Skip to contentSkip to conversation

Kernel 2.6.9 Page Allocation Failures w/TSO+rollup.patch

DISCUSSION Chuck Ebbert 2 participants 2 messages Oct 25, 2004

lore
Share & tools
lore
#0 lore raw
Quoted 5 lines · Nicj Piggin

Does it cause any noticable problems? If not, then stay with 2.6.9.

However, it would be nice to get to the bottom of it. It might just be happening by chance on 2.6.9 but not 2.6.8.1 though...

Isn't this the problem fixed by the below patch? (Sorry I didn't
get sender name when I collected it.) Some were skeptical this would fix it but it has worked for those who tried...

Oh and BTW what is rollup.patch?

# The following patch makes it allocate skb_headlen(skb) - len instead

# of skb->len - len.  When skb is linear there is no difference.  When
# it's non-linear we only ever copy the bytes in the header.
#
===== net/ipv4/tcp_output.c 1.67 vs edited =====
--- 1.67/net/ipv4/tcp_output.c  2004-10-01 13:56:45 +10:00
+++ edited/net/ipv4/tcp_output.c        2004-10-17 18:58:47 +10:00
@@ -455,8 +455,12 @@
 {
        struct tcp_opt *tp = tcp_sk(sk);
        struct sk_buff *buff;
-       int nsize = skb->len - len;
+       int nsize;
        u16 flags;
+
+       nsize = skb_headlen(skb) - len;
+       if (nsize < 0)
+               nsize = 0;
        if (skb_cloned(skb) &&
            skb_is_nonlinear(skb) &&

--Chuck Ebbert 25-Oct-04 14:54:36

#1 lore raw

No, the below patch did not fix the problem.

I am waiting for the error to re-occur then I will sysrq+M and dmesg & send the info to the mailing list.

Rollup patch was from Nick Piggin, he stated:

Date: Sat, 23 Oct 2004 19:14:46 +1000
From: Nick Piggin <[email protected]>
To: Justin Piszcz <[email protected]>
Cc: [email protected]
Subject: Re: Page Allocation Failures Return With 2.6.9+TSO patch. Parts/Attachments:
1 Shown 32 lines Text
2 Shown 356 lines Text
----------------------------------------

Quoted 17 lines · Justin Piszcz

Kernel 2.6.9 w/TSO patch.

(most likely do to the e1000/nic/issue)

$ dmesg
gaim: page allocation failure. order:4, mode:0x21
 [<c01391a7>] __alloc_pages+0x247/0x3b0
 [<c0139328>] __get_free_pages+0x18/0x40
 [<c035c33a>] sound_alloc_dmap+0xaa/0x1b0
 [<c03648c0>] ad_mute+0x20/0x40
 [<c035c70f>] open_dmap+0x1f/0x100
 [<c035cb58>] DMAbuf_open+0x178/0x1d0
 [<c035a4fa>] audio_open+0xba/0x280
 [<c015d863>] cdev_get+0x53/0xc0
 [<c035968c>] sound_open+0xac/0x110
 [<c035898e>] soundcore_open+0x1ce/0x300
 [<c03587c0>] soundcore_open+0x0/0x300
Quoted 8 lines
 [<c015d524>] chrdev_open+0x104/0x250
 [<c015d420>] chrdev_open+0x0/0x250
 [<c0152d82>] dentry_open+0x1d2/0x270
 [<c0152b9c>] filp_open+0x5c/0x70
 [<c01049c8>] common_interrupt+0x18/0x20
 [<c0152e75>] get_unused_fd+0x55/0xf0
 [<c0152fd9>] sys_open+0x49/0x90
 [<c010405b>] syscall_call+0x7/0xb

Ouch, 64K atomic DMA allocation.

The DMA zone barely even keeps that much total memory free.

The caller probably wants fixing, but you could try this patch...

[ Part 2: "Attached Text" ]

Pine -> 2 356 lines Text/X-PATCH (Name: "rollup.patch")

That was rollup.patch (patched 3 source files) - but this as well as the TSO has not seemed to have fixed the problem.

Quoted 39 lines · Chuck Ebbert
Quoted 5 lines · Nicj Piggin

Does it cause any noticable problems? If not, then stay with 2.6.9.

However, it would be nice to get to the bottom of it. It might just be happening by chance on 2.6.9 but not 2.6.8.1 though...

Isn't this the problem fixed by the below patch? (Sorry I didn't
get sender name when I collected it.) Some were skeptical this would fix it but it has worked for those who tried...

Oh and BTW what is rollup.patch?

# The following patch makes it allocate skb_headlen(skb) - len instead

# of skb->len - len.  When skb is linear there is no difference.  When
# it's non-linear we only ever copy the bytes in the header.
#
===== net/ipv4/tcp_output.c 1.67 vs edited =====
net/ipv4/tcp_output.c 2004-10-17 18:58:47 +10:00 renamed from net/ipv4/tcp_output.c 2004-10-01 13:56:45 +10:00 +5 -2
@@ -455,8 +455,12 @@
 {
       struct tcp_opt *tp = tcp_sk(sk);
       struct sk_buff *buff;
-       int nsize = skb->len - len;
+       int nsize;
       u16 flags;
+
+       nsize = skb_headlen(skb) - len;
+       if (nsize < 0)
+               nsize = 0;
 
       if (skb_cloned(skb) &&
           skb_is_nonlinear(skb) &&
 
--Chuck Ebbert  25-Oct-04  14:54:36