From a0c61bf02f50cc9a4e3e3703bb98ce2658794bce Mon Sep 17 00:00:00 2001 From: Su Yanjun Date: Mon, 3 Dec 2018 15:33:07 +0800 Subject: [PATCH] net: 8139cp: fix a BUG triggered by changing mtu with network traffic [ Upstream commit a5d4a89245ead1f37ed135213653c5beebea4237 ] When changing mtu many times with traffic, a bug is triggered: [ 1035.684037] kernel BUG at lib/dynamic_queue_limits.c:26! [ 1035.684042] invalid opcode: 0000 [#1] SMP [ 1035.684049] Modules linked in: loop binfmt_misc 8139cp(OE) macsec tcp_diag udp_diag inet_diag unix_diag af_packet_diag netlink_diag tcp_lp fuse uinput xt_CHECKSUM iptable_mangle ipt_MASQUERADE nf_nat_masquerade_ipv4 iptable_nat nf_nat_ipv4 nf_nat nf_conntrack_ipv4 nf_defrag_ipv4 xt_conntrack nf_conntrack ipt_REJECT nf_reject_ipv4 tun bridge stp llc ebtable_filter ebtables ip6table_filter devlink ip6_tables iptable_filter sunrpc snd_hda_codec_generic snd_hda_intel snd_hda_codec snd_hda_core snd_hwdep ppdev snd_seq iosf_mbi crc32_pclmul parport_pc snd_seq_device ghash_clmulni_intel parport snd_pcm aesni_intel joydev lrw snd_timer virtio_balloon sg gf128mul glue_helper ablk_helper cryptd snd soundcore i2c_piix4 pcspkr ip_tables xfs libcrc32c sr_mod sd_mod cdrom crc_t10dif crct10dif_generic ata_generic [ 1035.684102] pata_acpi virtio_console qxl drm_kms_helper syscopyarea sysfillrect sysimgblt floppy fb_sys_fops crct10dif_pclmul crct10dif_common ttm crc32c_intel serio_raw ata_piix drm libata 8139too virtio_pci drm_panel_orientation_quirks virtio_ring virtio mii dm_mirror dm_region_hash dm_log dm_mod [last unloaded: 8139cp] [ 1035.684132] CPU: 9 PID: 25140 Comm: if-mtu-change Kdump: loaded Tainted: G OE ------------ T 3.10.0-957.el7.x86_64 #1 [ 1035.684134] Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011 [ 1035.684136] task: ffff8f59b1f5a080 ti: ffff8f5a2e32c000 task.ti: ffff8f5a2e32c000 [ 1035.684149] RIP: 0010:[] [] dql_completed+0x180/0x190 [ 1035.684162] RSP: 0000:ffff8f5a75483e50 EFLAGS: 00010093 [ 1035.684162] RAX: 00000000000000c2 RBX: ffff8f5a6f91c000 RCX: 0000000000000000 [ 1035.684162] RDX: 0000000000000000 RSI: 0000000000000184 RDI: ffff8f599fea3ec0 [ 1035.684162] RBP: ffff8f5a75483ea8 R08: 00000000000000c2 R09: 0000000000000000 [ 1035.684162] R10: 00000000000616ef R11: ffff8f5a75483b56 R12: ffff8f599fea3e00 [ 1035.684162] R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000000184 [ 1035.684162] FS: 00007fa8434de740(0000) GS:ffff8f5a75480000(0000) knlGS:0000000000000000 [ 1035.684162] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 1035.684162] CR2: 00000000004305d0 CR3: 000000024eb66000 CR4: 00000000001406e0 [ 1035.684162] Call Trace: [ 1035.684162] [ 1035.684162] [] ? cp_interrupt+0x478/0x580 [8139cp] [ 1035.684162] [] __handle_irq_event_percpu+0x44/0x1c0 [ 1035.684162] [] handle_irq_event_percpu+0x32/0x80 [ 1035.684162] [] handle_irq_event+0x3c/0x60 [ 1035.684162] [] handle_fasteoi_irq+0x59/0x110 [ 1035.684162] [] handle_irq+0xe4/0x1a0 [ 1035.684162] [] do_IRQ+0x4d/0xf0 [ 1035.684162] [] common_interrupt+0x162/0x162 [ 1035.684162] [ 1035.684162] [] ? __wake_up_bit+0x24/0x70 [ 1035.684162] [] ? do_set_pte+0xd5/0x120 [ 1035.684162] [] unlock_page+0x2b/0x30 [ 1035.684162] [] do_read_fault.isra.61+0x139/0x1b0 [ 1035.684162] [] handle_pte_fault+0x2f4/0xd10 [ 1035.684162] [] handle_mm_fault+0x39d/0x9b0 [ 1035.684162] [] __do_page_fault+0x203/0x500 [ 1035.684162] [] trace_do_page_fault+0x56/0x150 [ 1035.684162] [] do_async_page_fault+0x22/0xf0 [ 1035.684162] [] async_page_fault+0x28/0x30 [ 1035.684162] Code: 54 c7 47 54 ff ff ff ff 44 0f 49 ce 48 8b 35 48 2f 9c 00 48 89 77 58 e9 fe fe ff ff 0f 1f 80 00 00 00 00 41 89 d1 e9 ef fe ff ff <0f> 0b 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 55 8d 42 ff 48 [ 1035.684162] RIP [] dql_completed+0x180/0x190 [ 1035.684162] RSP It's not the same as in 7fe0ee09 patch described. As 8139cp uses shared irq mode, other device irq will trigger cp_interrupt to execute. cp_change_mtu -> cp_close -> cp_open In cp_close routine just before free_irq(), some interrupt may occur. In my environment, cp_interrupt exectutes and IntrStatus is 0x4, exactly TxOk. That will cause cp_tx to wake device queue. As device queue is started, cp_start_xmit and cp_open will run at same time which will cause kernel BUG. For example: [#] for tx descriptor At start: [#][#][#] num_queued=3 After cp_init_hw->cp_start_hw->netdev_reset_queue: [#][#][#] num_queued=0 When 8139cp starts to work then cp_tx will check num_queued mismatchs the complete_bytes. The patch will check IntrMask before check IntrStatus in cp_interrupt. When 8139cp interrupt is disabled, just return. Signed-off-by: Su Yanjun Signed-off-by: David S. Miller Signed-off-by: Greg Kroah-Hartman --- drivers/net/ethernet/realtek/8139cp.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/net/ethernet/realtek/8139cp.c b/drivers/net/ethernet/realtek/8139cp.c index 75b1693..1c0c1a2 100644 --- a/drivers/net/ethernet/realtek/8139cp.c +++ b/drivers/net/ethernet/realtek/8139cp.c @@ -576,6 +576,7 @@ static irqreturn_t cp_interrupt (int irq, void *dev_instance) struct cp_private *cp; int handled = 0; u16 status; + u16 mask; if (unlikely(dev == NULL)) return IRQ_NONE; @@ -583,6 +584,10 @@ static irqreturn_t cp_interrupt (int irq, void *dev_instance) spin_lock(&cp->lock); + mask = cpr16(IntrMask); + if (!mask) + goto out_unlock; + status = cpr16(IntrStatus); if (!status || (status == 0xFFFF)) goto out_unlock; -- 2.7.4