Re: [PATCH 4/8] pipe: fix limit checking in pipe_set_size()
flat view
From: Vegard Nossum <hidden>
Date: 2016-08-19 08:31:46
Also in:
lkml
On 08/19/2016 07:25 AM, Michael Kerrisk (man-pages) wrote:
The limit checking in pipe_set_size() (used by fcntl(F_SETPIPE_SZ)) has the following problems:
[...]
quoted hunk ↗ jump to hunk
@@ -1030,6 +1030,7 @@ static long pipe_set_size(struct pipe_inode_info *pipe, unsigned long arg) { struct pipe_buffer *bufs; unsigned int size, nr_pages; + long ret = 0; size = round_pipe_size(arg); nr_pages = size >> PAGE_SHIFT;@@ -1037,13 +1038,26 @@ static long pipe_set_size(struct pipe_inode_info *pipe, unsigned long arg) if (!nr_pages) return -EINVAL; - if (!capable(CAP_SYS_RESOURCE) && size > pipe_max_size) - return -EPERM; + account_pipe_buffers(pipe->user, pipe->buffers, nr_pages); - if ((too_many_pipe_buffers_hard(pipe->user) || - too_many_pipe_buffers_soft(pipe->user)) && - !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) - return -EPERM; + /* + * If trying to increase the pipe capacity, check that an + * unprivileged user is not trying to exceed various limits. + * (Decreasing the pipe capacity is always permitted, even + * if the user is currently over a limit.) + */ + if (nr_pages > pipe->buffers) { + if (!capable(CAP_SYS_RESOURCE) && size > pipe_max_size) { + ret = -EPERM; + goto out_revert_acct; + } else if ((too_many_pipe_buffers_hard(pipe->user) || + too_many_pipe_buffers_soft(pipe->user)) && + !capable(CAP_SYS_RESOURCE) && + !capable(CAP_SYS_ADMIN)) { + ret = -EPERM; + goto out_revert_acct; + } + }
I'm slightly worried about not checking arg/nr_pages before we pass it on to account_pipe_buffers(). The potential problem happens if the user passes a very large number which will overflow pipe->user->pipe_bufs. On 32-bit, sizeof(int) == sizeof(long), so if they pass arg = INT_MAX then round_pipe_size() returns INT_MAX. Although it's true that the accounting is done in terms of pages and not bytes, so you'd need on the order of (1 << 13) = 8192 processes hitting the limit at the same time in order to make it overflow, which seems a bit unlikely. (See https://lkml.org/lkml/2016/8/12/215 for another discussion on the limit checking) Is there any reason why we couldn't do the (size > pipe_max_size) check before calling account_pipe_buffers()? Vegard