target/linux/generic/backport-5.15/795-v6.3-06-r8152-Add-__GFP_NOWARN-to-big-allocations.patch

   1 From 5cc33f139e11b893ff6dc60d8a0ae865a65521ac Mon Sep 17 00:00:00 2001
   2 From: Douglas Anderson <dianders@chromium.org>
   3 Date: Thu, 6 Apr 2023 17:14:26 -0700
   4 Subject: [PATCH] r8152: Add __GFP_NOWARN to big allocations
   5
   6 When memory is a little tight on my system, it's pretty easy to see
   7 warnings that look like this.
   8
   9   ksoftirqd/0: page allocation failure: order:3, mode:0x40a20(GFP_ATOMIC|__GFP_COMP), nodemask=(null),cpuset=/,mems_allowed=0
  10   ...
  11   Call trace:
  12    dump_backtrace+0x0/0x1e8
  13    show_stack+0x20/0x2c
  14    dump_stack_lvl+0x60/0x78
  15    dump_stack+0x18/0x38
  16    warn_alloc+0x104/0x174
  17    __alloc_pages+0x588/0x67c
  18    alloc_rx_agg+0xa0/0x190 [r8152 ...]
  19    r8152_poll+0x270/0x760 [r8152 ...]
  20    __napi_poll+0x44/0x1ec
  21    net_rx_action+0x100/0x300
  22    __do_softirq+0xec/0x38c
  23    run_ksoftirqd+0x38/0xec
  24    smpboot_thread_fn+0xb8/0x248
  25    kthread+0x134/0x154
  26    ret_from_fork+0x10/0x20
  27
  28 On a fragmented system it's normal that order 3 allocations will
  29 sometimes fail, especially atomic ones. The driver handles these
  30 failures fine and the WARN just creates spam in the logs for this
  31 case. The __GFP_NOWARN flag is exactly for this situation, so add it
  32 to the allocation.
  33
  34 NOTE: my testing is on a 5.15 system, but there should be no reason
  35 that this would be fundamentally different on a mainline kernel.
  36
  37 Signed-off-by: Douglas Anderson <dianders@chromium.org>
  38 Acked-by: Hayes Wang <hayeswang@realtek.com>
  39 Link: https://lore.kernel.org/r/20230406171411.1.I84dbef45786af440fd269b71e9436a96a8e7a152@changeid
  40 Signed-off-by: Jakub Kicinski <kuba@kernel.org>
  41 ---
  42  drivers/net/usb/r8152.c | 2 +-
  43  1 file changed, 1 insertion(+), 1 deletion(-)
  44
  45 --- a/drivers/net/usb/r8152.c
  46 +++ b/drivers/net/usb/r8152.c
  47 @@ -1944,7 +1944,7 @@ static struct rx_agg *alloc_rx_agg(struc
  48         if (!rx_agg)
  49                 return NULL;
  50
  51 -       rx_agg->page = alloc_pages(mflags | __GFP_COMP, order);
  52 +       rx_agg->page = alloc_pages(mflags | __GFP_COMP | __GFP_NOWARN, order);
  53         if (!rx_agg->page)
  54                 goto free_rx;
  55