On Thu, Jan 28, 2021 at 05:41:21PM -0800, Andy Lutomirski wrote: > task_user_regset_view() is fundamentally broken, but it's ABI for > PTRACE_GETREGSET and PTRACE_SETREGSET. > > We shouldn't be using it for PTRACE_GETREGS or PTRACE_SETREGS,
No "We" etc pls. > though. A native 64-bit ptrace() call and an x32 ptrace() call > should use the 64-bit regset views, and a 32-bit ptrace() call > (native or compat) should use the 32-bit regset. > task_user_regset_view() almost does this except that it will > malfunction if a ptracer is itself ptraced and the outer ptracer > modifies CS on entry to a ptrace() syscall. Is that the reason why task_user_regset_view() is fundamentally broken? It is somewhat unclear what exactly is broken. > Hopefully that has > never happened. (The compat ptrace() code already hardcoded the > 32-bit regset, so this patch has no effect on that path.) > > Fix it and deobfuscate the code by hardcoding the 64-bit view in the > x32 ptrace() and selecting the view based on the kernel config in > the native ptrace(). > > Signed-off-by: Andy Lutomirski <[email protected]> > --- > > Every time I look at ptrace, it grosses me out. This makes it slightly > more comprehensible. > > arch/x86/kernel/ptrace.c | 37 +++++++++++++++++++++++++++++-------- > 1 file changed, 29 insertions(+), 8 deletions(-) Well, did you run the gdb testsuite on this and a bunch of other tests we have? I don't want us to break gdb or something else using ptrace() in some sublte manner and then waste a bunch of time and energy chasing it, like the DR6 thing earlier this week. > +/* > + * This is used by PTRACE_GETREGSET and PTRACE_SETREGSET to decide which > + * regset format to use based on the register state of the tracee. > + * This makes no sense whatsoever, but there appears to be existing user > + * code that relies on it. ... because? It should use the native regset with which the kernel is built? Please explain yourself Lutomirski! :-))) Thx. -- Regards/Gruss, Boris. https://people.kernel.org/tglx/notes-about-netiquette
