From 119e9db233ad0f4cbd4cfb9f9385a7bca378b37d Mon Sep 17 00:00:00 2001 From: Zeng Chi Date: Mon, 21 Sep 2026 18:24:42 +0800 Subject: KVM: Don't pre-reserve xarray entries when storing empty/NULL attributes Skip the xarray reservation loop when clearing all memory attributes, as storing NULL only erases the entry and never needs to allocate, so no reservation (and no cleanup of a failed one) is required in that case. Suggested-by: Sean Christopherson Cc: David Ballesteros Signed-off-by: Zeng Chi Link: https://patch.msgid.link/20260921102442.1232375-1-zeng_chi911@163.com [sean: split to separate patch] Signed-off-by: Sean Christopherson --- fs/kernel_read_file.c | 180 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 180 insertions(+) create mode 100644 fs/kernel_read_file.c (limited to 'fs/kernel_read_file.c') diff --git a/fs/kernel_read_file.c b/fs/kernel_read_file.c new file mode 100644 index 000000000..9c2ba9240 --- /dev/null +++ b/fs/kernel_read_file.c @@ -0,0 +1,180 @@ +// SPDX-License-Identifier: GPL-2.0-only +#include +#include +#include +#include +#include + +/** + * kernel_read_file() - read file contents into a kernel buffer + * + * @file: file to read from + * @offset: where to start reading from (see below). + * @buf: pointer to a "void *" buffer for reading into (if + * *@buf is NULL, a buffer will be allocated, and + * @buf_size will be ignored) + * @buf_size: size of buf, if already allocated. If @buf not + * allocated, this is the largest size to allocate. + * @file_size: if non-NULL, the full size of @file will be + * written here. + * @id: the kernel_read_file_id identifying the type of + * file contents being read (for LSMs to examine) + * + * @offset must be 0 unless both @buf and @file_size are non-NULL + * (i.e. the caller must be expecting to read partial file contents + * via an already-allocated @buf, in at most @buf_size chunks, and + * will be able to determine when the entire file was read by + * checking @file_size). This isn't a recommended way to read a + * file, though, since it is possible that the contents might + * change between calls to kernel_read_file(). + * + * Returns number of bytes read (no single read will be bigger + * than SSIZE_MAX), or negative on error. + * + */ +ssize_t kernel_read_file(struct file *file, loff_t offset, void **buf, + size_t buf_size, size_t *file_size, + enum kernel_read_file_id id) +{ + loff_t i_size, pos; + ssize_t copied; + void *allocated = NULL; + bool whole_file; + int ret; + + if (offset != 0 && (!*buf || !file_size)) + return -EINVAL; + + if (!S_ISREG(file_inode(file)->i_mode)) + return -EINVAL; + + ret = deny_write_access(file); + if (ret) + return ret; + + i_size = i_size_read(file_inode(file)); + if (i_size <= 0) { + ret = -EINVAL; + goto out; + } + /* The file is too big for sane activities. */ + if (i_size > SSIZE_MAX) { + ret = -EFBIG; + goto out; + } + /* The entire file cannot be read in one buffer. */ + if (!file_size && offset == 0 && i_size > buf_size) { + ret = -EFBIG; + goto out; + } + + whole_file = (offset == 0 && i_size <= buf_size); + ret = security_kernel_read_file(file, id, whole_file); + if (ret) + goto out; + + if (file_size) + *file_size = i_size; + + if (!*buf) + *buf = allocated = vmalloc(i_size); + if (!*buf) { + ret = -ENOMEM; + goto out; + } + + pos = offset; + copied = 0; + while (copied < buf_size) { + ssize_t bytes; + size_t wanted = min_t(size_t, buf_size - copied, + i_size - pos); + + bytes = kernel_read(file, *buf + copied, wanted, &pos); + if (bytes < 0) { + ret = bytes; + goto out_free; + } + + if (bytes == 0) + break; + copied += bytes; + } + + if (whole_file) { + if (pos != i_size) { + ret = -EIO; + goto out_free; + } + + ret = security_kernel_post_read_file(file, *buf, i_size, id); + } + +out_free: + if (ret < 0) { + if (allocated) { + vfree(*buf); + *buf = NULL; + } + } + +out: + allow_write_access(file); + return ret == 0 ? copied : ret; +} +EXPORT_SYMBOL_GPL(kernel_read_file); + +ssize_t kernel_read_file_from_path(const char *path, loff_t offset, void **buf, + size_t buf_size, size_t *file_size, + enum kernel_read_file_id id) +{ + struct file *file; + ssize_t ret; + + if (!path || !*path) + return -EINVAL; + + file = filp_open(path, O_RDONLY, 0); + if (IS_ERR(file)) + return PTR_ERR(file); + + ret = kernel_read_file(file, offset, buf, buf_size, file_size, id); + fput(file); + return ret; +} +EXPORT_SYMBOL_GPL(kernel_read_file_from_path); + +ssize_t kernel_read_file_from_path_initns(const char *path, loff_t offset, + void **buf, size_t buf_size, + size_t *file_size, + enum kernel_read_file_id id) +{ + struct file *file; + ssize_t ret; + + if (!path || !*path) + return -EINVAL; + + scoped_with_init_fs() + file = filp_open(path, O_RDONLY, 0); + if (IS_ERR(file)) + return PTR_ERR(file); + + ret = kernel_read_file(file, offset, buf, buf_size, file_size, id); + fput(file); + return ret; +} +EXPORT_SYMBOL_GPL(kernel_read_file_from_path_initns); + +ssize_t kernel_read_file_from_fd(int fd, loff_t offset, void **buf, + size_t buf_size, size_t *file_size, + enum kernel_read_file_id id) +{ + CLASS(fd, f)(fd); + + if (fd_empty(f) || !(fd_file(f)->f_mode & FMODE_READ)) + return -EBADF; + + return kernel_read_file(fd_file(f), offset, buf, buf_size, file_size, id); +} +EXPORT_SYMBOL_GPL(kernel_read_file_from_fd); -- cgit v1.3.1