On Tue, Dec 5, 2017 at 5:36 AM, Russell King - ARM Linux <li...@armlinux.org.uk> wrote: > On Tue, Dec 05, 2017 at 01:30:11PM +0000, Phil Elwell wrote: >> This was my initial explanation: >> >> 1. Data which is marked __ro_after_init is initially writeable. >> >> 2. The ro_perms data covers kernel text, read-only data and __ro_after_init >> data. >> >> 3. set_kernel_text_rw marks everything in ro_perms as writeable. >> >> 4. set_kernel_text_ro marks everything in ro_perms as read-only, including >> the __ro_after_init data. >> >> 5. Using the function tracing code involves code modification, resulting in >> calls to >> __ftrace_modify_code and set_kernel_text_ro. >> >> 6. Therefore if function tracing is enabled before kernel_init has completed >> then the __ro_after_init >> data is made read-only prematurely. > > My question still stands, but let me rephrase. Do we need > set_kernel_text_*() to touch the read-only data?
We don't _need_ to, but they're all contiguous, so the ro_perms array used by set_kernel_text_*() is actually only a single entry: static struct section_perm ro_perms[] = { /* Make kernel code and rodata RX (set RO). */ { .name = "text/rodata RO", .start = (unsigned long)_stext, .end = (unsigned long)__init_begin, ... -Kees -- Kees Cook Pixel Security