Jakob Goldbach wrote:
> That would be great. Thanks.There are usually a few days  between it gets stuck.

Ok. Happily, I've managed to invent what I need to check first
before it's too late here in Moscow ;)

I presume, that the infinite loop is really somewhere near the
__d_lookup. Please, apply this patch in attach (I made it against
2.6.18-028stab053.5, but should fit OK all the other 028stab053
releases) and check for warnings in dmesg ;)

Let's see whether this is really __d_lookup.

> /jakob
> - oprindelig besked -
> Emne:	Re: [Users] Infinite loop in __d_lookup ?
> Fra:	Pavel Emelyanov <xemul@openvz.org>
> Dato:		15-05-2008 12:34
> 
> Jakob Goldbach wrote:
>> Hi,
>>
>> I regularly have processes that gets stock eating all cpu. SysRq-p says
>> it is stock in __d_lookup+0x10b as seen in dmesg output below.
> 
> If you can reproduce this in a reasonable time I can send you
> a debugging patch to find out what's going on there. 
> 
> Let's try with it?
> 
>> I run vanilla 2.6.18 with 028stab053 and the lustre filesystem. I also
>> run lustre on non-openvz kernel without problems, hence this mail to
>> this group. 
>>
>> I believe I've found where the problem is, but I'm not a kernel hacker
>> so I don't know what to do about this information. 
>>
>> I'd appreciate any hints on what to do next to get this solved.
>>
>> Below is what I could find out. 
>>
>> Thanks,
>> Jakob 
>>
>> gdb find that the process is in the hlist_for_each_entry_rcu loop:
>>
>> (gdb) list *__d_lookup+0x10b
>> 0x12f0 is in __d_lookup (fs/dcache.c:1153).
>> 1148            struct dentry *dentry, *found;
>> 1149
>> 1150            rcu_read_lock();
>> 1151
>> 1152            found = NULL;
>> 1153            hlist_for_each_entry_rcu(dentry, node, head, d_hash) {
>> 1154                    struct qstr *qstr;
>> 1155
>> 1156                    if (dentry->d_name.hash != hash)
>> 1157                            continue;
>>
>> I believe this is the relevant part (0x12f0) of the disassembled object:
>>  
>>     12e0:       4d 8b 24 24             mov    (%r12),%r12
>>     12e4:       4d 85 e4                test   %r12,%r12
>>     12e7:       74 2c                   je     1315 <__d_lookup+0x130>
>>     12e9:       49 8b 04 24             mov    (%r12),%rax
>>     12ed:       0f 18 08                prefetcht0 (%rax)
>>     12f0:       49 8d 5c 24 d8          lea    0xffffffffffffffd8(%r12),
>> %rbx
>>     12f5:       8b 45 cc                mov    0xffffffffffffffcc(%rbp),
>> %eax
>>     12f8:       39 43 40                cmp    %eax,0x40(%rbx)
>>     12fb:       75 e3                   jne    12e0 <__d_lookup+0xfb>
>>
>>
>> Dmesg after sysrq-p:
>>
>>
>>
>>
>> [186124.494329] SysRq: Show Regs
>> [186124.495218] ----------- IPI show regs -----------
>> [186124.496136] CPU 3, VCPU 0:1
>> [186124.496804] Modules linked in: simfs vznetdev vzethdev vzrst ip_nat
>> vzcpt ip_conntrack nfnetlink vzdquota vzmon vzdev xt_length ipt_ttl xt_
>> tcpmss ipt_TCPMSS iptable_mangle xt_multiport xt_limit ipt_tos
>> ipt_REJECT iptable_filter ip_tables x_tables 8021q osc mgc lustre lov
>> lquota mdc
>>  ksocklnd ptlrpc obdclass lnet lvfs libcfs bonding xfs
>> [186124.503636] Pid: 22699, comm: find Not tainted
>> 2.6.18.8-openvz-028stab053-bnx2-1.6.7b-arpannounce1 #3 028stab053
>> [186124.505535] RIP: 0060:[<ffffffff8029b314>]  [<ffffffff8029b314>]
>> __d_lookup+0x10b/0x142
>> [186124.507265] RSP: 0068:ffff810073d63bc8  EFLAGS: 00000282
>> [186124.508296] RAX: ffff8101016dc298 RBX: ffff8101016dc270 RCX:
>> 0000000000000013
>> [186124.509768] RDX: 0000000000025ff5 RSI: 00c38320c56a5ff5 RDI:
>> ffff810118b056b0
>> [186124.511480] RBP: ffff810073d63c08 R08: ffff8100ac9e8000 R09:
>> ffff810118b056b0
>> [186124.512963] R10: 0000000000000000 R11: 0000000000000000 R12:
>> ffff8101016dc298
>> [186124.514452] R13: ffff810073d63e38 R14: ffff810118b056b0 R15:
>> ffff810073d63c78
>> [186124.515931] FS:  00002ba786cb56d0(0000) GS:ffff81012a693340(0000)
>> knlGS:0000000000000000
>> [186124.517538] CS:  0060 DS: 0000 ES: 0000 CR0: 0000000080050033
>> [186124.518587] CR2: 0000000000539938 CR3: 0000000073f06000 CR4:
>> 00000000000006e0
>> [186124.520022] 
>> [186124.520023] Call Trace:
>> [186124.521245]  [<ffffffff8029105d>] do_lookup+0x2c/0x193
>> [186124.522363]  [<ffffffff80293122>] __link_path_walk+0xb07/0x10ac
>> [186124.523642]  [<ffffffff8029374e>] link_path_walk+0x87/0x140
>> [186124.524818]  [<ffffffff80293c76>] do_path_lookup+0x2d3/0x2f8
>> [186124.526000]  [<ffffffff802945e2>] __user_walk_fd+0x41/0x62
>> [186124.527156]  [<ffffffff8028cecb>] vfs_lstat_fd+0x24/0x5a
>> [186124.528278]  [<ffffffff8028cf23>] sys_newlstat+0x22/0x3c
>> [186124.529383]  [<ffffffff80209902>] system_call+0x7e/0x83
>> [186124.530362] DWARF2 unwinder stuck at system_call+0x7e/0x83
>> [186124.531460] Leftover inexact backtrace:
>> [186124.532563] 
>>
>>
--- ./fs/dcache.c.loopdebug	2008-05-15 20:09:04.000000000 +0400
+++ ./fs/dcache.c	2008-05-15 20:16:19.000000000 +0400
@@ -1128,12 +1128,24 @@ struct dentry * d_lookup(struct dentry *
 {
 	struct dentry * dentry = NULL;
 	unsigned long seq;
+	unsigned long loops = 0;
+	static int once = 1;
 
         do {
                 seq = read_seqbegin(&rename_lock);
                 dentry = __d_lookup(parent, name);
                 if (dentry)
 			break;
+
+		if (loops++ > 200) {
+			printk("%s: Abort on 200 seq-retry iteration\n",
+					__func__);
+			if (once) {
+				once = 0;
+				dump_stack();
+			}
+			break;
+		}
 	} while (read_seqretry(&rename_lock, seq));
 	return dentry;
 }
@@ -1146,6 +1158,8 @@ struct dentry * __d_lookup(struct dentry
 	struct hlist_head *head = d_hash(parent,hash);
 	struct hlist_node *node;
 	struct dentry *dentry, *found;
+	unsigned long loops = 0;
+	static int once = 1;
 
 	rcu_read_lock();
 	
@@ -1154,9 +1168,9 @@ struct dentry * __d_lookup(struct dentry
 		struct qstr *qstr;
 
 		if (dentry->d_name.hash != hash)
-			continue;
+			goto next_nolock;
 		if (dentry->d_parent != parent)
-			continue;
+			goto next_nolock;
 
 		spin_lock(&dentry->d_lock);
 
@@ -1193,6 +1207,16 @@ struct dentry * __d_lookup(struct dentry
 		break;
 next:
 		spin_unlock(&dentry->d_lock);
+next_nolock:
+		if (loops++ > 5000) {
+			printk("%s: Abort on 5000 loop iteration in a chain\n",
+					__func__);
+			if (once) {
+				once = 0;
+				dump_stack();
+			}
+			break;
+		}
  	}
  	rcu_read_unlock();