tree: https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git sched/wip.umcg
head: 0bcd439a3202373f9497c08b30801fb81bb99528
commit: 806dfbbc75579b4a633ff79e93e42b7b4ec6d9c6 [3/12] entry,x86: Create common IRQ operations for exceptions
config: x86_64-allyesconfig
compiler: gcc-11 (Debian 11.3.0-3) 11.3.0
reproduce (this is a W=1 build):
# https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git/commit/?id=806dfbbc75579b4a633ff79e93e42b7b4ec6d9c6
git remote add peterz-queue https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git
git fetch --no-tags peterz-queue sched/wip.umcg
git checkout 806dfbbc75579b4a633ff79e93e42b7b4ec6d9c6
# save the config file
mkdir build_dir && cp config build_dir/.config
make W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash arch/x86/
If you fix the issue, kindly add following tag where applicable
Reported-by: kernel test robot <[email protected]>
All errors (new ones prefixed by >>):
arch/x86/kernel/traps.c: In function '__exc_virtualization_exception':
>> arch/x86/kernel/traps.c:1406:9: error: implicit declaration of function 'cond_local_irq_enable'; did you mean 'arch_local_irq_enable'? [-Werror=implicit-function-declaration]
1406 | cond_local_irq_enable(regs);
| ^~~~~~~~~~~~~~~~~~~~~
| arch_local_irq_enable
>> arch/x86/kernel/traps.c:1415:9: error: implicit declaration of function 'cond_local_irq_disable'; did you mean 'arch_local_irq_disable'? [-Werror=implicit-function-declaration]
1415 | cond_local_irq_disable(regs);
| ^~~~~~~~~~~~~~~~~~~~~~
| arch_local_irq_disable
cc1: some warnings being treated as errors
vim +1406 arch/x86/kernel/traps.c
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1351
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1352 /*
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1353 * Virtualization Exceptions (#VE) are delivered to TDX guests due to
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1354 * specific guest actions which may happen in either user space or the
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1355 * kernel:
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1356 *
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1357 * * Specific instructions (WBINVD, for example)
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1358 * * Specific MSR accesses
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1359 * * Specific CPUID leaf accesses
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1360 * * Access to specific guest physical addresses
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1361 *
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1362 * In the settings that Linux will run in, virtualization exceptions are
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1363 * never generated on accesses to normal, TD-private memory that has been
7dbde763162989 Kirill A. Shutemov 2022-04-06 1364 * accepted (by BIOS or with tdx_enc_status_changed()).
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1365 *
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1366 * Syscall entry code has a critical window where the kernel stack is not
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1367 * yet set up. Any exception in this window leads to hard to debug issues
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1368 * and can be exploited for privilege escalation. Exceptions in the NMI
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1369 * entry code also cause issues. Returning from the exception handler with
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1370 * IRET will re-enable NMIs and nested NMI will corrupt the NMI stack.
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1371 *
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1372 * For these reasons, the kernel avoids #VEs during the syscall gap and
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1373 * the NMI entry code. Entry code paths do not access TD-shared memory,
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1374 * MMIO regions, use #VE triggering MSRs, instructions, or CPUID leaves
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1375 * that might generate #VE. VMM can remove memory from TD at any point,
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1376 * but access to unaccepted (or missing) private memory leads to VM
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1377 * termination, not to #VE.
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1378 *
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1379 * Similarly to page faults and breakpoints, #VEs are allowed in NMI
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1380 * handlers once the kernel is ready to deal with nested NMIs.
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1381 *
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1382 * During #VE delivery, all interrupts, including NMIs, are blocked until
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1383 * TDGETVEINFO is called. It prevents #VE nesting until the kernel reads
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1384 * the VE info.
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1385 *
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1386 * If a guest kernel action which would normally cause a #VE occurs in
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1387 * the interrupt-disabled region before TDGETVEINFO, a #DF (fault
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1388 * exception) is delivered to the guest which will result in an oops.
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1389 *
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1390 * The entry code has been audited carefully for following these expectations.
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1391 * Changes in the entry code have to be audited for correctness vs. this
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1392 * aspect. Similarly to #PF, #VE in these places will expose kernel to
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1393 * privilege escalation or may lead to random crashes.
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1394 */
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1395 DEFINE_IDTENTRY(exc_virtualization_exception)
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1396 {
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1397 struct ve_info ve;
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1398
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1399 /*
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1400 * NMIs/Machine-checks/Interrupts will be in a disabled state
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1401 * till TDGETVEINFO TDCALL is executed. This ensures that VE
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1402 * info cannot be overwritten by a nested #VE.
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1403 */
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1404 tdx_get_ve_info(&ve);
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1405
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 @1406 cond_local_irq_enable(regs);
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1407
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1408 /*
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1409 * If tdx_handle_virt_exception() could not process
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1410 * it successfully, treat it as #GP(0) and handle it.
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1411 */
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1412 if (!tdx_handle_virt_exception(regs, &ve))
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1413 ve_raise_fault(regs, 0);
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1414
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 @1415 cond_local_irq_disable(regs);
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1416 }
9a22bf6debbf51 Kirill A. Shutemov 2022-04-06 1417
:::::: The code at line 1406 was first introduced by commit
:::::: 9a22bf6debbf5169f750af53c7f86eb4e3cd6712 x86/traps: Add #VE support for TDX guest
:::::: TO: Kirill A. Shutemov <[email protected]>
:::::: CC: Dave Hansen <[email protected]>
--
0-DAY CI Kernel Test Service
https://01.org/lkp