On 2020-06-05 22:19, [email protected] wrote:
> The patch titled
> Subject: exec: open code copy_string_kernel
> has been removed from the -mm tree. Its filename was
> exec-open-code-copy_string_kernel.patch
>
> This patch was dropped because it was merged into mainline or a subsystem tree
>
> ------------------------------------------------------
> From: Christoph Hellwig <[email protected]>
> Subject: exec: open code copy_string_kernel
>
> Currently copy_string_kernel is just a wrapper around copy_strings that
> simplifies the calling conventions and uses set_fs to allow passing a
> kernel pointer. But due to the fact the we only need to handle a single
> kernel argument pointer, the logic can be sigificantly simplified while
> getting rid of the set_fs.
>
> Link: http://lkml.kernel.org/r/[email protected]
> Signed-off-by: Christoph Hellwig <[email protected]>
> Cc: Alexander Viro <[email protected]>
> Signed-off-by: Andrew Morton <[email protected]>
> ---
>
> fs/exec.c | 45 +++++++++++++++++++++++++++++++++++----------
> 1 file changed, 35 insertions(+), 10 deletions(-)
>
> --- a/fs/exec.c~exec-open-code-copy_string_kernel
> +++ a/fs/exec.c
> @@ -592,17 +592,42 @@ out:
> */
> int copy_string_kernel(const char *arg, struct linux_binprm *bprm)
> {
> - int r;
> - mm_segment_t oldfs = get_fs();
> - struct user_arg_ptr argv = {
> - .ptr.native = (const char __user *const __user *)&arg,
> - };
> -
> - set_fs(KERNEL_DS);
> - r = copy_strings(1, argv, bprm);
> - set_fs(oldfs);
> + int len = strnlen(arg, MAX_ARG_STRLEN) + 1 /* terminating NUL */;
> + unsigned long pos = bprm->p;
>
> - return r;
> + if (len == 0)
> + return -EFAULT;
Just a quick question, how can len ever be 0 here when len was set to
strnlen() + 1? Should the test be different?
The old version (i.e. copy_strings()) seems to return -EFAULT when
strnlen() returns 0.
Vegard
> + if (!valid_arg_len(bprm, len))
> + return -E2BIG;
> +
> + /* We're going to work our way backwards. */
> + arg += len;
> + bprm->p -= len;
> + if (IS_ENABLED(CONFIG_MMU) && bprm->p < bprm->argmin)
> + return -E2BIG;
> +
> + while (len > 0) {
> + unsigned int bytes_to_copy = min_t(unsigned int, len,
> + min_not_zero(offset_in_page(pos), PAGE_SIZE));
> + struct page *page;
> + char *kaddr;
> +
> + pos -= bytes_to_copy;
> + arg -= bytes_to_copy;
> + len -= bytes_to_copy;
> +
> + page = get_arg_page(bprm, pos, 1);
> + if (!page)
> + return -E2BIG;
> + kaddr = kmap_atomic(page);
> + flush_arg_page(bprm, pos & PAGE_MASK, page);
> + memcpy(kaddr + offset_in_page(pos), arg, bytes_to_copy);
> + flush_kernel_dcache_page(page);
> + kunmap_atomic(kaddr);
> + put_arg_page(page);
> + }
> +
> + return 0;
> }
> EXPORT_SYMBOL(copy_string_kernel);
>
On Tue, Jun 16, 2020 at 03:14:44PM +0200, Vegard Nossum wrote:
>
> On 2020-06-05 22:19, [email protected] wrote:
>> The patch titled
>> Subject: exec: open code copy_string_kernel
>> has been removed from the -mm tree. Its filename was
>> exec-open-code-copy_string_kernel.patch
>>
>> This patch was dropped because it was merged into mainline or a subsystem tree
>>
>> ------------------------------------------------------
>> From: Christoph Hellwig <[email protected]>
>> Subject: exec: open code copy_string_kernel
>>
>> Currently copy_string_kernel is just a wrapper around copy_strings that
>> simplifies the calling conventions and uses set_fs to allow passing a
>> kernel pointer. But due to the fact the we only need to handle a single
>> kernel argument pointer, the logic can be sigificantly simplified while
>> getting rid of the set_fs.
>>
>> Link: http://lkml.kernel.org/r/[email protected]
>> Signed-off-by: Christoph Hellwig <[email protected]>
>> Cc: Alexander Viro <[email protected]>
>> Signed-off-by: Andrew Morton <[email protected]>
>> ---
>>
>> fs/exec.c | 45 +++++++++++++++++++++++++++++++++++----------
>> 1 file changed, 35 insertions(+), 10 deletions(-)
>>
>> --- a/fs/exec.c~exec-open-code-copy_string_kernel
>> +++ a/fs/exec.c
>> @@ -592,17 +592,42 @@ out:
>> */
>> int copy_string_kernel(const char *arg, struct linux_binprm *bprm)
>> {
>> - int r;
>> - mm_segment_t oldfs = get_fs();
>> - struct user_arg_ptr argv = {
>> - .ptr.native = (const char __user *const __user *)&arg,
>> - };
>> -
>> - set_fs(KERNEL_DS);
>> - r = copy_strings(1, argv, bprm);
>> - set_fs(oldfs);
>> + int len = strnlen(arg, MAX_ARG_STRLEN) + 1 /* terminating NUL */;
>> + unsigned long pos = bprm->p;
>> - return r;
>> + if (len == 0)
>> + return -EFAULT;
>
> Just a quick question, how can len ever be 0 here when len was set to
> strnlen() + 1? Should the test be different?
>
> The old version (i.e. copy_strings()) seems to return -EFAULT when
> strnlen() returns 0.
So, the nasty part here is that strnlen_user has different semantics
from strnlen:
- strlen excludes the terminating null byte and never returns error
codes
- strnlen_user includes the terminating null byte, and a 0 return
means it could not access the user address (a condition that can't
happen for strlen).
Now with that back to your original question: I think then len == 0
check can just be removed without replacement.