[dvyukov:dvyukov-rseq-pkey 3/3] kernel/rseq.c:406:9: error: unknown type name 'pkey_reg_t'

[kernel test robot <lkp@intel.com>]

tree:   https://github.com/dvyukov/linux dvyukov-rseq-pkey
head:   b991732d23294a961eed6a4a2758c16f0430de9c
commit: b991732d23294a961eed6a4a2758c16f0430de9c [3/3] rseq: Make rseq work with protection keys
config: powerpc-allnoconfig (https://download.01.org/0day-ci/archive/20250208/202502082017.0GixUctp-lkp@intel.com/config)
compiler: powerpc-linux-gcc (GCC) 14.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250208/202502082017.0GixUctp-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202502082017.0GixUctp-lkp@intel.com/

All errors (new ones prefixed by >>):

   kernel/rseq.c: In function '__rseq_handle_notify_resume':
>> kernel/rseq.c:406:9: error: unknown type name 'pkey_reg_t'
     406 |         pkey_reg_t saved;
         |         ^~~~~~~~~~
>> kernel/rseq.c:426:17: error: implicit declaration of function 'write_pkey_reg' [-Wimplicit-function-declaration]
     426 |                 write_pkey_reg(saved);
         |                 ^~~~~~~~~~~~~~
>> kernel/rseq.c:455:25: error: implicit declaration of function 'switch_to_permissive_pkey_reg' [-Wimplicit-function-declaration]
     455 |                 saved = switch_to_permissive_pkey_reg();
         |                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~


vim +/pkey_reg_t +406 kernel/rseq.c

   390	
   391	/*
   392	 * This resume handler must always be executed between any of:
   393	 * - preemption,
   394	 * - signal delivery,
   395	 * and return to user-space.
   396	 *
   397	 * This is how we can ensure that the entire rseq critical section
   398	 * will issue the commit instruction only if executed atomically with
   399	 * respect to other threads scheduled on the same CPU, and with respect
   400	 * to signal handlers.
   401	 */
   402	void __rseq_handle_notify_resume(struct ksignal *ksig, struct pt_regs *regs)
   403	{
   404		struct task_struct *t = current;
   405		int ret, sig;
 > 406		pkey_reg_t saved;
   407		bool switched_pkey_reg = false;
   408	
   409		if (unlikely(t->flags & PF_EXITING))
   410			return;
   411	
   412	retry:
   413		/*
   414		 * regs is NULL if and only if the caller is in a syscall path.  Skip
   415		 * fixup and leave rseq_cs as is so that rseq_sycall() will detect and
   416		 * kill a misbehaving userspace on debug kernels.
   417		 */
   418		if (regs) {
   419			ret = rseq_ip_fixup(regs);
   420			if (unlikely(ret < 0))
   421				goto error;
   422		}
   423		if (unlikely(rseq_update_cpu_node_id(t)))
   424			goto error;
   425		if (switched_pkey_reg)
 > 426			write_pkey_reg(saved);
   427		return;
   428	
   429	error:
   430		/*
   431		 * If the application registers rseq, and ever switches to another
   432		 * pkey protection (such that the rseq becomes inaccessible), then
   433		 * any context switch will cause failure here attemping to read/write
   434		 * struct rseq and/or rseq_cs. Since context switches are
   435		 * asynchronous and are outside of the application control
   436		 * (not part of the restricted code scope), we temporarily switch
   437		 * to premissive pkey register to read/write rseq/rseq_cs,
   438		 * similarly to signal delivery.
   439		 *
   440		 * We don't bother to check if the failure really happened due to
   441		 * pkeys or not, since it does not matter (performance-wise and
   442		 * otherwise).
   443		 *
   444		 * If the restricted code installs rseq_cs in inaccessible to it
   445		 * due to pkeys page, we still let this function read the rseq_cs.
   446		 * It's unclear what benefits the resticted code gets by doing this
   447		 * (it probably already hijacked control flow at this point), and
   448		 * presumably any sane sandbox should prohibit restricted code
   449		 * from accessing struct rseq, and this is still better than
   450		 * terminating the app unconditionally (it always has a choice
   451		 * of not using rseq and pkeys together).
   452		 */
   453		if (!switched_pkey_reg) {
   454			switched_pkey_reg = true;
 > 455			saved = switch_to_permissive_pkey_reg();
   456			goto retry;
   457		} else {
   458			write_pkey_reg(saved);
   459		}
   460		sig = ksig ? ksig->sig : 0;
   461		force_sigsegv(sig);
   462	}
   463	

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki