Kernel+LibCore+LibC: Split the mount syscall into multiple syscalls

This is a preparation before we can create a usable mechanism to use filesystem-specific mount flags. To keep some compatibility with userland code, LibC and LibCore mount functions are kept being usable, but now instead of doing an "atomic" syscall, they do multiple syscalls to perform the complete procedure of mounting a filesystem. The FileBackedFileSystem IntrusiveList in the VFS code is now changed to be protected by a Mutex, because when we mount a new filesystem, we need to check if a filesystem is already created for a given source_fd so we do a scan for that OpenFileDescription in that list. If we fail to find an already-created filesystem we create a new one and register it in the list if we successfully mounted it. We use a Mutex because we might need to initiate disk access during the filesystem creation, which will take other mutexes in other parts of the kernel, therefore making it not possible to take a spinlock while doing this.
2024-10-04 15:09:40 +00:00 · 2022-12-15 11:42:40 +02:00 · 2022-12-15 11:42:40 +02:00 · 23a7ccf607
parent 63fc36d00d
commit 23a7ccf607
39 changed files with 651 additions and 214 deletions
--- a/Base/usr/share/man/man2/pledge.md
+++ b/Base/usr/share/man/man2/pledge.md
@ -54,6 +54,7 @@ If the process later attempts to use any system functionality it has previously
 * `ptrace`: The [`ptrace`(2)](help://man/2/ptrace) syscall (\*)
 * `prot_exec`: [`mmap`(2)](help://man/2/mmap) and [`mprotect`(2)](help://man/2/mprotect) with `PROT_EXEC`
 * `map_fixed`: [`mmap`(2)](help://man/2/mmap) with `MAP_FIXED` or `MAP_FIXED_NOREPLACE` (\*)
+* `mount`: [`mount`(2)](help://man/2/mount) Various filesystem mount related syscalls (\*)
 * `no_error`: Ignore requests of pledge elevation going forwards, this is useful for enforcing _execpromises_ while the child process wants to ask for more upfront (Note that the elevation requests are _not_ granted, merely ignored), this is similar to the `error` pledge in OpenBSD.
 * `jail`: Various jail-specific syscalls (\*)

--- a/Kernel/API/FileSystem/MountSpecificFlags.h
+++ b/Kernel/API/FileSystem/MountSpecificFlags.h
@ -0,0 +1,30 @@
+/*
+ * Copyright (c) 2022-2023, Liav A. <liavalb@hotmail.co.il>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include <AK/Types.h>
+
+#define MOUNT_SPECIFIC_FLAG_KEY_STRING_MAX_LENGTH 64
+
+#define MOUNT_SPECIFIC_FLAG_NON_ASCII_STRING_TYPE_MAX_LENGTH 64
+#define MOUNT_SPECIFIC_FLAG_ASCII_STRING_TYPE_MAX_LENGTH 1024
+
+struct MountSpecificFlag {
+    u32 key_string_length;
+    u32 value_length;
+
+    enum class ValueType : u32 {
+        Boolean = 0,
+        UnsignedInteger,
+        SignedInteger,
+        ASCIIString,
+    };
+
+    ValueType value_type;
+    unsigned char const* key_string_addr;
+    void const* value_addr;
+};
--- a/Kernel/API/Ioctl.h
+++ b/Kernel/API/Ioctl.h
@ -108,6 +108,7 @@ enum IOCtlNumber {
    KEYBOARD_IOCTL_SET_NUM_LOCK,
    KEYBOARD_IOCTL_GET_CAPS_LOCK,
    KEYBOARD_IOCTL_SET_CAPS_LOCK,
+    MOUNT_IOCTL_SET_MOUNT_SPECIFIC_FLAG,
    SIOCATMARK,
    SIOCSIFADDR,
    SIOCGIFADDR,
@ -190,6 +191,7 @@ enum IOCtlNumber {
 #define FIBMAP FIBMAP
 #define FIONBIO FIONBIO
 #define FIONREAD FIONREAD
+#define MOUNT_IOCTL_SET_MOUNT_SPECIFIC_FLAG MOUNT_IOCTL_SET_MOUNT_SPECIFIC_FLAG
 #define SOUNDCARD_IOCTL_SET_SAMPLE_RATE SOUNDCARD_IOCTL_SET_SAMPLE_RATE
 #define SOUNDCARD_IOCTL_GET_SAMPLE_RATE SOUNDCARD_IOCTL_GET_SAMPLE_RATE
 #define STORAGE_DEVICE_GET_SIZE STORAGE_DEVICE_GET_SIZE
--- a/Kernel/API/Syscall.h
+++ b/Kernel/API/Syscall.h
@ -82,6 +82,8 @@ enum class NeedsBigProcessLock {
    S(fork, NeedsBigProcessLock::No)                       \
    S(fstat, NeedsBigProcessLock::No)                      \
    S(fstatvfs, NeedsBigProcessLock::No)                   \
+    S(fsopen, NeedsBigProcessLock::No)                     \
+    S(fsmount, NeedsBigProcessLock::No)                    \
    S(fsync, NeedsBigProcessLock::No)                      \
    S(ftruncate, NeedsBigProcessLock::No)                  \
    S(futex, NeedsBigProcessLock::Yes)                     \
@ -128,7 +130,6 @@ enum class NeedsBigProcessLock {
    S(mkdir, NeedsBigProcessLock::No)                      \
    S(mknod, NeedsBigProcessLock::No)                      \
    S(mmap, NeedsBigProcessLock::No)                       \
-    S(mount, NeedsBigProcessLock::Yes)                     \
    S(mprotect, NeedsBigProcessLock::No)                   \
    S(mremap, NeedsBigProcessLock::No)                     \
    S(msync, NeedsBigProcessLock::No)                      \
@ -432,9 +433,19 @@ struct SC_rename_params {
    StringArgument new_path;
 };

-struct SC_mount_params {
-    StringArgument target;
+struct SC_fsopen_params {
    StringArgument fs_type;
+    int flags;
+};
+
+struct SC_fsmount_params {
+    int mount_fd;
+    StringArgument target;
+    int source_fd;
+};
+
+struct SC_bindmount_params {
+    StringArgument target;
    int source_fd;
    int flags;
 };
@ -444,12 +455,6 @@ struct SC_remount_params {
    int flags;
 };

-struct SC_bindmount_params {
-    StringArgument target;
-    int source_fd;
-    int flags;
-};
-
 struct SC_pledge_params {
    StringArgument promises;
    StringArgument execpromises;
--- a/Kernel/CMakeLists.txt
+++ b/Kernel/CMakeLists.txt
@ -138,6 +138,7 @@ set(KERNEL_SOURCES
    FileSystem/ISO9660FS/FileSystem.cpp
    FileSystem/ISO9660FS/Inode.cpp
    FileSystem/Mount.cpp
+    FileSystem/MountFile.cpp
    FileSystem/OpenFileDescription.cpp
    FileSystem/Plan9FS/FileSystem.cpp
    FileSystem/Plan9FS/Inode.cpp
--- a/Kernel/Devices/Storage/StorageManagement.cpp
+++ b/Kernel/Devices/Storage/StorageManagement.cpp
@ -456,7 +456,9 @@ NonnullRefPtr<FileSystem> StorageManagement::root_filesystem() const
    auto description_or_error = OpenFileDescription::try_create(boot_device_description.release_nonnull());
    VERIFY(!description_or_error.is_error());

-    auto file_system = Ext2FS::try_create(description_or_error.release_value()).release_value();
+    Array<u8, PAGE_SIZE> mount_specific_data;
+    mount_specific_data.fill(0);
+    auto file_system = Ext2FS::try_create(description_or_error.release_value(), mount_specific_data.span()).release_value();

    if (auto result = file_system->initialize(); result.is_error()) {
        dump_storage_devices_and_partitions();
--- a/Kernel/FileSystem/DevPtsFS/FileSystem.cpp
+++ b/Kernel/FileSystem/DevPtsFS/FileSystem.cpp
@ -13,7 +13,7 @@

 namespace Kernel {

-ErrorOr<NonnullRefPtr<FileSystem>> DevPtsFS::try_create()
+ErrorOr<NonnullRefPtr<FileSystem>> DevPtsFS::try_create(ReadonlyBytes)
 {
    return TRY(adopt_nonnull_ref_or_enomem(new (nothrow) DevPtsFS));
 }
--- a/Kernel/FileSystem/DevPtsFS/FileSystem.h
+++ b/Kernel/FileSystem/DevPtsFS/FileSystem.h
@ -20,7 +20,7 @@ class DevPtsFS final : public FileSystem {

 public:
    virtual ~DevPtsFS() override;
-    static ErrorOr<NonnullRefPtr<FileSystem>> try_create();
+    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(ReadonlyBytes);

    virtual ErrorOr<void> initialize() override;
    virtual StringView class_name() const override { return "DevPtsFS"sv; }
--- a/Kernel/FileSystem/Ext2FS/FileSystem.cpp
+++ b/Kernel/FileSystem/Ext2FS/FileSystem.cpp
@ -13,7 +13,7 @@

 namespace Kernel {

-ErrorOr<NonnullRefPtr<FileSystem>> Ext2FS::try_create(OpenFileDescription& file_description)
+ErrorOr<NonnullRefPtr<FileSystem>> Ext2FS::try_create(OpenFileDescription& file_description, ReadonlyBytes)
 {
    return TRY(adopt_nonnull_ref_or_enomem(new (nothrow) Ext2FS(file_description)));
 }
--- a/Kernel/FileSystem/Ext2FS/FileSystem.h
+++ b/Kernel/FileSystem/Ext2FS/FileSystem.h
@ -27,7 +27,7 @@ public:
        FileSize64bits = 1 << 1,
    };

-    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(OpenFileDescription&);
+    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(OpenFileDescription&, ReadonlyBytes);

    virtual ~Ext2FS() override;

--- a/Kernel/FileSystem/FATFS/FileSystem.cpp
+++ b/Kernel/FileSystem/FATFS/FileSystem.cpp
@ -10,7 +10,7 @@

 namespace Kernel {

-ErrorOr<NonnullRefPtr<FileSystem>> FATFS::try_create(OpenFileDescription& file_description)
+ErrorOr<NonnullRefPtr<FileSystem>> FATFS::try_create(OpenFileDescription& file_description, ReadonlyBytes)
 {
    return TRY(adopt_nonnull_ref_or_enomem(new (nothrow) FATFS(file_description)));
 }
--- a/Kernel/FileSystem/FATFS/FileSystem.h
+++ b/Kernel/FileSystem/FATFS/FileSystem.h
@ -21,7 +21,7 @@ class FATFS final : public BlockBasedFileSystem {
    friend FATInode;

 public:
-    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(OpenFileDescription&);
+    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(OpenFileDescription&, ReadonlyBytes);

    virtual ~FATFS() override = default;
    virtual StringView class_name() const override { return "FATFS"sv; }
--- a/Kernel/FileSystem/File.h
+++ b/Kernel/FileSystem/File.h
@ -114,6 +114,7 @@ public:
    virtual bool is_character_device() const { return false; }
    virtual bool is_socket() const { return false; }
    virtual bool is_inode_watcher() const { return false; }
+    virtual bool is_mount_file() const { return false; }

    virtual bool is_regular_file() const { return false; }

--- a/Kernel/FileSystem/ISO9660FS/FileSystem.cpp
+++ b/Kernel/FileSystem/ISO9660FS/FileSystem.cpp
@ -16,7 +16,7 @@ constexpr u32 first_data_area_block = 16;
 constexpr u32 logical_sector_size = 2048;
 constexpr u32 max_cached_directory_entries = 128;

-ErrorOr<NonnullRefPtr<FileSystem>> ISO9660FS::try_create(OpenFileDescription& description)
+ErrorOr<NonnullRefPtr<FileSystem>> ISO9660FS::try_create(OpenFileDescription& description, ReadonlyBytes)
 {
    return TRY(adopt_nonnull_ref_or_enomem(new (nothrow) ISO9660FS(description)));
 }
--- a/Kernel/FileSystem/ISO9660FS/FileSystem.h
+++ b/Kernel/FileSystem/ISO9660FS/FileSystem.h
@ -29,7 +29,7 @@ class ISO9660FS final : public BlockBasedFileSystem {
    friend ISO9660DirectoryIterator;

 public:
-    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(OpenFileDescription&);
+    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(OpenFileDescription&, ReadonlyBytes);

    virtual ~ISO9660FS() override;
    virtual StringView class_name() const override { return "ISO9660FS"sv; }
--- a/Kernel/FileSystem/Initializer.h
+++ b/Kernel/FileSystem/Initializer.h
@ -0,0 +1,32 @@
+/*
+ * Copyright (c) 2022, Liav A. <liavalb@hotmail.co.il>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include <AK/Error.h>
+#include <AK/Span.h>
+#include <AK/StringView.h>
+#include <AK/Types.h>
+#include <Kernel/FileSystem/FileSystem.h>
+#include <Kernel/FileSystem/OpenFileDescription.h>
+
+namespace Kernel {
+
+struct FileSystemInitializer {
+    StringView short_name;
+    StringView name;
+    bool requires_open_file_description { false };
+    bool requires_block_device { false };
+    bool requires_seekable_file { false };
+    ErrorOr<NonnullRefPtr<FileSystem>> (*create_with_fd)(OpenFileDescription&, ReadonlyBytes) = nullptr;
+    ErrorOr<NonnullRefPtr<FileSystem>> (*create)(ReadonlyBytes) = nullptr;
+    ErrorOr<void> (*handle_mount_boolean_flag)(Span<u8>, StringView key, bool) = nullptr;
+    ErrorOr<void> (*handle_mount_unsigned_integer_flag)(Span<u8>, StringView key, u64) = nullptr;
+    ErrorOr<void> (*handle_mount_signed_integer_flag)(Span<u8>, StringView key, i64) = nullptr;
+    ErrorOr<void> (*handle_mount_ascii_string_flag)(Span<u8>, StringView key, StringView value) = nullptr;
+};
+
+}
--- a/Kernel/FileSystem/MountFile.cpp
+++ b/Kernel/FileSystem/MountFile.cpp
@ -0,0 +1,122 @@
+/*
+ * Copyright (c) 2022-2023, Liav A. <liavalb@hotmail.co.il>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#include <AK/StringView.h>
+#include <Kernel/API/FileSystem/MountSpecificFlags.h>
+#include <Kernel/API/Ioctl.h>
+#include <Kernel/API/POSIX/errno.h>
+#include <Kernel/API/POSIX/unistd.h>
+#include <Kernel/FileSystem/Inode.h>
+#include <Kernel/FileSystem/MountFile.h>
+#include <Kernel/FileSystem/OpenFileDescription.h>
+#include <Kernel/FileSystem/VirtualFileSystem.h>
+#include <Kernel/Library/StdLib.h>
+#include <Kernel/Memory/PrivateInodeVMObject.h>
+#include <Kernel/Memory/SharedInodeVMObject.h>
+
+namespace Kernel {
+
+ErrorOr<NonnullLockRefPtr<MountFile>> MountFile::create(FileSystemInitializer const& file_system_initializer, int flags)
+{
+    // NOTE: We should not open a MountFile if someone wants to either remount or bindmount.
+    // There's a check for this in the fsopen syscall entry handler, but here we just assert
+    // to ensure this never happens.
+    VERIFY(!(flags & MS_BIND));
+    VERIFY(!(flags & MS_REMOUNT));
+    auto mount_specific_data_buffer = TRY(KBuffer::try_create_with_size("Mount Specific Data"sv, PAGE_SIZE, Memory::Region::Access::ReadWrite, AllocationStrategy::AllocateNow));
+    return TRY(adopt_nonnull_lock_ref_or_enomem(new (nothrow) MountFile(file_system_initializer, flags, move(mount_specific_data_buffer))));
+}
+
+MountFile::MountFile(FileSystemInitializer const& file_system_initializer, int flags, NonnullOwnPtr<KBuffer> mount_specific_data)
+    : m_flags(flags)
+    , m_file_system_initializer(file_system_initializer)
+{
+    m_mount_specific_data.with_exclusive([&](auto& our_mount_specific_data) {
+        our_mount_specific_data = move(mount_specific_data);
+        memset(our_mount_specific_data->data(), 0, our_mount_specific_data->size());
+    });
+}
+
+MountFile::~MountFile() = default;
+
+ErrorOr<void> MountFile::ioctl(OpenFileDescription&, unsigned request, Userspace<void*> arg)
+{
+    return m_mount_specific_data.with_exclusive([&](auto& our_mount_specific_data) -> ErrorOr<void> {
+        switch (request) {
+        case MOUNT_IOCTL_SET_MOUNT_SPECIFIC_FLAG: {
+            auto user_mount_specific_data = static_ptr_cast<MountSpecificFlag const*>(arg);
+            auto mount_specific_data = TRY(copy_typed_from_user(user_mount_specific_data));
+            if ((mount_specific_data.value_type == MountSpecificFlag::ValueType::SignedInteger || mount_specific_data.value_type == MountSpecificFlag::ValueType::UnsignedInteger) && mount_specific_data.value_length != 8)
+                return EDOM;
+            if (mount_specific_data.key_string_length > MOUNT_SPECIFIC_FLAG_KEY_STRING_MAX_LENGTH)
+                return ENAMETOOLONG;
+            if (mount_specific_data.value_type != MountSpecificFlag::ValueType::Boolean && mount_specific_data.value_length == 0)
+                return EINVAL;
+            if (mount_specific_data.value_type != MountSpecificFlag::ValueType::Boolean && mount_specific_data.value_addr == nullptr)
+                return EFAULT;
+
+            // NOTE: We put these limits in place because we assume that don't need to handle huge
+            // amounts of bytes when trying to handle a mount fs-specific flag.
+            // Anything larger than these constants (which could be changed if needed) is deemed to
+            // potentially cause OOM condition, and cannot represent any reasonable and "honest" data
+            // from userspace.
+            if (mount_specific_data.value_type != MountSpecificFlag::ValueType::ASCIIString && mount_specific_data.value_length > MOUNT_SPECIFIC_FLAG_NON_ASCII_STRING_TYPE_MAX_LENGTH)
+                return E2BIG;
+            if (mount_specific_data.value_type == MountSpecificFlag::ValueType::ASCIIString && mount_specific_data.value_length > MOUNT_SPECIFIC_FLAG_ASCII_STRING_TYPE_MAX_LENGTH)
+                return E2BIG;
+
+            // NOTE: We enforce that the passed argument will be either i64 or u64, so it will always be
+            // exactly 8 bytes. We do that to simplify handling of integers as well as to ensure ABI correctness
+            // in all possible cases.
+            auto key_string = TRY(try_copy_kstring_from_user(reinterpret_cast<FlatPtr>(mount_specific_data.key_string_addr), static_cast<size_t>(mount_specific_data.key_string_length)));
+            switch (mount_specific_data.value_type) {
+            // NOTE: This is actually considered as simply boolean flag.
+            case MountSpecificFlag::ValueType::Boolean: {
+                VERIFY(m_file_system_initializer.handle_mount_boolean_flag);
+                Userspace<u64*> user_value_addr(reinterpret_cast<FlatPtr>(mount_specific_data.value_addr));
+                auto value_integer = TRY(copy_typed_from_user(user_value_addr));
+                if (value_integer != 0 && value_integer != 1)
+                    return EDOM;
+                bool value = (value_integer == 1) ? true : false;
+                TRY(m_file_system_initializer.handle_mount_boolean_flag(our_mount_specific_data->bytes(), key_string->view(), value));
+                return {};
+            }
+            case MountSpecificFlag::ValueType::UnsignedInteger: {
+                VERIFY(m_file_system_initializer.handle_mount_unsigned_integer_flag);
+                Userspace<u64*> user_value_addr(reinterpret_cast<FlatPtr>(mount_specific_data.value_addr));
+                auto value_integer = TRY(copy_typed_from_user(user_value_addr));
+                TRY(m_file_system_initializer.handle_mount_unsigned_integer_flag(our_mount_specific_data->bytes(), key_string->view(), value_integer));
+                return {};
+            }
+            case MountSpecificFlag::ValueType::SignedInteger: {
+                VERIFY(m_file_system_initializer.handle_mount_signed_integer_flag);
+                Userspace<i64*> user_value_addr(reinterpret_cast<FlatPtr>(mount_specific_data.value_addr));
+                auto value_integer = TRY(copy_typed_from_user(user_value_addr));
+                TRY(m_file_system_initializer.handle_mount_signed_integer_flag(our_mount_specific_data->bytes(), key_string->view(), value_integer));
+                return {};
+            }
+            case MountSpecificFlag::ValueType::ASCIIString: {
+                VERIFY(m_file_system_initializer.handle_mount_ascii_string_flag);
+                auto value_string = TRY(try_copy_kstring_from_user(reinterpret_cast<FlatPtr>(mount_specific_data.value_addr), static_cast<size_t>(mount_specific_data.value_length)));
+                TRY(m_file_system_initializer.handle_mount_ascii_string_flag(our_mount_specific_data->bytes(), key_string->view(), value_string->view()));
+                return {};
+            }
+            default:
+                return EINVAL;
+            }
+        }
+        default:
+            return EINVAL;
+        }
+    });
+}
+
+ErrorOr<NonnullOwnPtr<KString>> MountFile::pseudo_path(OpenFileDescription const&) const
+{
+    return KString::try_create(":mount-file:"sv);
+}
+
+}
--- a/Kernel/FileSystem/MountFile.h
+++ b/Kernel/FileSystem/MountFile.h
@ -0,0 +1,43 @@
+/*
+ * Copyright (c) 2022, Liav A. <liavalb@hotmail.co.il>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include <Kernel/FileSystem/File.h>
+#include <Kernel/FileSystem/Initializer.h>
+#include <Kernel/Locking/MutexProtected.h>
+
+namespace Kernel {
+
+class MountFile final : public File {
+public:
+    static ErrorOr<NonnullLockRefPtr<MountFile>> create(FileSystemInitializer const&, int flags);
+    virtual ~MountFile() override;
+
+    virtual bool can_read(OpenFileDescription const&, u64) const override { return true; }
+    virtual bool can_write(OpenFileDescription const&, u64) const override { return true; }
+    virtual ErrorOr<size_t> read(OpenFileDescription&, u64, UserOrKernelBuffer&, size_t) override { return ENOTSUP; }
+    virtual ErrorOr<size_t> write(OpenFileDescription&, u64, UserOrKernelBuffer const&, size_t) override { return ENOTSUP; }
+    virtual ErrorOr<void> ioctl(OpenFileDescription&, unsigned request, Userspace<void*> arg) override;
+    virtual ErrorOr<NonnullOwnPtr<KString>> pseudo_path(OpenFileDescription const&) const override;
+    virtual StringView class_name() const override { return "MountFile"sv; }
+
+    int mount_flags() const { return m_flags; }
+
+    MutexProtected<OwnPtr<KBuffer>>& mount_file_system_specific_data() { return m_mount_specific_data; }
+    FileSystemInitializer const& file_system_initializer() const { return m_file_system_initializer; }
+
+private:
+    virtual bool is_mount_file() const override { return true; }
+
+    MountFile(FileSystemInitializer const&, int flags, NonnullOwnPtr<KBuffer>);
+
+    int const m_flags;
+    FileSystemInitializer const& m_file_system_initializer;
+    MutexProtected<OwnPtr<KBuffer>> m_mount_specific_data;
+};
+
+}
--- a/Kernel/FileSystem/OpenFileDescription.cpp
+++ b/Kernel/FileSystem/OpenFileDescription.cpp
@ -12,6 +12,7 @@
 #include <Kernel/FileSystem/FIFO.h>
 #include <Kernel/FileSystem/InodeFile.h>
 #include <Kernel/FileSystem/InodeWatcher.h>
+#include <Kernel/FileSystem/MountFile.h>
 #include <Kernel/FileSystem/OpenFileDescription.h>
 #include <Kernel/FileSystem/VirtualFileSystem.h>
 #include <Kernel/Memory/MemoryManager.h>
@ -313,6 +314,25 @@ InodeWatcher* OpenFileDescription::inode_watcher()
    return static_cast<InodeWatcher*>(m_file.ptr());
 }

+bool OpenFileDescription::is_mount_file() const
+{
+    return m_file->is_mount_file();
+}
+
+MountFile const* OpenFileDescription::mount_file() const
+{
+    if (!is_mount_file())
+        return nullptr;
+    return static_cast<MountFile const*>(m_file.ptr());
+}
+
+MountFile* OpenFileDescription::mount_file()
+{
+    if (!is_mount_file())
+        return nullptr;
+    return static_cast<MountFile*>(m_file.ptr());
+}
+
 bool OpenFileDescription::is_master_pty() const
 {
    return m_file->is_master_pty();
--- a/Kernel/FileSystem/OpenFileDescription.h
+++ b/Kernel/FileSystem/OpenFileDescription.h
@ -80,6 +80,10 @@ public:
    InodeWatcher const* inode_watcher() const;
    InodeWatcher* inode_watcher();

+    bool is_mount_file() const;
+    MountFile const* mount_file() const;
+    MountFile* mount_file();
+
    bool is_master_pty() const;
    MasterPTY const* master_pty() const;
    MasterPTY* master_pty();
--- a/Kernel/FileSystem/Plan9FS/FileSystem.cpp
+++ b/Kernel/FileSystem/Plan9FS/FileSystem.cpp
@ -10,7 +10,7 @@

 namespace Kernel {

-ErrorOr<NonnullRefPtr<FileSystem>> Plan9FS::try_create(OpenFileDescription& file_description)
+ErrorOr<NonnullRefPtr<FileSystem>> Plan9FS::try_create(OpenFileDescription& file_description, ReadonlyBytes)
 {
    return TRY(adopt_nonnull_ref_or_enomem(new (nothrow) Plan9FS(file_description)));
 }
--- a/Kernel/FileSystem/Plan9FS/FileSystem.h
+++ b/Kernel/FileSystem/Plan9FS/FileSystem.h
@ -22,7 +22,7 @@ class Plan9FS final : public FileBackedFileSystem {

 public:
    virtual ~Plan9FS() override;
-    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(OpenFileDescription&);
+    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(OpenFileDescription&, ReadonlyBytes);

    virtual bool supports_watchers() const override { return false; }

--- a/Kernel/FileSystem/ProcFS/FileSystem.cpp
+++ b/Kernel/FileSystem/ProcFS/FileSystem.cpp
@ -11,7 +11,7 @@

 namespace Kernel {

-ErrorOr<NonnullRefPtr<FileSystem>> ProcFS::try_create()
+ErrorOr<NonnullRefPtr<FileSystem>> ProcFS::try_create(ReadonlyBytes)
 {
    return TRY(adopt_nonnull_ref_or_enomem(new (nothrow) ProcFS));
 }
--- a/Kernel/FileSystem/ProcFS/FileSystem.h
+++ b/Kernel/FileSystem/ProcFS/FileSystem.h
@ -20,7 +20,7 @@ class ProcFS final : public FileSystem {

 public:
    virtual ~ProcFS() override;
-    static ErrorOr<NonnullRefPtr<FileSystem>> try_create();
+    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(ReadonlyBytes);

    virtual ErrorOr<void> initialize() override;
    virtual StringView class_name() const override { return "ProcFS"sv; }
--- a/Kernel/FileSystem/RAMFS/FileSystem.cpp
+++ b/Kernel/FileSystem/RAMFS/FileSystem.cpp
@ -10,7 +10,7 @@

 namespace Kernel {

-ErrorOr<NonnullRefPtr<FileSystem>> RAMFS::try_create()
+ErrorOr<NonnullRefPtr<FileSystem>> RAMFS::try_create(ReadonlyBytes)
 {
    return TRY(adopt_nonnull_ref_or_enomem(new (nothrow) RAMFS));
 }
--- a/Kernel/FileSystem/RAMFS/FileSystem.h
+++ b/Kernel/FileSystem/RAMFS/FileSystem.h
@ -18,7 +18,7 @@ class RAMFS final : public FileSystem {

 public:
    virtual ~RAMFS() override;
-    static ErrorOr<NonnullRefPtr<FileSystem>> try_create();
+    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(ReadonlyBytes);
    virtual ErrorOr<void> initialize() override;

    virtual StringView class_name() const override { return "RAMFS"sv; }
--- a/Kernel/FileSystem/SysFS/FileSystem.cpp
+++ b/Kernel/FileSystem/SysFS/FileSystem.cpp
@ -11,7 +11,7 @@

 namespace Kernel {

-ErrorOr<NonnullRefPtr<FileSystem>> SysFS::try_create()
+ErrorOr<NonnullRefPtr<FileSystem>> SysFS::try_create(ReadonlyBytes)
 {
    return TRY(adopt_nonnull_ref_or_enomem(new (nothrow) SysFS));
 }
--- a/Kernel/FileSystem/SysFS/FileSystem.h
+++ b/Kernel/FileSystem/SysFS/FileSystem.h
@ -20,7 +20,7 @@ class SysFS final : public FileSystem {

 public:
    virtual ~SysFS() override;
-    static ErrorOr<NonnullRefPtr<FileSystem>> try_create();
+    static ErrorOr<NonnullRefPtr<FileSystem>> try_create(ReadonlyBytes);

    virtual ErrorOr<void> initialize() override;
    virtual StringView class_name() const override { return "SysFS"sv; }
--- a/Kernel/FileSystem/VirtualFileSystem.cpp
+++ b/Kernel/FileSystem/VirtualFileSystem.cpp
@ -1,5 +1,6 @@
 /*
 * Copyright (c) 2018-2020, Andreas Kling <kling@serenityos.org>
+ * Copyright (c) 2022-2023, Liav A. <liavalb@hotmail.co.il>
 *
 * SPDX-License-Identifier: BSD-2-Clause
 */
@ -23,11 +24,60 @@
 #include <Kernel/Sections.h>
 #include <Kernel/Tasks/Process.h>

+#include <Kernel/FileSystem/DevPtsFS/FileSystem.h>
+#include <Kernel/FileSystem/Ext2FS/FileSystem.h>
+#include <Kernel/FileSystem/FATFS/FileSystem.h>
+#include <Kernel/FileSystem/ISO9660FS/FileSystem.h>
+#include <Kernel/FileSystem/Plan9FS/FileSystem.h>
+#include <Kernel/FileSystem/ProcFS/FileSystem.h>
+#include <Kernel/FileSystem/RAMFS/FileSystem.h>
+#include <Kernel/FileSystem/SysFS/FileSystem.h>
+
 namespace Kernel {

 static Singleton<VirtualFileSystem> s_the;
 static constexpr int root_mount_flags = 0;

+static ErrorOr<void> handle_mount_boolean_flag_as_invalid(Span<u8>, StringView, bool)
+{
+    return EINVAL;
+}
+
+static ErrorOr<void> handle_mount_unsigned_integer_flag_as_invalid(Span<u8>, StringView, u64)
+{
+    return EINVAL;
+}
+
+static ErrorOr<void> handle_mount_signed_integer_flag_as_invalid(Span<u8>, StringView, i64)
+{
+    return EINVAL;
+}
+
+static ErrorOr<void> handle_mount_ascii_string_flag_as_invalid(Span<u8>, StringView, StringView)
+{
+    return EINVAL;
+}
+
+static constexpr FileSystemInitializer s_initializers[] = {
+    { "proc"sv, "ProcFS"sv, false, false, false, {}, ProcFS::try_create, handle_mount_boolean_flag_as_invalid, handle_mount_unsigned_integer_flag_as_invalid, handle_mount_signed_integer_flag_as_invalid, handle_mount_ascii_string_flag_as_invalid },
+    { "devpts"sv, "DevPtsFS"sv, false, false, false, {}, DevPtsFS::try_create, handle_mount_boolean_flag_as_invalid, handle_mount_unsigned_integer_flag_as_invalid, handle_mount_signed_integer_flag_as_invalid, handle_mount_ascii_string_flag_as_invalid },
+    { "sys"sv, "SysFS"sv, false, false, false, {}, SysFS::try_create, handle_mount_boolean_flag_as_invalid, handle_mount_unsigned_integer_flag_as_invalid, handle_mount_signed_integer_flag_as_invalid, handle_mount_ascii_string_flag_as_invalid },
+    { "ram"sv, "RAMFS"sv, false, false, false, {}, RAMFS::try_create, handle_mount_boolean_flag_as_invalid, handle_mount_unsigned_integer_flag_as_invalid, handle_mount_signed_integer_flag_as_invalid, handle_mount_ascii_string_flag_as_invalid },
+    { "ext2"sv, "Ext2FS"sv, true, true, true, Ext2FS::try_create, {}, handle_mount_boolean_flag_as_invalid, handle_mount_unsigned_integer_flag_as_invalid, handle_mount_signed_integer_flag_as_invalid, handle_mount_ascii_string_flag_as_invalid },
+    { "9p"sv, "Plan9FS"sv, true, true, true, Plan9FS::try_create, {}, handle_mount_boolean_flag_as_invalid, handle_mount_unsigned_integer_flag_as_invalid, handle_mount_signed_integer_flag_as_invalid, handle_mount_ascii_string_flag_as_invalid },
+    { "iso9660"sv, "ISO9660FS"sv, true, true, true, ISO9660FS::try_create, {}, handle_mount_boolean_flag_as_invalid, handle_mount_unsigned_integer_flag_as_invalid, handle_mount_signed_integer_flag_as_invalid, handle_mount_ascii_string_flag_as_invalid },
+    { "fat"sv, "FATFS"sv, true, true, true, FATFS::try_create, {}, handle_mount_boolean_flag_as_invalid, handle_mount_unsigned_integer_flag_as_invalid, handle_mount_signed_integer_flag_as_invalid, handle_mount_ascii_string_flag_as_invalid },
+};
+
+ErrorOr<FileSystemInitializer const*> VirtualFileSystem::find_filesystem_type_initializer(StringView fs_type)
+{
+    for (auto& initializer_entry : s_initializers) {
+        if (fs_type == initializer_entry.short_name || fs_type == initializer_entry.name)
+            return &initializer_entry;
+    }
+    return ENODEV;
+}
+
 UNMAP_AFTER_INIT void VirtualFileSystem::initialize()
 {
    s_the.ensure_instance();
@ -59,14 +109,14 @@ bool VirtualFileSystem::mount_point_exists_at_inode(InodeIdentifier inode_identi
    });
 }

-ErrorOr<void> VirtualFileSystem::mount(FileSystem& fs, Custody& mount_point, int flags)
+ErrorOr<void> VirtualFileSystem::add_file_system_to_mount_table(FileSystem& file_system, Custody& mount_point, int flags)
 {
-    auto new_mount = TRY(adopt_nonnull_own_or_enomem(new (nothrow) Mount(fs, &mount_point, flags)));
+    auto new_mount = TRY(adopt_nonnull_own_or_enomem(new (nothrow) Mount(file_system, &mount_point, flags)));
    return m_mounts.with([&](auto& mounts) -> ErrorOr<void> {
        auto& inode = mount_point.inode();
-        dbgln("VirtualFileSystem: FileSystemID {}, Mounting {} at inode {} with flags {}",
-            fs.fsid(),
-            fs.class_name(),
+        dbgln("VirtualFileSystem: FileSystemID {} (non file-backed), Mounting {} at inode {} with flags {}",
+            file_system.fsid(),
+            file_system.class_name(),
            inode.identifier(),
            flags);
        if (mount_point_exists_at_inode(inode.identifier())) {
@ -84,14 +134,8 @@ ErrorOr<void> VirtualFileSystem::mount(FileSystem& fs, Custody& mount_point, int
            // the FileSystem once it is no longer mounted).
            if (mounted_count == 1) {
                m_file_systems_list.with([&](auto& fs_list) {
-                    fs_list.append(fs);
+                    fs_list.append(file_system);
                });
-                if (fs.is_file_backed()) {
-                    auto& file_backed_fs = static_cast<FileBackedFileSystem&>(fs);
-                    m_file_backed_file_systems_list.with([&](auto& fs_list) {
-                        fs_list.append(file_backed_fs);
-                    });
-                }
            }
        });

@ -102,6 +146,65 @@ ErrorOr<void> VirtualFileSystem::mount(FileSystem& fs, Custody& mount_point, int
    });
 }

+ErrorOr<void> VirtualFileSystem::mount(MountFile& mount_file, OpenFileDescription* source_description, Custody& mount_point, int flags)
+{
+    auto const& file_system_initializer = mount_file.file_system_initializer();
+    if (!source_description) {
+        if (file_system_initializer.requires_open_file_description)
+            return ENOTSUP;
+        if (!file_system_initializer.create)
+            return ENOTSUP;
+        RefPtr<FileSystem> fs;
+        TRY(mount_file.mount_file_system_specific_data().with_exclusive([&](auto& mount_specific_data) -> ErrorOr<void> {
+            fs = TRY(file_system_initializer.create(mount_specific_data->bytes()));
+            return {};
+        }));
+        VERIFY(fs);
+        TRY(fs->initialize());
+        TRY(add_file_system_to_mount_table(*fs, mount_point, flags));
+        return {};
+    }
+
+    // NOTE: Although it might be OK to support creating filesystems
+    // without providing an actual file descriptor to their create() method
+    // because the caller of this function actually supplied a valid file descriptor,
+    // this will only make things complicated in the future, so we should block
+    // this kind of behavior.
+    if (!file_system_initializer.requires_open_file_description)
+        return ENOTSUP;
+
+    if (file_system_initializer.requires_block_device && !source_description->file().is_block_device())
+        return ENOTBLK;
+    if (file_system_initializer.requires_seekable_file && !source_description->file().is_seekable()) {
+        dbgln("mount: this is not a seekable file");
+        return ENODEV;
+    }
+
+    // NOTE: If there's an associated file description with the filesystem, we could
+    // try to first find it from the VirtualFileSystem filesystem list and if it was not found,
+    // then create it and add it.
+    VERIFY(file_system_initializer.create_with_fd);
+    return m_file_backed_file_systems_list.with_exclusive([&](auto& list) -> ErrorOr<void> {
+        RefPtr<FileSystem> fs;
+        for (auto& node : list) {
+            if ((&node.file_description() == source_description) || (&node.file() == &source_description->file())) {
+                fs = node;
+                break;
+            }
+        }
+        if (!fs) {
+            TRY(mount_file.mount_file_system_specific_data().with_exclusive([&](auto& mount_specific_data) -> ErrorOr<void> {
+                fs = TRY(file_system_initializer.create_with_fd(*source_description, mount_specific_data->bytes()));
+                return {};
+            }));
+            TRY(fs->initialize());
+        }
+        TRY(add_file_system_to_mount_table(*fs, mount_point, flags));
+        list.append(static_cast<FileBackedFileSystem&>(*fs));
+        return {};
+    });
+}
+
 ErrorOr<void> VirtualFileSystem::bind_mount(Custody& source, Custody& mount_point, int flags)
 {
    auto new_mount = TRY(adopt_nonnull_own_or_enomem(new (nothrow) Mount(source.inode(), mount_point, flags)));
@ -163,41 +266,42 @@ ErrorOr<void> VirtualFileSystem::unmount(Custody& mountpoint_custody)
    auto custody_path = TRY(mountpoint_custody.try_serialize_absolute_path());
    dbgln("VirtualFileSystem: unmount called with inode {} on mountpoint {}", guest_inode.identifier(), custody_path->view());

-    return m_mounts.with([&](auto& mounts) -> ErrorOr<void> {
-        for (auto& mount : mounts) {
-            if (&mount.guest() != &guest_inode)
-                continue;
-            auto mountpoint_path = TRY(mount.absolute_path());
-            if (custody_path->view() != mountpoint_path->view())
-                continue;
-            NonnullRefPtr<FileSystem> fs = mount.guest_fs();
-            TRY(fs->prepare_to_unmount());
-            fs->mounted_count({}).with([&](auto& mounted_count) {
-                VERIFY(mounted_count > 0);
-                if (mounted_count == 1) {
-                    dbgln("VirtualFileSystem: Unmounting file system {} for the last time...", fs->fsid());
-                    m_file_systems_list.with([&](auto& list) {
-                        list.remove(*fs);
-                    });
-                    if (fs->is_file_backed()) {
-                        dbgln("VirtualFileSystem: Unmounting file backed file system {} for the last time...", fs->fsid());
-                        auto& file_backed_fs = static_cast<FileBackedFileSystem&>(*fs);
-                        m_file_backed_file_systems_list.with([&](auto& list) {
-                            list.remove(file_backed_fs);
+    return m_file_backed_file_systems_list.with_exclusive([&](auto& file_backed_fs_list) -> ErrorOr<void> {
+        TRY(m_mounts.with([&](auto& mounts) -> ErrorOr<void> {
+            for (auto& mount : mounts) {
+                if (&mount.guest() != &guest_inode)
+                    continue;
+                auto mountpoint_path = TRY(mount.absolute_path());
+                if (custody_path->view() != mountpoint_path->view())
+                    continue;
+                NonnullRefPtr<FileSystem> fs = mount.guest_fs();
+                TRY(fs->prepare_to_unmount());
+                fs->mounted_count({}).with([&](auto& mounted_count) {
+                    VERIFY(mounted_count > 0);
+                    if (mounted_count == 1) {
+                        dbgln("VirtualFileSystem: Unmounting file system {} for the last time...", fs->fsid());
+                        m_file_systems_list.with([&](auto& list) {
+                            list.remove(*fs);
                        });
+                        if (fs->is_file_backed()) {
+                            dbgln("VirtualFileSystem: Unmounting file backed file system {} for the last time...", fs->fsid());
+                            auto& file_backed_fs = static_cast<FileBackedFileSystem&>(*fs);
+                            file_backed_fs_list.remove(file_backed_fs);
+                        }
+                    } else {
+                        mounted_count--;
                    }
-                } else {
-                    mounted_count--;
-                }
-            });
-            dbgln("VirtualFileSystem: Unmounting file system {}...", fs->fsid());
-            mount.m_vfs_list_node.remove();
-            // Note: This is balanced by a `new` statement that is happening in various places before inserting the Mount object to the list.
-            delete &mount;
-            return {};
-        }
-        dbgln("VirtualFileSystem: Nothing mounted on inode {}", guest_inode.identifier());
-        return ENODEV;
+                });
+                dbgln("VirtualFileSystem: Unmounting file system {}...", fs->fsid());
+                mount.m_vfs_list_node.remove();
+                // NOTE: This is balanced by a `new` statement that is happening in various places before inserting the Mount object to the list.
+                delete &mount;
+                return {};
+            }
+            dbgln("VirtualFileSystem: Nothing mounted on inode {}", guest_inode.identifier());
+            return ENODEV;
+        }));
+        return {};
    });
 }

@ -219,7 +323,7 @@ ErrorOr<void> VirtualFileSystem::mount_root(FileSystem& fs)
    if (fs.is_file_backed()) {
        auto pseudo_path = TRY(static_cast<FileBackedFileSystem&>(fs).file_description().pseudo_path());
        dmesgln("VirtualFileSystem: mounted root({}) from {} ({})", fs.fsid(), fs.class_name(), pseudo_path);
-        m_file_backed_file_systems_list.with([&](auto& list) {
+        m_file_backed_file_systems_list.with_exclusive([&](auto& list) {
            list.append(static_cast<FileBackedFileSystem&>(fs));
        });
    } else {
@ -339,28 +443,6 @@ ErrorOr<InodeMetadata> VirtualFileSystem::lookup_metadata(Credentials const& cre
    return custody->inode().metadata();
 }

-ErrorOr<NonnullRefPtr<FileBackedFileSystem>> VirtualFileSystem::find_already_existing_or_create_file_backed_file_system(OpenFileDescription& description, Function<ErrorOr<NonnullRefPtr<FileSystem>>(OpenFileDescription&)> callback)
-{
-    return TRY(m_file_backed_file_systems_list.with([&](auto& list) -> ErrorOr<NonnullRefPtr<FileBackedFileSystem>> {
-        for (auto& node : list) {
-            if (&node.file_description() == &description) {
-                return node;
-            }
-            if (&node.file() == &description.file()) {
-                return node;
-            }
-        }
-        auto fs = TRY(callback(description));
-
-        // The created FileSystem is only added to the file_systems_lists
-        // when the FS has been successfully initialized and mounted
-        // (in VirtualFileSystem::mount()). This prevents file systems which
-        // fail to initialize or mount from existing in the list when the
-        // FileSystem is destroyed after failure.
-        return static_ptr_cast<FileBackedFileSystem>(fs);
-    }));
-}
-
 ErrorOr<NonnullRefPtr<OpenFileDescription>> VirtualFileSystem::open(Credentials const& credentials, StringView path, int options, mode_t mode, Custody& base, Optional<UidAndGid> owner)
 {
    return open(Process::current(), credentials, path, options, mode, base, owner);
--- a/Kernel/FileSystem/VirtualFileSystem.h
+++ b/Kernel/FileSystem/VirtualFileSystem.h
@ -14,11 +14,14 @@
 #include <AK/RefPtr.h>
 #include <Kernel/FileSystem/FileBackedFileSystem.h>
 #include <Kernel/FileSystem/FileSystem.h>
+#include <Kernel/FileSystem/Initializer.h>
 #include <Kernel/FileSystem/InodeIdentifier.h>
 #include <Kernel/FileSystem/InodeMetadata.h>
 #include <Kernel/FileSystem/Mount.h>
+#include <Kernel/FileSystem/MountFile.h>
 #include <Kernel/FileSystem/UnveilNode.h>
 #include <Kernel/Forward.h>
+#include <Kernel/Locking/MutexProtected.h>
 #include <Kernel/Locking/SpinlockProtected.h>

 namespace Kernel {
@ -49,11 +52,13 @@ public:
    static void initialize();
    static VirtualFileSystem& the();

+    static ErrorOr<FileSystemInitializer const*> find_filesystem_type_initializer(StringView fs_type);
+
    VirtualFileSystem();
    ~VirtualFileSystem();

    ErrorOr<void> mount_root(FileSystem&);
-    ErrorOr<void> mount(FileSystem&, Custody& mount_point, int flags);
+    ErrorOr<void> mount(MountFile&, OpenFileDescription*, Custody& mount_point, int flags);
    ErrorOr<void> bind_mount(Custody& source, Custody& mount_point, int flags);
    ErrorOr<void> remount(Custody& mount_point, int new_flags);
    ErrorOr<void> unmount(Custody& mount_point);
@ -82,8 +87,6 @@ public:

    ErrorOr<void> for_each_mount(Function<ErrorOr<void>(Mount const&)>) const;

-    ErrorOr<NonnullRefPtr<FileBackedFileSystem>> find_already_existing_or_create_file_backed_file_system(OpenFileDescription& description, Function<ErrorOr<NonnullRefPtr<FileSystem>>(OpenFileDescription&)> callback);
-
    InodeIdentifier root_inode_id() const;

    void sync_filesystems();
@ -105,6 +108,8 @@ private:
    ErrorOr<void> validate_path_against_process_veil(Custody const& path, int options);
    ErrorOr<void> validate_path_against_process_veil(StringView path, int options);

+    ErrorOr<void> add_file_system_to_mount_table(FileSystem& file_system, Custody& mount_point, int flags);
+
    bool is_vfs_root(InodeIdentifier) const;

    ErrorOr<void> traverse_directory_inode(Inode&, Function<ErrorOr<void>(FileSystem::DirectoryEntryView const&)>);
@ -120,7 +125,14 @@ private:
    SpinlockProtected<RefPtr<Custody>, LockRank::None> m_root_custody {};

    SpinlockProtected<IntrusiveList<&Mount::m_vfs_list_node>, LockRank::None> m_mounts {};
-    SpinlockProtected<IntrusiveList<&FileBackedFileSystem::m_file_backed_file_system_node>, LockRank::None> m_file_backed_file_systems_list {};
+
+    // NOTE: The FileBackedFileSystem list is protected by a mutex because we need to scan it
+    // to search for existing filesystems for already used block devices and therefore when doing
+    // that we could fail to find a filesystem so we need to create a new filesystem which might
+    // need to do disk access (i.e. taking Mutexes in other places) and then register that new filesystem
+    // in this list, to avoid TOCTOU bugs.
+    MutexProtected<IntrusiveList<&FileBackedFileSystem::m_file_backed_file_system_node>> m_file_backed_file_systems_list {};
+
    SpinlockProtected<IntrusiveList<&FileSystem::m_file_system_node>, LockRank::FileSystem> m_file_systems_list {};
 };

--- a/Kernel/Forward.h
+++ b/Kernel/Forward.h
@ -31,6 +31,7 @@ class IPv4Socket;
 class Inode;
 class InodeIdentifier;
 class InodeWatcher;
+class MountFile;
 class Jail;
 class KBuffer;
 class KString;
--- a/Kernel/Syscalls/mount.cpp
+++ b/Kernel/Syscalls/mount.cpp
@ -1,120 +1,68 @@
 /*
 * Copyright (c) 2018-2020, Andreas Kling <kling@serenityos.org>
+ * Copyright (c) 2022, Liav A. <liavalb@hotmail.co.il>
 *
 * SPDX-License-Identifier: BSD-2-Clause
 */

 #include <Kernel/FileSystem/Custody.h>
-#include <Kernel/FileSystem/DevPtsFS/FileSystem.h>
-#include <Kernel/FileSystem/Ext2FS/FileSystem.h>
-#include <Kernel/FileSystem/FATFS/FileSystem.h>
-#include <Kernel/FileSystem/ISO9660FS/FileSystem.h>
-#include <Kernel/FileSystem/Plan9FS/FileSystem.h>
-#include <Kernel/FileSystem/ProcFS/FileSystem.h>
-#include <Kernel/FileSystem/RAMFS/FileSystem.h>
-#include <Kernel/FileSystem/SysFS/FileSystem.h>
+#include <Kernel/FileSystem/MountFile.h>
 #include <Kernel/FileSystem/VirtualFileSystem.h>
 #include <Kernel/Tasks/Process.h>

 namespace Kernel {

-struct FileSystemInitializer {
-    StringView short_name;
-    StringView name;
-    bool requires_open_file_description { false };
-    bool requires_block_device { false };
-    bool requires_seekable_file { false };
-    ErrorOr<NonnullRefPtr<FileSystem>> (*create_with_fd)(OpenFileDescription&) = nullptr;
-    ErrorOr<NonnullRefPtr<FileSystem>> (*create)(void) = nullptr;
-};
-
-static constexpr FileSystemInitializer s_initializers[] = {
-    { "proc"sv, "ProcFS"sv, false, false, false, {}, ProcFS::try_create },
-    { "devpts"sv, "DevPtsFS"sv, false, false, false, {}, DevPtsFS::try_create },
-    { "sys"sv, "SysFS"sv, false, false, false, {}, SysFS::try_create },
-    { "ram"sv, "RAMFS"sv, false, false, false, {}, RAMFS::try_create },
-    { "ext2"sv, "Ext2FS"sv, true, true, true, Ext2FS::try_create, {} },
-    { "9p"sv, "Plan9FS"sv, true, true, true, Plan9FS::try_create, {} },
-    { "iso9660"sv, "ISO9660FS"sv, true, true, true, ISO9660FS::try_create, {} },
-    { "fat"sv, "FATFS"sv, true, true, true, FATFS::try_create, {} },
-};
-
-static ErrorOr<NonnullRefPtr<FileSystem>> find_or_create_filesystem_instance(StringView fs_type, OpenFileDescription* possible_description)
+ErrorOr<FlatPtr> Process::sys$fsopen(Userspace<Syscall::SC_fsopen_params const*> user_params)
 {
-    for (auto& initializer_entry : s_initializers) {
-        if (fs_type != initializer_entry.short_name && fs_type != initializer_entry.name)
-            continue;
-        if (!initializer_entry.requires_open_file_description) {
-            VERIFY(initializer_entry.create);
-            NonnullRefPtr<FileSystem> fs = TRY(initializer_entry.create());
-            return fs;
-        }
-        // Note: If there's an associated file description with the filesystem, we could
-        // try to first find it from the VirtualFileSystem filesystem list and if it was not found,
-        // then create it and add it.
-        VERIFY(initializer_entry.create_with_fd);
-        if (!possible_description)
-            return EBADF;
-        OpenFileDescription& description = *possible_description;
-
-        if (initializer_entry.requires_block_device && !description.file().is_block_device())
-            return ENOTBLK;
-        if (initializer_entry.requires_seekable_file && !description.file().is_seekable()) {
-            dbgln("mount: this is not a seekable file");
-            return ENODEV;
-        }
-        return TRY(VirtualFileSystem::the().find_already_existing_or_create_file_backed_file_system(description, initializer_entry.create_with_fd));
-    }
-    return ENODEV;
-}
-
-ErrorOr<FlatPtr> Process::sys$mount(Userspace<Syscall::SC_mount_params const*> user_params)
-{
-    VERIFY_PROCESS_BIG_LOCK_ACQUIRED(this);
-    TRY(require_no_promises());
+    VERIFY_NO_PROCESS_BIG_LOCK(this);
+    TRY(require_promise(Pledge::mount));
    auto credentials = this->credentials();
    if (!credentials->is_superuser())
-        return EPERM;
+        return Error::from_errno(EPERM);
+    auto params = TRY(copy_typed_from_user(user_params));
+    auto fs_type_string = TRY(try_copy_kstring_from_user(params.fs_type));
+
+    // NOTE: If some userspace program uses MS_REMOUNT, return EINVAL to indicate that we never want this
+    // flag to appear in the mount table...
+    if (params.flags & MS_REMOUNT || params.flags & MS_BIND)
+        return Error::from_errno(EINVAL);
+
+    auto const* fs_type_initializer = TRY(VirtualFileSystem::find_filesystem_type_initializer(fs_type_string->view()));
+    VERIFY(fs_type_initializer);
+    auto mount_file = TRY(MountFile::create(*fs_type_initializer, params.flags));
+    auto description = TRY(OpenFileDescription::try_create(move(mount_file)));
+    return m_fds.with_exclusive([&](auto& fds) -> ErrorOr<FlatPtr> {
+        auto new_fd = TRY(fds.allocate());
+        fds[new_fd.fd].set(move(description), FD_CLOEXEC);
+        return new_fd.fd;
+    });
+}
+
+ErrorOr<FlatPtr> Process::sys$fsmount(Userspace<Syscall::SC_fsmount_params const*> user_params)
+{
+    VERIFY_NO_PROCESS_BIG_LOCK(this);
+    TRY(require_promise(Pledge::mount));
+    auto credentials = this->credentials();
+    if (!credentials->is_superuser())
+        return Error::from_errno(EPERM);

    auto params = TRY(copy_typed_from_user(user_params));
-    if (params.flags & MS_REMOUNT)
-        return EINVAL;
-    if (params.flags & MS_BIND)
-        return EINVAL;
+    auto mount_description = TRY(open_file_description(params.mount_fd));
+    if (!mount_description->is_mount_file())
+        return Error::from_errno(EINVAL);

-    auto source_fd = params.source_fd;
+    RefPtr<OpenFileDescription> source_description = TRY(open_file_description_ignoring_negative(params.source_fd));
    auto target = TRY(try_copy_kstring_from_user(params.target));
-    auto fs_type_string = TRY(try_copy_kstring_from_user(params.fs_type));
-    auto fs_type = fs_type_string->view();
-
-    auto description_or_error = open_file_description(source_fd);
-    if (!description_or_error.is_error())
-        dbgln("mount {}: source fd {} @ {}", fs_type, source_fd, target);
-    else
-        dbgln("mount {} @ {}", fs_type, target);
-
    auto target_custody = TRY(VirtualFileSystem::the().resolve_path(credentials, target->view(), current_directory()));
-
-    RefPtr<FileSystem> fs;
-
-    if (!description_or_error.is_error()) {
-        auto description = description_or_error.release_value();
-        fs = TRY(find_or_create_filesystem_instance(fs_type, description.ptr()));
-        auto source_pseudo_path = TRY(description->pseudo_path());
-        dbgln("mount: attempting to mount {} on {}", source_pseudo_path, target);
-    } else {
-        fs = TRY(find_or_create_filesystem_instance(fs_type, {}));
-    }
-
-    TRY(fs->initialize());
-    TRY(VirtualFileSystem::the().mount(*fs, target_custody, params.flags));
+    auto flags = mount_description->mount_file()->mount_flags();
+    TRY(VirtualFileSystem::the().mount(*mount_description->mount_file(), source_description.ptr(), target_custody, flags));
    return 0;
 }

 ErrorOr<FlatPtr> Process::sys$remount(Userspace<Syscall::SC_remount_params const*> user_params)
 {
    VERIFY_NO_PROCESS_BIG_LOCK(this);
-    TRY(require_no_promises());
+    TRY(require_promise(Pledge::mount));
    auto credentials = this->credentials();
    if (!credentials->is_superuser())
        return EPERM;
@ -134,7 +82,7 @@ ErrorOr<FlatPtr> Process::sys$remount(Userspace<Syscall::SC_remount_params const
 ErrorOr<FlatPtr> Process::sys$bindmount(Userspace<Syscall::SC_bindmount_params const*> user_params)
 {
    VERIFY_NO_PROCESS_BIG_LOCK(this);
-    TRY(require_no_promises());
+    TRY(require_promise(Pledge::mount));
    auto credentials = this->credentials();
    if (!credentials->is_superuser())
        return EPERM;
@ -166,7 +114,7 @@ ErrorOr<FlatPtr> Process::sys$umount(Userspace<char const*> user_mountpoint, siz
    if (!credentials->is_superuser())
        return EPERM;

-    TRY(require_no_promises());
+    TRY(require_promise(Pledge::mount));

    auto mountpoint = TRY(get_syscall_path_argument(user_mountpoint, mountpoint_length));
    auto custody = TRY(VirtualFileSystem::the().resolve_path(credentials, mountpoint->view(), current_directory()));
--- a/Kernel/Tasks/Process.h
+++ b/Kernel/Tasks/Process.h
@ -70,6 +70,7 @@ UnixDateTime kgettimeofday();
    __ENUMERATE_PLEDGE_PROMISE(map_fixed) \
    __ENUMERATE_PLEDGE_PROMISE(getkeymap) \
    __ENUMERATE_PLEDGE_PROMISE(jail)      \
+    __ENUMERATE_PLEDGE_PROMISE(mount)     \
    __ENUMERATE_PLEDGE_PROMISE(no_error)

 enum class Pledge : u32 {
@ -395,7 +396,8 @@ public:
    ErrorOr<FlatPtr> sys$unlink(int dirfd, Userspace<char const*> pathname, size_t path_length, int flags);
    ErrorOr<FlatPtr> sys$symlink(Userspace<Syscall::SC_symlink_params const*>);
    ErrorOr<FlatPtr> sys$rmdir(Userspace<char const*> pathname, size_t path_length);
-    ErrorOr<FlatPtr> sys$mount(Userspace<Syscall::SC_mount_params const*>);
+    ErrorOr<FlatPtr> sys$fsmount(Userspace<Syscall::SC_fsmount_params const*>);
+    ErrorOr<FlatPtr> sys$fsopen(Userspace<Syscall::SC_fsopen_params const*>);
    ErrorOr<FlatPtr> sys$umount(Userspace<char const*> mountpoint, size_t mountpoint_length);
    ErrorOr<FlatPtr> sys$chmod(Userspace<Syscall::SC_chmod_params const*>);
    ErrorOr<FlatPtr> sys$fchmod(int fd, mode_t);
@ -826,11 +828,25 @@ public:
        return m_fds.with_shared([fd](auto& fds) { return fds.open_file_description(fd); });
    }

+    ErrorOr<RefPtr<OpenFileDescription>> open_file_description_ignoring_negative(int fd)
+    {
+        if (fd < 0)
+            return nullptr;
+        return open_file_description(fd);
+    }
+
    ErrorOr<NonnullRefPtr<OpenFileDescription>> open_file_description(int fd) const
    {
        return m_fds.with_shared([fd](auto& fds) { return fds.open_file_description(fd); });
    }

+    ErrorOr<RefPtr<OpenFileDescription>> open_file_description_ignoring_negative(int fd) const
+    {
+        if (fd < 0)
+            return nullptr;
+        return open_file_description(fd);
+    }
+
    ErrorOr<ScopedDescriptionAllocation> allocate_fd()
    {
        return m_fds.with_exclusive([](auto& fds) { return fds.allocate(); });
--- a/Userland/DevTools/UserspaceEmulator/Emulator.h
+++ b/Userland/DevTools/UserspaceEmulator/Emulator.h
@ -147,6 +147,7 @@ private:
    int virt$anon_create(size_t, int);
    int virt$beep();
    int virt$bind(int sockfd, FlatPtr address, socklen_t address_length);
+    u32 virt$bindmount(u32 params_addr);
    int virt$chdir(FlatPtr, size_t);
    int virt$chmod(FlatPtr);
    int virt$chown(FlatPtr);
@ -167,6 +168,8 @@ private:
    int virt$fchown(int, uid_t, gid_t);
    u32 virt$fcntl(int fd, int, u32);
    int virt$fork();
+    u32 virt$fsopen(u32);
+    u32 virt$fsmount(u32);
    int virt$fstat(int, FlatPtr);
    int virt$ftruncate(int fd, FlatPtr length_addr);
    int virt$futex(FlatPtr);
@ -199,7 +202,6 @@ private:
    u32 virt$madvise(FlatPtr, size_t, int);
    int virt$mkdir(FlatPtr path, size_t path_length, mode_t mode);
    u32 virt$mmap(u32);
-    u32 virt$mount(u32);
    u32 virt$mprotect(FlatPtr, size_t, int);
    FlatPtr virt$mremap(FlatPtr);
    int virt$annotate_mapping(FlatPtr);
@ -219,6 +221,7 @@ private:
    int virt$recvfd(int, int);
    int virt$recvmsg(int sockfd, FlatPtr msg_addr, int flags);
    int virt$rename(FlatPtr address);
+    u32 virt$remount(u32);
    int virt$rmdir(FlatPtr path, size_t path_length);
    int virt$scheduler_get_parameters(FlatPtr);
    int virt$scheduler_set_parameters(FlatPtr);
--- a/Userland/DevTools/UserspaceEmulator/Emulator_syscalls.cpp
+++ b/Userland/DevTools/UserspaceEmulator/Emulator_syscalls.cpp
@ -51,6 +51,8 @@ u32 Emulator::virt_syscall(u32 function, u32 arg1, u32 arg2, u32 arg3)
        return virt$beep();
    case SC_bind:
        return virt$bind(arg1, arg2, arg3);
+    case SC_bindmount:
+        return virt$bindmount(arg1);
    case SC_chdir:
        return virt$chdir(arg1, arg2);
    case SC_chmod:
@ -94,6 +96,10 @@ u32 Emulator::virt_syscall(u32 function, u32 arg1, u32 arg2, u32 arg3)
        return virt$fork();
    case SC_fstat:
        return virt$fstat(arg1, arg2);
+    case SC_fsmount:
+        return virt$fsopen(arg1);
+    case SC_fsopen:
+        return virt$fsopen(arg1);
    case SC_ftruncate:
        return virt$ftruncate(arg1, arg2);
    case SC_futex:
@ -158,8 +164,6 @@ u32 Emulator::virt_syscall(u32 function, u32 arg1, u32 arg2, u32 arg3)
        return virt$mkdir(arg1, arg2, arg3);
    case SC_mmap:
        return virt$mmap(arg1);
-    case SC_mount:
-        return virt$mount(arg1);
    case SC_mprotect:
        return virt$mprotect(arg1, arg2, arg3);
    case SC_mremap:
@ -196,6 +200,8 @@ u32 Emulator::virt_syscall(u32 function, u32 arg1, u32 arg2, u32 arg3)
        return virt$recvmsg(arg1, arg2, arg3);
    case SC_rename:
        return virt$rename(arg1);
+    case SC_remount:
+        return virt$remount(arg1);
    case SC_rmdir:
        return virt$rmdir(arg1, arg2);
    case SC_scheduler_get_parameters:
@ -906,18 +912,48 @@ FlatPtr Emulator::virt$mremap(FlatPtr params_addr)
    return -EINVAL;
 }

-u32 Emulator::virt$mount(u32 params_addr)
+u32 Emulator::virt$bindmount(u32 params_addr)
 {
-    Syscall::SC_mount_params params;
+    Syscall::SC_bindmount_params params;
    mmu().copy_from_vm(&params, params_addr, sizeof(params));
    auto target = mmu().copy_buffer_from_vm((FlatPtr)params.target.characters, params.target.length);
-    auto fs_path = mmu().copy_buffer_from_vm((FlatPtr)params.fs_type.characters, params.fs_type.length);
-    params.fs_type.characters = (char*)fs_path.data();
-    params.fs_type.length = fs_path.size();
    params.target.characters = (char*)target.data();
    params.target.length = target.size();

-    return syscall(SC_mount, &params);
+    return syscall(SC_bindmount, &params);
+}
+
+u32 Emulator::virt$remount(u32 params_addr)
+{
+    Syscall::SC_remount_params params;
+    mmu().copy_from_vm(&params, params_addr, sizeof(params));
+    auto target = mmu().copy_buffer_from_vm((FlatPtr)params.target.characters, params.target.length);
+    params.target.characters = (char*)target.data();
+    params.target.length = target.size();
+
+    return syscall(SC_remount, &params);
+}
+
+u32 Emulator::virt$fsopen(u32 params_addr)
+{
+    Syscall::SC_fsopen_params params;
+    mmu().copy_from_vm(&params, params_addr, sizeof(params));
+    auto fs_type = mmu().copy_buffer_from_vm((FlatPtr)params.fs_type.characters, params.fs_type.length);
+    params.fs_type.characters = (char*)fs_type.data();
+    params.fs_type.length = fs_type.size();
+
+    return syscall(SC_fsopen, &params);
+}
+
+u32 Emulator::virt$fsmount(u32 params_addr)
+{
+    Syscall::SC_fsmount_params params;
+    mmu().copy_from_vm(&params, params_addr, sizeof(params));
+    auto target = mmu().copy_buffer_from_vm((FlatPtr)params.target.characters, params.target.length);
+    params.target.characters = (char*)target.data();
+    params.target.length = target.size();
+
+    return syscall(SC_fsmount, &params);
 }

 u32 Emulator::virt$gettid()
--- a/Userland/Libraries/LibC/unistd.cpp
+++ b/Userland/Libraries/LibC/unistd.cpp
@ -940,23 +940,65 @@ int fsync(int fd)
    __RETURN_WITH_ERRNO(rc, rc, -1);
 }

-int mount(int source_fd, char const* target, char const* fs_type, int flags)
+int fsopen(char const* fs_type, int flags)
 {
-    if (!target || !fs_type) {
+    if (!fs_type) {
        errno = EFAULT;
        return -1;
    }

-    Syscall::SC_mount_params params {
-        { target, strlen(target) },
+    Syscall::SC_fsopen_params params {
        { fs_type, strlen(fs_type) },
-        source_fd,
-        flags
+        flags,
    };
-    int rc = syscall(SC_mount, &params);
+    int rc = syscall(SC_fsopen, &params);
    __RETURN_WITH_ERRNO(rc, rc, -1);
 }

+int fsmount(int mount_fd, int source_fd, char const* target)
+{
+    if (!target) {
+        errno = EFAULT;
+        return -1;
+    }
+
+    Syscall::SC_fsmount_params params {
+        mount_fd,
+        { target, strlen(target) },
+        source_fd,
+    };
+    int rc = syscall(SC_fsmount, &params);
+    __RETURN_WITH_ERRNO(rc, rc, -1);
+}
+
+int bindmount(int source_fd, char const* target, int flags)
+{
+    if (!target) {
+        errno = EFAULT;
+        return -1;
+    }
+
+    Syscall::SC_bindmount_params params {
+        { target, strlen(target) },
+        source_fd,
+        flags,
+    };
+    int rc = syscall(SC_bindmount, &params);
+    __RETURN_WITH_ERRNO(rc, rc, -1);
+}
+
+int mount(int source_fd, char const* target, char const* fs_type, int flags)
+{
+    if (flags & MS_BIND)
+        return bindmount(source_fd, target, flags);
+
+    int mount_fd = fsopen(fs_type, flags);
+    if (mount_fd < 0)
+        return -1;
+
+    return fsmount(mount_fd, source_fd, target);
+}
+
 int umount(char const* mountpoint)
 {
    int rc = syscall(SC_umount, mountpoint, strlen(mountpoint));
--- a/Userland/Libraries/LibC/unistd.h
+++ b/Userland/Libraries/LibC/unistd.h
@ -117,6 +117,9 @@ int fchown(int fd, uid_t, gid_t);
 int fchownat(int fd, char const* pathname, uid_t uid, gid_t gid, int flags);
 int ftruncate(int fd, off_t length);
 int truncate(char const* path, off_t length);
+int fsopen(char const* fs_type, int flags);
+int fsmount(int mount_fd, int source_fd, char const* target);
+int bindmount(int source_fd, char const* target, int flags);
 int mount(int source_fd, char const* target, char const* fs_type, int flags);
 int umount(char const* mountpoint);
 int pledge(char const* promises, char const* execpromises);
--- a/Userland/Libraries/LibCore/System.cpp
+++ b/Userland/Libraries/LibCore/System.cpp
@ -263,14 +263,43 @@ ErrorOr<void> mount(int source_fd, StringView target, StringView fs_type, int fl
    if (target.is_null() || fs_type.is_null())
        return Error::from_errno(EFAULT);

-    Syscall::SC_mount_params params {
-        { target.characters_without_null_termination(), target.length() },
+    if (flags & MS_REMOUNT) {
+        TRY(remount(target, flags));
+        return {};
+    }
+    if (flags & MS_BIND) {
+        TRY(bindmount(source_fd, target, flags));
+        return {};
+    }
+    int mount_fd = TRY(fsopen(fs_type, flags));
+    return fsmount(mount_fd, source_fd, target);
+}
+
+ErrorOr<int> fsopen(StringView fs_type, int flags)
+{
+    if (fs_type.is_null())
+        return Error::from_errno(EFAULT);
+
+    Syscall::SC_fsopen_params params {
        { fs_type.characters_without_null_termination(), fs_type.length() },
-        source_fd,
-        flags
+        flags,
    };
-    int rc = syscall(SC_mount, &params);
-    HANDLE_SYSCALL_RETURN_VALUE("mount", rc, {});
+    int rc = syscall(SC_fsopen, &params);
+    HANDLE_SYSCALL_RETURN_VALUE("fsopen", rc, rc);
+}
+
+ErrorOr<void> fsmount(int mount_fd, int source_fd, StringView target)
+{
+    if (target.is_null())
+        return Error::from_errno(EFAULT);
+
+    Syscall::SC_fsmount_params params {
+        mount_fd,
+        { target.characters_without_null_termination(), target.length() },
+        source_fd,
+    };
+    int rc = syscall(SC_fsmount, &params);
+    HANDLE_SYSCALL_RETURN_VALUE("fsmount", rc, {});
 }

 ErrorOr<void> umount(StringView mount_point)
--- a/Userland/Libraries/LibCore/System.h
+++ b/Userland/Libraries/LibCore/System.h
@ -60,6 +60,8 @@ ErrorOr<int> recvfd(int sockfd, int options);
 ErrorOr<void> ptrace_peekbuf(pid_t tid, void const* tracee_addr, Bytes destination_buf);
 ErrorOr<void> mount(int source_fd, StringView target, StringView fs_type, int flags);
 ErrorOr<void> bindmount(int source_fd, StringView target, int flags);
+ErrorOr<int> fsopen(StringView fs_type, int flags);
+ErrorOr<void> fsmount(int mount_fd, int source_fd, StringView target_path);
 ErrorOr<void> remount(StringView target, int flags);
 ErrorOr<void> umount(StringView mount_point);
 ErrorOr<long> ptrace(int request, pid_t tid, void* address, void* data);