Re: 5.4.143-rt63 fscache_cookie_hash initialisation

Re: 5.4.143-rt63 fscache_cookie_hash initialisation

[Sebastian Andrzej Siewior]

On 2021-09-06 09:55:59 [+0200], Gregor Beck wrote:
Your proposed change is

|index 5508d92e3f8f..cba2a226897f 100644
|--- a/fs/fscache/cookie.c
|+++ b/fs/fscache/cookie.c
|@@ -963,6 +963,6 @@ void __init fscache_cookie_init(void)
| {
|        int i;
|
|-       for (i = 0; i < (1 << fscache_cookie_hash_shift) - 1; i++)
|+       for (i = 0; i < ARRAY_SIZE(fscache_cookie_hash); i++)
|                INIT_HLIST_BL_HEAD(&fscache_cookie_hash[i]);
| }
|

So that -1 at the end killed us.

fscache-initialize-cookie-hash-table-raw-spinlocks.patch was removed in
the v5.6-RT series so this should be applied to 5.4 and 4.19.

Sebastian

Re: 5.4.143-rt63 fscache_cookie_hash initialisation

[Tom Zanussi]

On Mon, 2021-09-06 at 10:36 +0200, Sebastian Andrzej Siewior wrote:
> On 2021-09-06 09:55:59 [+0200], Gregor Beck wrote:
> Your proposed change is
> 
> > index 5508d92e3f8f..cba2a226897f 100644
> > --- a/fs/fscache/cookie.c
> > +++ b/fs/fscache/cookie.c
> > @@ -963,6 +963,6 @@ void __init fscache_cookie_init(void)
> > {
> >        int i;
> > 
> > -       for (i = 0; i < (1 << fscache_cookie_hash_shift) - 1; i++)
> > +       for (i = 0; i < ARRAY_SIZE(fscache_cookie_hash); i++)
> >                INIT_HLIST_BL_HEAD(&fscache_cookie_hash[i]);
> > }
> > 
> 
> So that -1 at the end killed us.
> 
> fscache-initialize-cookie-hash-table-raw-spinlocks.patch was removed
> in
> the v5.6-RT series so this should be applied to 5.4 and 4.19.
> 

OK, thanks, I'll add this to the current 5.4 update.

Tom

> Sebastian

Re: 5.4.143-rt63 fscache_cookie_hash initialisation

[Tom Zanussi]

Hi Gregor,

On Mon, 2021-09-06 at 09:55 +0200, Gregor Beck wrote:
> Hi,
> 
> Patch 602660600bcdb53e56a87edee6b4dbbcc248a0cc "fscache: initialize
> cookie hash table raw spinlocks" fails to initialize the last element
> of fscache_cookie_hash.
> 
> [   79.218128] 003: BUG: spinlock bad magic on CPU#3, bash/2090 
> [   79.218132] 003:  lock:
> fscache_cookie_hash+0xfffe8/0xffffffffffff4fa0 [fscache], .magic:
> 00000000, .owner: <none>/-1
> , .owner_cpu: 0 
> [   79.218140] 003: CPU: 3 PID: 2090 Comm: bash Not tainted 5.4.143-
> rt63 #1 
> [   79.218141] 003: Hardware name: To be filled by O.E.M. To be
> filled by O.E.M./MAHOBAY, BIOS 4.6.5 11/07/2013 
> [   79.218142] 003: Call Trace: 
> [   79.218145] 003:  dump_stack+0x50/0x63 
> [   79.218149] 003:  do_raw_spin_lock+0x71/0xc0 
> [   79.218152] 003:  fscache_hash_cookie+0x3e/0x390 [fscache] 
> [   79.218156] 003:  __fscache_acquire_cookie+0x9f/0x2d0 [fscache] 
> [   79.218159] 003:  nfs_fscache_init_inode+0xbf/0xe0 [nfs] 
> [   79.218171] 003:  nfs_fhget+0x416/0x600 [nfs] 
> [   79.218177] 003:  nfs_readdir_page_filler+0x551/0x690 [nfs] 
> [   79.218183] 003:  nfs_readdir_xdr_to_array+0x2ba/0x420 [nfs] 
> [   79.218188] 003:  ? preempt_count_sub+0x9b/0xd0 
> [   79.218190] 003:  nfs_readdir_filler+0x1d/0x80 [nfs] 
> [   79.218195] 003:  do_read_cache_page+0x33c/0x780 
> [   79.218197] 003:  ? nfs_readdir_xdr_to_array+0x420/0x420 [nfs] 
> [   79.218201] 003:  ? verify_dirent_name+0x1c/0x30 
> [   79.218203] 003:  ? filldir64+0x3b/0x1c0 
> [   79.218204] 003:  nfs_readdir+0x20d/0x700 [nfs] 
> [   79.218209] 003:  ? nfs4_xdr_dec_server_caps+0xa0/0xa0 [nfsv4] 
> [   79.218222] 003:  iterate_dir+0x94/0x1a0 
> [   79.218223] 003:  ksys_getdents64+0xc1/0x170 
> [   79.218224] 003:  ? filldir+0x1c0/0x1c0 
> [   79.218226] 003:  __x64_sys_getdents64+0x16/0x20 
> [   79.218227] 003:  do_syscall_64+0x81/0x1b0 
> [   79.218229] 003:  entry_SYSCALL_64_after_hwframe+0x44/0xa9 
> [   79.218230] 003: RIP: 0033:0x7f3b7141781b
> 
> Gregor Beck
> 
> 

Thanks for the patch.  I'll apply it to the 5.4 rt stable kernel, but
it's missing your Signed-off-by:

Can you resend it after adding your Signed-off-by?

Thanks,

Tom