[LTP] [PATCH v2 2/2] kallsyms01: Utilize ksymbol table for unauthorized address access

[Li Wang]

Hi Cyril,

All comments make sense, patch v3 is coming.

On Thu, May 2, 2024 at 6:30 PM Cyril Hrubis <chrubis@suse.cz> wrote:

> Hi!
> > @@ -0,0 +1,141 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
>
> Any reason this is GPL-2.0-only and not 2.0-or-later?
>
> > +/*
> > + * Copyright (C) 2024 Red Hat, Inc.
> > + */
> > +
> > +/*\
> > + * [Description]
> > + *
> > + *  Utilize kernel's symbol table for unauthorized address access.
> > + *
> > + *  Access the system symbols with root permission to test whether it's
> > + *  possible to read and write the memory addresses of kernel-space
> > + *  from user-space. This helps in identifying potential vulnerabilities
> > + *  where user-space processes can inappropriately access kernel memory.
> > + *
> > + * Steps:
> > + *  1. Start a process that reads all symbols and their addresses from
> > + *     '/proc/kallsyms' and stores them in a linked list.
> > + *
> > + *  2. Attempt to write to each kernel address found in the linked list.
> > + *     The expectation is that each attempt will fail with a SIGSEGV
> > + *     (segmentation fault), indicating that the user-space process
> > + *     cannot write to kernel memory.
> > + *
> > + *  3. Handle each SIGSEGV using a signal handler that sets a flag and
> > + *     long jumps out of the faulting context.
> > + *
> > + *  4. If any write operation does not result in a SIGSEGV, log this as
> > + *     a potential security vulnerability.
> > + *
> > + *  5. Observe and log the behavior and any system responses to these
> > + *     unauthorized access attempts.
> > + *
> > + */
> > +
> > +#include <stdio.h>
> > +#include <stdlib.h>
> > +#include <assert.h>
> > +#include <unistd.h>
> > +#include <string.h>
> > +#include <setjmp.h>
> > +#include <signal.h>
> > +
> > +#include "tst_test.h"
> > +#include "tst_safe_stdio.h"
> > +
> > +struct kallsym {
> > +     unsigned long addr;
> > +     char type;
> > +     char name[128];
> > +};
> > +
> > +static struct kallsym *sym_table;
> > +static unsigned int nr_symbols;
> > +static sigjmp_buf jmpbuf;
> > +volatile sig_atomic_t segv_caught = 0;
> > +
> > +static void segv_handler(int sig)
> > +{
> > +     if (sig == SIGSEGV)
> > +             segv_caught++;
> > +     else
> > +             tst_res(TFAIL, "Unexpected signal %s", strsignal(sig));
> > +
> > +     siglongjmp(jmpbuf, 1);
> > +}
> > +
> > +static unsigned int read_kallsyms(struct kallsym *table, unsigned int
> table_size)
> > +{
> > +     char *line = NULL;
> > +     size_t len = 0;
> > +     unsigned int nr_syms = 0;
> > +     FILE *stream = SAFE_FOPEN("/proc/kallsyms", "r");
> > +
> > +     while (getline(&line, &len, stream) != -1) {
> > +
> > +             if (table && nr_syms < table_size) {
> > +                     sscanf(line, "%lx %c %s",
> > +                                     &table[nr_syms].addr,
> > +                                     &table[nr_syms].type,
> > +                                     table[nr_syms].name);
> > +             }
> > +
> > +             nr_syms++;
> > +     }
> > +
> > +     SAFE_FCLOSE(stream);
> > +
> > +     return nr_syms;
> > +}
> > +
> > +static void setup(void)
> > +{
> > +     nr_symbols = read_kallsyms(NULL, 0);
> > +     sym_table = SAFE_CALLOC(nr_symbols, sizeof(*sym_table));
> > +     unsigned int read_symbols = read_kallsyms(sym_table, nr_symbols);
> > +
> > +     if (nr_symbols != read_symbols)
> > +             tst_res(TWARN, "/proc/kallsyms changed size!?");
> > +}
> > +
> > +static void access_ksymbols_address(struct kallsym *table)
> > +{
> > +     tst_res(TDEBUG, "Access kernel addr: 0x%lx (%c) (%s)",
> > +                             table->addr, table->type, table->name);
> > +
> > +     if (sigsetjmp(jmpbuf, 1) == 0) {
> > +             *(volatile unsigned long *)table->addr = 0;
> > +
> > +             tst_res(TFAIL, "Successfully accessed kernel addr 0x%lx
> (%c) (%s)",
> > +                             table->addr, table->type, table->name);
> > +     }
> > +}
> > +
> > +static void test_access_kernel_address(void)
> > +{
> > +     struct sigaction sa;
> > +     memset(&sa, 0, sizeof(sa));
> > +     sa.sa_handler = segv_handler;
> > +     sigaction(SIGSEGV, &sa, NULL);
>
> We can move the sigaction to the test setup.
>
> > +     for (unsigned int i = 0; i < nr_symbols; i++)
> > +             access_ksymbols_address(&sym_table[i]);
>
> And we have to set the segv_caught to 0 before this loop, otherwise the
> test does not work with -i 2
>
> > +     if (segv_caught == (sig_atomic_t)nr_symbols)
> > +             tst_res(TPASS, "Caught %d times SIGSEGV in access ksymbols
> addr", segv_caught);
>
> And also in a case that we got wrong number on segfaults this does not
> produce any results. I guess that we need to do:
>
>         if (...)
>                 tst_res(TPASS, "...");
>         else
>                 tst_res(TFAIL, "Caught %d SIGSEGV expected %d", ...);
>
>
> > +}
> > +
> > +static void cleanup(void)
> > +{
> > +     if (sym_table)
> > +             free(sym_table);
> > +}
> > +
> > +static struct tst_test test = {
> > +     .needs_root = 1,
> > +     .setup = setup,
> > +     .cleanup = cleanup,
> > +     .max_runtime = 60,
> > +     .test_all = test_access_kernel_address,
> > +};
> > --
> > 2.40.1
> >
> >
> > --
> > Mailing list info: https://lists.linux.it/listinfo/ltp
>
> --
> Cyril Hrubis
> chrubis@suse.cz
>
>

-- 
Regards,
Li Wang