Re: [PATCH v6 10/18] sh/tlb: Convert SH to generic mmu_gather

[John Paul Adrian Glaubitz]

Hi Rob!

On 12/5/19 8:24 PM, Rob Landley wrote:
> No, but most people running this kind of hardware tend not to upgrade to current
> kernels on a regular basis.
> 
> The j-core guys tested the 5.3 release. I can't find an email about 5.4 so I
> dunno if that's been tested yet?
> 
> I just tested yesterday's git and it works fine with
> http://lkml.iu.edu/hypermail/linux/kernel/1912.0/01554.html installed, modulo it
> _still_ has the suprious stack dump shortly before calling init, which I've
> complained about on linux-sh and off for a year now?
> 
> ------------[ cut here ]------------
> WARNING: CPU: 0 PID: 1 at mm/slub.c:2451 kmem_cache_free_bulk+0x2c2/0x37c
> 
> CPU: 0 PID: 1 Comm: swapper Not tainted 5.4.0 #1
> PC is at kmem_cache_free_bulk+0x2c2/0x37c
> PR is at kmem_cache_alloc_bulk+0x36/0x1a0
> PC  : 8c0a6fae SP  : 8f829e9c SR  : 400080f0
> TEA : c0001240
> R0  : 8c0a6de4 R1  : 00000100 R2  : 00000100 R3  : 00000000
> R4  : 8f8020a0 R5  : 00000dc0 R6  : 8c01d66c R7  : 8fff5180
> R8  : 8c011a00 R9  : 8fff5180 R10 : 8c01d66c R11 : 80000000
> R12 : 00007fff R13 : 00000dc0 R14 : 8f8020a0
> MACH: 0000017a MACL: 0ae4849d GBR : 00000000 PR  : 8c0a709e
> 
> Call trace:
>  [<(ptrval)>] copy_process+0x218/0x1094
>  [<(ptrval)>] copy_process+0x7ba/0x1094
>  [<(ptrval)>] kmem_cache_alloc_bulk+0x36/0x1a0
>  [<(ptrval)>] restore_sigcontext+0x94/0x1b0
>  [<(ptrval)>] restore_sigcontext+0x70/0x1b0
>  [<(ptrval)>] copy_process+0x218/0x1094
>  [<(ptrval)>] sysfs_slab_add+0x106/0x354
>  [<(ptrval)>] restore_sigcontext+0x70/0x1b0
>  [<(ptrval)>] copy_process+0x218/0x1094
>  [<(ptrval)>] copy_process+0x218/0x1094
>  [<(ptrval)>] fprop_fraction_single+0x38/0xa4
>  [<(ptrval)>] pipe_read+0x7a/0x23c
>  [<(ptrval)>] restore_sigcontext+0x70/0x1b0
>  [<(ptrval)>] restore_sigcontext+0x94/0x1b0
>  [<(ptrval)>] alloc_pipe_info+0x162/0x1c8
>  [<(ptrval)>] restore_sigcontext+0x94/0x1b0
>  [<(ptrval)>] restore_sigcontext+0x70/0x1b0
>  [<(ptrval)>] handle_bad_irq+0x154/0x188
>  [<(ptrval)>] raw6_exit_net+0x0/0x14
>  [<(ptrval)>] prepare_stack+0xe4/0x2fc
>  [<(ptrval)>] sys_sched_get_priority_min+0x18/0x28
>  [<(ptrval)>] ndisc_net_exit+0x4/0x24
> 
> ---[ end trace 6ce4eefeb577b078 ]---
> 
> But it's cosmetic...

This is fixed by the following patch [1]:

sh: Fix unneeded constructor in page table allocation

The pgd kmem_cache allocation both specified __GFP_ZERO and had a
constructor which makes no sense.  Remove __GFP_ZERO and zero the user
parts of the pgd explicitly.

Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
Signed-off-by: Rich Felker <dalias@xxxxxxxx>

Adrian

> [1] https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?id=73c348f31b63d28d176ed290eb1aa2a648f3e51e

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaubitz@xxxxxxxxxx
`. `'   Freie Universitaet Berlin - glaubitz@xxxxxxxxxxxxxxxxxxx
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913