mm/hotplug: silence a lockdep splat with printk()
authorQian Cai <cai@lca.pw>
Fri, 31 Jan 2020 06:14:57 +0000 (22:14 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 31 Jan 2020 18:30:39 +0000 (10:30 -0800)
commit4a55c0474a92d5c418bcbbe122368de0910aeac2
tree79c66943cc7e37de9c2e15d4ecfbdc4259d7ca41
parentbd5c2344f9eb1ebf7ff2501ddb13d83151939780
mm/hotplug: silence a lockdep splat with printk()

It is not that hard to trigger lockdep splats by calling printk from
under zone->lock.  Most of them are false positives caused by lock
chains introduced early in the boot process and they do not cause any
real problems (although most of the early boot lock dependencies could
happen after boot as well).  There are some console drivers which do
allocate from the printk context as well and those should be fixed.  In
any case, false positives are not that trivial to workaround and it is
far from optimal to lose lockdep functionality for something that is a
non-issue.

So change has_unmovable_pages() so that it no longer calls dump_page()
itself - instead it returns a "struct page *" of the unmovable page back
to the caller so that in the case of a has_unmovable_pages() failure,
the caller can call dump_page() after releasing zone->lock.  Also, make
dump_page() is able to report a CMA page as well, so the reason string
from has_unmovable_pages() can be removed.

Even though has_unmovable_pages doesn't hold any reference to the
returned page this should be reasonably safe for the purpose of
reporting the page (dump_page) because it cannot be hotremoved in the
context of memory unplug.  The state of the page might change but that
is the case even with the existing code as zone->lock only plays role
for free pages.

While at it, remove a similar but unnecessary debug-only printk() as
well.  A sample of one of those lockdep splats is,

  WARNING: possible circular locking dependency detected
  ------------------------------------------------------
  test.sh/8653 is trying to acquire lock:
  ffffffff865a4460 (console_owner){-.-.}, at:
  console_unlock+0x207/0x750

  but task is already holding lock:
  ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
  __offline_isolated_pages+0x179/0x3e0

  which lock already depends on the new lock.

  the existing dependency chain (in reverse order) is:

  -> #3 (&(&zone->lock)->rlock){-.-.}:
         __lock_acquire+0x5b3/0xb40
         lock_acquire+0x126/0x280
         _raw_spin_lock+0x2f/0x40
         rmqueue_bulk.constprop.21+0xb6/0x1160
         get_page_from_freelist+0x898/0x22c0
         __alloc_pages_nodemask+0x2f3/0x1cd0
         alloc_pages_current+0x9c/0x110
         allocate_slab+0x4c6/0x19c0
         new_slab+0x46/0x70
         ___slab_alloc+0x58b/0x960
         __slab_alloc+0x43/0x70
         __kmalloc+0x3ad/0x4b0
         __tty_buffer_request_room+0x100/0x250
         tty_insert_flip_string_fixed_flag+0x67/0x110
         pty_write+0xa2/0xf0
         n_tty_write+0x36b/0x7b0
         tty_write+0x284/0x4c0
         __vfs_write+0x50/0xa0
         vfs_write+0x105/0x290
         redirected_tty_write+0x6a/0xc0
         do_iter_write+0x248/0x2a0
         vfs_writev+0x106/0x1e0
         do_writev+0xd4/0x180
         __x64_sys_writev+0x45/0x50
         do_syscall_64+0xcc/0x76c
         entry_SYSCALL_64_after_hwframe+0x49/0xbe

  -> #2 (&(&port->lock)->rlock){-.-.}:
         __lock_acquire+0x5b3/0xb40
         lock_acquire+0x126/0x280
         _raw_spin_lock_irqsave+0x3a/0x50
         tty_port_tty_get+0x20/0x60
         tty_port_default_wakeup+0xf/0x30
         tty_port_tty_wakeup+0x39/0x40
         uart_write_wakeup+0x2a/0x40
         serial8250_tx_chars+0x22e/0x440
         serial8250_handle_irq.part.8+0x14a/0x170
         serial8250_default_handle_irq+0x5c/0x90
         serial8250_interrupt+0xa6/0x130
         __handle_irq_event_percpu+0x78/0x4f0
         handle_irq_event_percpu+0x70/0x100
         handle_irq_event+0x5a/0x8b
         handle_edge_irq+0x117/0x370
         do_IRQ+0x9e/0x1e0
         ret_from_intr+0x0/0x2a
         cpuidle_enter_state+0x156/0x8e0
         cpuidle_enter+0x41/0x70
         call_cpuidle+0x5e/0x90
         do_idle+0x333/0x370
         cpu_startup_entry+0x1d/0x1f
         start_secondary+0x290/0x330
         secondary_startup_64+0xb6/0xc0

  -> #1 (&port_lock_key){-.-.}:
         __lock_acquire+0x5b3/0xb40
         lock_acquire+0x126/0x280
         _raw_spin_lock_irqsave+0x3a/0x50
         serial8250_console_write+0x3e4/0x450
         univ8250_console_write+0x4b/0x60
         console_unlock+0x501/0x750
         vprintk_emit+0x10d/0x340
         vprintk_default+0x1f/0x30
         vprintk_func+0x44/0xd4
         printk+0x9f/0xc5

  -> #0 (console_owner){-.-.}:
         check_prev_add+0x107/0xea0
         validate_chain+0x8fc/0x1200
         __lock_acquire+0x5b3/0xb40
         lock_acquire+0x126/0x280
         console_unlock+0x269/0x750
         vprintk_emit+0x10d/0x340
         vprintk_default+0x1f/0x30
         vprintk_func+0x44/0xd4
         printk+0x9f/0xc5
         __offline_isolated_pages.cold.52+0x2f/0x30a
         offline_isolated_pages_cb+0x17/0x30
         walk_system_ram_range+0xda/0x160
         __offline_pages+0x79c/0xa10
         offline_pages+0x11/0x20
         memory_subsys_offline+0x7e/0xc0
         device_offline+0xd5/0x110
         state_store+0xc6/0xe0
         dev_attr_store+0x3f/0x60
         sysfs_kf_write+0x89/0xb0
         kernfs_fop_write+0x188/0x240
         __vfs_write+0x50/0xa0
         vfs_write+0x105/0x290
         ksys_write+0xc6/0x160
         __x64_sys_write+0x43/0x50
         do_syscall_64+0xcc/0x76c
         entry_SYSCALL_64_after_hwframe+0x49/0xbe

  other info that might help us debug this:

  Chain exists of:
    console_owner --> &(&port->lock)->rlock --> &(&zone->lock)->rlock

   Possible unsafe locking scenario:

         CPU0                    CPU1
         ----                    ----
    lock(&(&zone->lock)->rlock);
                                 lock(&(&port->lock)->rlock);
                                 lock(&(&zone->lock)->rlock);
    lock(console_owner);

   *** DEADLOCK ***

  9 locks held by test.sh/8653:
   #0: ffff88839ba7d408 (sb_writers#4){.+.+}, at:
  vfs_write+0x25f/0x290
   #1: ffff888277618880 (&of->mutex){+.+.}, at:
  kernfs_fop_write+0x128/0x240
   #2: ffff8898131fc218 (kn->count#115){.+.+}, at:
  kernfs_fop_write+0x138/0x240
   #3: ffffffff86962a80 (device_hotplug_lock){+.+.}, at:
  lock_device_hotplug_sysfs+0x16/0x50
   #4: ffff8884374f4990 (&dev->mutex){....}, at:
  device_offline+0x70/0x110
   #5: ffffffff86515250 (cpu_hotplug_lock.rw_sem){++++}, at:
  __offline_pages+0xbf/0xa10
   #6: ffffffff867405f0 (mem_hotplug_lock.rw_sem){++++}, at:
  percpu_down_write+0x87/0x2f0
   #7: ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
  __offline_isolated_pages+0x179/0x3e0
   #8: ffffffff865a4920 (console_lock){+.+.}, at:
  vprintk_emit+0x100/0x340

  stack backtrace:
  Hardware name: HPE ProLiant DL560 Gen10/ProLiant DL560 Gen10,
  BIOS U34 05/21/2019
  Call Trace:
   dump_stack+0x86/0xca
   print_circular_bug.cold.31+0x243/0x26e
   check_noncircular+0x29e/0x2e0
   check_prev_add+0x107/0xea0
   validate_chain+0x8fc/0x1200
   __lock_acquire+0x5b3/0xb40
   lock_acquire+0x126/0x280
   console_unlock+0x269/0x750
   vprintk_emit+0x10d/0x340
   vprintk_default+0x1f/0x30
   vprintk_func+0x44/0xd4
   printk+0x9f/0xc5
   __offline_isolated_pages.cold.52+0x2f/0x30a
   offline_isolated_pages_cb+0x17/0x30
   walk_system_ram_range+0xda/0x160
   __offline_pages+0x79c/0xa10
   offline_pages+0x11/0x20
   memory_subsys_offline+0x7e/0xc0
   device_offline+0xd5/0x110
   state_store+0xc6/0xe0
   dev_attr_store+0x3f/0x60
   sysfs_kf_write+0x89/0xb0
   kernfs_fop_write+0x188/0x240
   __vfs_write+0x50/0xa0
   vfs_write+0x105/0x290
   ksys_write+0xc6/0x160
   __x64_sys_write+0x43/0x50
   do_syscall_64+0xcc/0x76c
   entry_SYSCALL_64_after_hwframe+0x49/0xbe

Link: http://lkml.kernel.org/r/20200117181200.20299-1-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/page-isolation.h
mm/debug.c
mm/page_alloc.c
mm/page_isolation.c