg/sentry/context"
"gvisor.dev/gvisor/pkg/sentry/memmap"
- "gvisor.dev/gvisor/pkg/sentry/usermem"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
+ "gvisor.dev/gvisor/pkg/usermem"
"gvisor.dev/gvisor/pkg/waiter"
)
diff --git a/pkg/sentry/vfs/file_description_impl_util_test.go b/pkg/sentry/vfs/file_description_impl_util_test.go
index 9ed58512f..1720d325d 100644
--- a/pkg/sentry/vfs/file_description_impl_util_test.go
+++ b/pkg/sentry/vfs/file_description_impl_util_test.go
@@ -22,11 +22,11 @@ import (
"testing"
"gvisor.dev/gvisor/pkg/abi/linux"
- "gvisor.dev/gvisor/pkg/sentry/context"
- "gvisor.dev/gvisor/pkg/sentry/context/contexttest"
+ "gvisor.dev/gvisor/pkg/context"
+ "gvisor.dev/gvisor/pkg/sentry/contexttest"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
- "gvisor.dev/gvisor/pkg/sentry/usermem"
"gvisor.dev/gvisor/pkg/syserror"
+ "gvisor.dev/gvisor/pkg/usermem"
)
// fileDescription is the common fd struct which a filesystem implementation
diff --git a/pkg/sentry/vfs/filesystem.go b/pkg/sentry/vfs/filesystem.go
index ea78f555b..a06a6caf3 100644
--- a/pkg/sentry/vfs/filesystem.go
+++ b/pkg/sentry/vfs/filesystem.go
@@ -18,8 +18,8 @@ import (
"sync/atomic"
"gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/fspath"
- "gvisor.dev/gvisor/pkg/sentry/context"
)
// A Filesystem is a tree of nodes represented by Dentries, which forms part of
diff --git a/pkg/sentry/vfs/filesystem_type.go b/pkg/sentry/vfs/filesystem_type.go
index 023301780..c58b70728 100644
--- a/pkg/sentry/vfs/filesystem_type.go
+++ b/pkg/sentry/vfs/filesystem_type.go
@@ -18,7 +18,7 @@ import (
"bytes"
"fmt"
- "gvisor.dev/gvisor/pkg/sentry/context"
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
)
diff --git a/pkg/sentry/vfs/mount.go b/pkg/sentry/vfs/mount.go
index 00177b371..d39528051 100644
--- a/pkg/sentry/vfs/mount.go
+++ b/pkg/sentry/vfs/mount.go
@@ -19,7 +19,7 @@ import (
"sync/atomic"
"gvisor.dev/gvisor/pkg/abi/linux"
- "gvisor.dev/gvisor/pkg/sentry/context"
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/syserror"
)
diff --git a/pkg/sentry/vfs/pathname.go b/pkg/sentry/vfs/pathname.go
index cf80df90e..b318c681a 100644
--- a/pkg/sentry/vfs/pathname.go
+++ b/pkg/sentry/vfs/pathname.go
@@ -15,8 +15,8 @@
package vfs
import (
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/fspath"
- "gvisor.dev/gvisor/pkg/sentry/context"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
)
diff --git a/pkg/sentry/vfs/permissions.go b/pkg/sentry/vfs/permissions.go
index f1edb0680..f664581f4 100644
--- a/pkg/sentry/vfs/permissions.go
+++ b/pkg/sentry/vfs/permissions.go
@@ -30,6 +30,26 @@ const (
MayExec = 1
)
+// OnlyRead returns true if access _only_ allows read.
+func (a AccessTypes) OnlyRead() bool {
+ return a == MayRead
+}
+
+// MayRead returns true if access allows read.
+func (a AccessTypes) MayRead() bool {
+ return a&MayRead != 0
+}
+
+// MayWrite returns true if access allows write.
+func (a AccessTypes) MayWrite() bool {
+ return a&MayWrite != 0
+}
+
+// MayExec returns true if access allows exec.
+func (a AccessTypes) MayExec() bool {
+ return a&MayExec != 0
+}
+
// GenericCheckPermissions checks that creds has the given access rights on a
// file with the given permissions, UID, and GID, subject to the rules of
// fs/namei.c:generic_permission(). isDir is true if the file is a directory.
@@ -53,7 +73,7 @@ func GenericCheckPermissions(creds *auth.Credentials, ats AccessTypes, isDir boo
}
// CAP_DAC_READ_SEARCH allows the caller to read and search arbitrary
// directories, and read arbitrary non-directory files.
- if (isDir && (ats&MayWrite == 0)) || ats == MayRead {
+ if (isDir && !ats.MayWrite()) || ats.OnlyRead() {
if creds.HasCapability(linux.CAP_DAC_READ_SEARCH) {
return nil
}
@@ -61,7 +81,7 @@ func GenericCheckPermissions(creds *auth.Credentials, ats AccessTypes, isDir boo
// CAP_DAC_OVERRIDE allows arbitrary access to directories, read/write
// access to non-directory files, and execute access to non-directory files
// for which at least one execute bit is set.
- if isDir || (ats&MayExec == 0) || (mode&0111 != 0) {
+ if isDir || !ats.MayExec() || (mode&0111 != 0) {
if creds.HasCapability(linux.CAP_DAC_OVERRIDE) {
return nil
}
@@ -74,14 +94,13 @@ func GenericCheckPermissions(creds *auth.Credentials, ats AccessTypes, isDir boo
// the set of accesses permitted for the opened file:
//
// - O_TRUNC causes MayWrite to be set in the returned AccessTypes (since it
-// mutates the file), but does not permit the opened to write to the file
+// mutates the file), but does not permit writing to the open file description
// thereafter.
//
// - "Linux reserves the special, nonstandard access mode 3 (binary 11) in
// flags to mean: check for read and write permission on the file and return a
// file descriptor that can't be used for reading or writing." - open(2). Thus
-// AccessTypesForOpenFlags returns MayRead|MayWrite in this case, but
-// filesystems are responsible for ensuring that access is denied.
+// AccessTypesForOpenFlags returns MayRead|MayWrite in this case.
//
// Use May{Read,Write}FileWithOpenFlags() for these checks instead.
func AccessTypesForOpenFlags(flags uint32) AccessTypes {
diff --git a/pkg/sentry/vfs/testutil.go b/pkg/sentry/vfs/testutil.go
index ee5c8b9e2..392c7611e 100644
--- a/pkg/sentry/vfs/testutil.go
+++ b/pkg/sentry/vfs/testutil.go
@@ -18,8 +18,8 @@ import (
"fmt"
"gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/fspath"
- "gvisor.dev/gvisor/pkg/sentry/context"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/syserror"
)
diff --git a/pkg/sentry/vfs/vfs.go b/pkg/sentry/vfs/vfs.go
index 1f21b0b31..b2bf48853 100644
--- a/pkg/sentry/vfs/vfs.go
+++ b/pkg/sentry/vfs/vfs.go
@@ -31,8 +31,8 @@ import (
"fmt"
"gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/fspath"
- "gvisor.dev/gvisor/pkg/sentry/context"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
@@ -80,6 +80,14 @@ type VirtualFilesystem struct {
devicesMu sync.RWMutex
devices map[devTuple]*registeredDevice
+ // anonBlockDevMinor contains all allocated anonymous block device minor
+ // numbers. anonBlockDevMinorNext is a lower bound for the smallest
+ // unallocated anonymous block device number. anonBlockDevMinorNext and
+ // anonBlockDevMinor are protected by anonBlockDevMinorMu.
+ anonBlockDevMinorMu sync.Mutex
+ anonBlockDevMinorNext uint32
+ anonBlockDevMinor map[uint32]struct{}
+
// fsTypes contains all registered FilesystemTypes. fsTypes is protected by
// fsTypesMu.
fsTypesMu sync.RWMutex
@@ -94,10 +102,12 @@ type VirtualFilesystem struct {
// New returns a new VirtualFilesystem with no mounts or FilesystemTypes.
func New() *VirtualFilesystem {
vfs := &VirtualFilesystem{
- mountpoints: make(map[*Dentry]map[*Mount]struct{}),
- devices: make(map[devTuple]*registeredDevice),
- fsTypes: make(map[string]*registeredFilesystemType),
- filesystems: make(map[*Filesystem]struct{}),
+ mountpoints: make(map[*Dentry]map[*Mount]struct{}),
+ devices: make(map[devTuple]*registeredDevice),
+ anonBlockDevMinorNext: 1,
+ anonBlockDevMinor: make(map[uint32]struct{}),
+ fsTypes: make(map[string]*registeredFilesystemType),
+ filesystems: make(map[*Filesystem]struct{}),
}
vfs.mounts.Init()
return vfs
generated by cgit v1.2.3 (git 2.39.1) at 2025-05-28 18:32:46 +0000