1
0
Fork 0
alistair23-linux/drivers/visorbus/visorchannel.c

435 lines
12 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (C) 2010 - 2015 UNISYS CORPORATION
* All rights reserved.
*/
/*
* This provides s-Par channel communication primitives, which are
* independent of the mechanism used to access the channel data.
*/
#include <linux/uuid.h>
#include <linux/io.h>
#include <linux/slab.h>
#include <linux/visorbus.h>
#include "visorbus_private.h"
#include "controlvmchannel.h"
#define VISOR_DRV_NAME "visorchannel"
#define VISOR_CONSOLEVIDEO_CHANNEL_GUID \
GUID_INIT(0x3cd6e705, 0xd6a2, 0x4aa5, \
0xad, 0x5c, 0x7b, 0x8, 0x88, 0x9d, 0xff, 0xe2)
static const guid_t visor_video_guid = VISOR_CONSOLEVIDEO_CHANNEL_GUID;
struct visorchannel {
u64 physaddr;
ulong nbytes;
void *mapped;
bool requested;
struct channel_header chan_hdr;
guid_t guid;
/*
* channel creator knows if more than one thread will be inserting or
* removing
*/
bool needs_lock;
/* protect head writes in chan_hdr */
spinlock_t insert_lock;
/* protect tail writes in chan_hdr */
spinlock_t remove_lock;
guid_t type;
guid_t inst;
};
void visorchannel_destroy(struct visorchannel *channel)
{
if (!channel)
return;
if (channel->mapped) {
memunmap(channel->mapped);
if (channel->requested)
release_mem_region(channel->physaddr, channel->nbytes);
}
kfree(channel);
}
u64 visorchannel_get_physaddr(struct visorchannel *channel)
{
return channel->physaddr;
}
ulong visorchannel_get_nbytes(struct visorchannel *channel)
{
return channel->nbytes;
}
char *visorchannel_guid_id(const guid_t *guid, char *s)
{
sprintf(s, "%pUL", guid);
return s;
}
char *visorchannel_id(struct visorchannel *channel, char *s)
{
return visorchannel_guid_id(&channel->guid, s);
}
char *visorchannel_zoneid(struct visorchannel *channel, char *s)
{
return visorchannel_guid_id(&channel->chan_hdr.zone_guid, s);
}
u64 visorchannel_get_clientpartition(struct visorchannel *channel)
{
return channel->chan_hdr.partition_handle;
}
int visorchannel_set_clientpartition(struct visorchannel *channel,
u64 partition_handle)
{
channel->chan_hdr.partition_handle = partition_handle;
return 0;
}
/**
* visorchannel_get_guid() - queries the GUID of the designated channel
* @channel: the channel to query
*
* Return: the GUID of the provided channel
*/
const guid_t *visorchannel_get_guid(struct visorchannel *channel)
{
return &channel->guid;
}
EXPORT_SYMBOL_GPL(visorchannel_get_guid);
int visorchannel_read(struct visorchannel *channel, ulong offset, void *dest,
ulong nbytes)
{
if (offset + nbytes > channel->nbytes)
return -EIO;
memcpy(dest, channel->mapped + offset, nbytes);
return 0;
}
int visorchannel_write(struct visorchannel *channel, ulong offset, void *dest,
ulong nbytes)
{
size_t chdr_size = sizeof(struct channel_header);
size_t copy_size;
if (offset + nbytes > channel->nbytes)
return -EIO;
if (offset < chdr_size) {
copy_size = min(chdr_size - offset, nbytes);
memcpy(((char *)(&channel->chan_hdr)) + offset,
dest, copy_size);
}
memcpy(channel->mapped + offset, dest, nbytes);
return 0;
}
void *visorchannel_get_header(struct visorchannel *channel)
{
return &channel->chan_hdr;
}
/*
* Return offset of a specific SIGNAL_QUEUE_HEADER from the beginning of a
* channel header
*/
static int sig_queue_offset(struct channel_header *chan_hdr, int q)
{
return ((chan_hdr)->ch_space_offset +
((q) * sizeof(struct signal_queue_header)));
}
/*
* Return offset of a specific queue entry (data) from the beginning of a
* channel header
*/
static int sig_data_offset(struct channel_header *chan_hdr, int q,
struct signal_queue_header *sig_hdr, int slot)
{
return (sig_queue_offset(chan_hdr, q) + sig_hdr->sig_base_offset +
(slot * sig_hdr->signal_size));
}
/*
* Write the contents of a specific field within a SIGNAL_QUEUE_HEADER back into
* host memory
*/
#define SIG_WRITE_FIELD(channel, queue, sig_hdr, FIELD) \
visorchannel_write(channel, \
sig_queue_offset(&channel->chan_hdr, queue) + \
offsetof(struct signal_queue_header, FIELD), \
&((sig_hdr)->FIELD), \
sizeof((sig_hdr)->FIELD))
static int sig_read_header(struct visorchannel *channel, u32 queue,
struct signal_queue_header *sig_hdr)
{
if (channel->chan_hdr.ch_space_offset < sizeof(struct channel_header))
return -EINVAL;
/* Read the appropriate SIGNAL_QUEUE_HEADER into local memory. */
return visorchannel_read(channel,
sig_queue_offset(&channel->chan_hdr, queue),
sig_hdr, sizeof(struct signal_queue_header));
}
static int sig_read_data(struct visorchannel *channel, u32 queue,
struct signal_queue_header *sig_hdr, u32 slot,
void *data)
{
int signal_data_offset = sig_data_offset(&channel->chan_hdr, queue,
sig_hdr, slot);
return visorchannel_read(channel, signal_data_offset,
data, sig_hdr->signal_size);
}
static int sig_write_data(struct visorchannel *channel, u32 queue,
struct signal_queue_header *sig_hdr, u32 slot,
void *data)
{
int signal_data_offset = sig_data_offset(&channel->chan_hdr, queue,
sig_hdr, slot);
return visorchannel_write(channel, signal_data_offset,
data, sig_hdr->signal_size);
}
static int signalremove_inner(struct visorchannel *channel, u32 queue,
void *msg)
{
struct signal_queue_header sig_hdr;
int error;
error = sig_read_header(channel, queue, &sig_hdr);
if (error)
return error;
/* No signals to remove; have caller try again. */
if (sig_hdr.head == sig_hdr.tail)
return -EAGAIN;
sig_hdr.tail = (sig_hdr.tail + 1) % sig_hdr.max_slots;
error = sig_read_data(channel, queue, &sig_hdr, sig_hdr.tail, msg);
if (error)
return error;
sig_hdr.num_received++;
/*
* For each data field in SIGNAL_QUEUE_HEADER that was modified, update
* host memory. Required for channel sync.
*/
mb();
error = SIG_WRITE_FIELD(channel, queue, &sig_hdr, tail);
if (error)
return error;
error = SIG_WRITE_FIELD(channel, queue, &sig_hdr, num_received);
if (error)
return error;
return 0;
}
/**
* visorchannel_signalremove() - removes a message from the designated
* channel/queue
* @channel: the channel the message will be removed from
* @queue: the queue the message will be removed from
* @msg: the message to remove
*
* Return: integer error code indicating the status of the removal
*/
int visorchannel_signalremove(struct visorchannel *channel, u32 queue,
void *msg)
{
int rc;
staging: unisys: fix random hangs with network stress in visornic We learned that it was possible for the core networking code to call visornic_xmit() within ISR context, resulting in the need for us to use spin_lock_irqsave() / spin_lock_irqrestore() to lock accesses to our virtual device channels. Without the correct locking added in this patch, random hangs would occur on typical kernels while stressing the netork. When using a kernel with CONFIG_DEBUG_SPINLOCK=y, a stackdump would occur at the time of the hang reporting: BUG: spinlock recursion on CPU#0, vnic_incoming/<pid> (see below for more details) We considered the possibility of adding a protocol between a visordriver and visorbus where the visordriver could specify which type of locking it required for its virtual device channels (essentially indicating whether or not it was possible for the channel to be accessed in ISR context), but decided this extra complexity was NOT needed, and that channel queues should always be accessed with the most-stringent locking. So that is what is implemented in this commit. Below is an example stackdump illustrating the spinlock recursion that is fixed by this commit. Note that we are first in virtnic_rx() writing to the device channel when an APIC timer interrupt occurs. Within the core networking code, net_rx_action() calls process_backlog(), which eventually lands up back up in virtnic_xmit() in the code attempting to also write to the device channel. BUG: spinlock recursion on CPU#0, vnic_incoming/262 lock: 0xffff88002db810c0, .magic: dead4ead, .owner: vnic_incoming/262, .owner_cpu: 0 CPU: 0 PID: 262 Comm: vnic_incoming Tainted: G C 4.2.0-rc1-ARCH+ #56 Hardware name: Dell Inc. PowerEdge T110/ , BIOS 1.23 12/15/2009 ffff8800216ac200 ffff88002c803388 ffffffff81476364 0000000000000106 ffff88002db810c0 ffff88002c8033a8 ffffffff8109e2bc ffff88002db810c0 ffffffff817631d4 ffff88002c8033c8 ffffffff8109e330 ffff88002db810c0 Call Trace: <IRQ> [<ffffffff81476364>] dump_stack+0x4f/0x73 [<ffffffff8109e2bc>] spin_dump+0x7c/0xc0 [<ffffffff8109e330>] spin_bug+0x30/0x40 [<ffffffff8109e547>] do_raw_spin_lock+0x127/0x140 [<ffffffff8147bad0>] _raw_spin_lock+0x40/0x50 [<ffffffffa0151fa6>] ? visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa0151fa6>] visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa01683a2>] visornic_xmit+0x302/0x5d0 [visornic] [<ffffffff813b2f30>] dev_hard_start_xmit+0x2e0/0x510 [<ffffffff813b2b75>] ? validate_xmit_skb+0x235/0x310 [<ffffffff813d79e7>] sch_direct_xmit+0xf7/0x1d0 [<ffffffff813b34d3>] __dev_queue_xmit+0x203/0x640 [<ffffffff813b3320>] ? __dev_queue_xmit+0x50/0x640 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff813b3933>] dev_queue_xmit_sk+0x13/0x20 [<ffffffff813f3a5c>] ip_finish_output2+0x22c/0x470 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813f3f6f>] ip_finish_output+0x1df/0x310 [<ffffffff813f4c31>] ip_output+0xb1/0x100 [<ffffffff813f41be>] ip_local_out_sk+0x3e/0x80 [<ffffffff813f4388>] ip_queue_xmit+0x188/0x4a0 [<ffffffff813f4200>] ? ip_local_out_sk+0x80/0x80 [<ffffffff8139fcd6>] ? __alloc_skb+0x86/0x1e0 [<ffffffff8140bd5b>] tcp_transmit_skb+0x4cb/0x9c0 [<ffffffff8139f0dc>] ? __kmalloc_reserve+0x3c/0x90 [<ffffffff8139fcea>] ? __alloc_skb+0x9a/0x1e0 [<ffffffff8140c47d>] tcp_send_ack+0x10d/0x150 [<ffffffff814060ee>] __tcp_ack_snd_check+0x5e/0x90 [<ffffffff81408eb4>] tcp_rcv_established+0x354/0x710 [<ffffffff81412182>] tcp_v4_do_rcv+0x162/0x3f0 [<ffffffff81414412>] tcp_v4_rcv+0xb22/0xb50 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee350>] ip_local_deliver_finish+0xe0/0x2d0 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee72e>] ip_local_deliver+0xae/0xc0 [<ffffffff813edeaf>] ip_rcv_finish+0x14f/0x510 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff813eea13>] ip_rcv+0x2d3/0x3b0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff813ab0f3>] __netif_receive_skb_core+0x663/0xb70 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff810971a9>] ? cpuacct_charge+0x99/0xb0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813ab72c>] ? process_backlog+0xbc/0x150 [<ffffffff813ab78b>] ? process_backlog+0x11b/0x150 [<ffffffff813ab627>] __netif_receive_skb+0x27/0x70 [<ffffffff813ab702>] process_backlog+0x92/0x150 [<ffffffff813afffd>] net_rx_action+0x13d/0x350 [<ffffffff81036b2d>] ? lapic_next_event+0x1d/0x30 [<ffffffff81058694>] __do_softirq+0x104/0x320 [<ffffffff810c0788>] ? hrtimer_interrupt+0xc8/0x1a0 [<ffffffff81074e70>] ? blocking_notifier_chain_cond_register+0x70/0x70 [<ffffffff81058ab9>] irq_exit+0x79/0xa0 [<ffffffff8147ecca>] smp_apic_timer_interrupt+0x4a/0x60 [<ffffffff8147d2c8>] apic_timer_interrupt+0x68/0x70 <EOI> [<ffffffff81271c02>] ? __memcpy+0x12/0x20 [<ffffffffa01517da>] ? visorchannel_write+0x4a/0x80 [visorbus] [<ffffffffa0151eb8>] signalinsert_inner+0x88/0x130 [visorbus] [<ffffffffa0151fb5>] visorchannel_signalinsert+0x55/0x70 [visorbus] [<ffffffffa0166e57>] visornic_rx+0x12e7/0x19d0 [visornic] [<ffffffffa01677c9>] process_incoming_rsps+0x289/0x690 [visornic] [<ffffffff814771c5>] ? preempt_schedule+0x25/0x30 [<ffffffff81001026>] ? ___preempt_schedule+0x12/0x14 [<ffffffff81093080>] ? wait_woken+0x90/0x90 [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffff81073a39>] kthread+0xe9/0x110 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 [<ffffffff8147c89f>] ret_from_fork+0x3f/0x70 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 Fixes: b12fdf7da ('staging: unisys: rework signal remove/insert to avoid sparse lock warnings') Signed-off-by: Tim Sell <Timothy.Sell@unisys.com> Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-07-14 12:43:30 -06:00
unsigned long flags;
if (channel->needs_lock) {
staging: unisys: fix random hangs with network stress in visornic We learned that it was possible for the core networking code to call visornic_xmit() within ISR context, resulting in the need for us to use spin_lock_irqsave() / spin_lock_irqrestore() to lock accesses to our virtual device channels. Without the correct locking added in this patch, random hangs would occur on typical kernels while stressing the netork. When using a kernel with CONFIG_DEBUG_SPINLOCK=y, a stackdump would occur at the time of the hang reporting: BUG: spinlock recursion on CPU#0, vnic_incoming/<pid> (see below for more details) We considered the possibility of adding a protocol between a visordriver and visorbus where the visordriver could specify which type of locking it required for its virtual device channels (essentially indicating whether or not it was possible for the channel to be accessed in ISR context), but decided this extra complexity was NOT needed, and that channel queues should always be accessed with the most-stringent locking. So that is what is implemented in this commit. Below is an example stackdump illustrating the spinlock recursion that is fixed by this commit. Note that we are first in virtnic_rx() writing to the device channel when an APIC timer interrupt occurs. Within the core networking code, net_rx_action() calls process_backlog(), which eventually lands up back up in virtnic_xmit() in the code attempting to also write to the device channel. BUG: spinlock recursion on CPU#0, vnic_incoming/262 lock: 0xffff88002db810c0, .magic: dead4ead, .owner: vnic_incoming/262, .owner_cpu: 0 CPU: 0 PID: 262 Comm: vnic_incoming Tainted: G C 4.2.0-rc1-ARCH+ #56 Hardware name: Dell Inc. PowerEdge T110/ , BIOS 1.23 12/15/2009 ffff8800216ac200 ffff88002c803388 ffffffff81476364 0000000000000106 ffff88002db810c0 ffff88002c8033a8 ffffffff8109e2bc ffff88002db810c0 ffffffff817631d4 ffff88002c8033c8 ffffffff8109e330 ffff88002db810c0 Call Trace: <IRQ> [<ffffffff81476364>] dump_stack+0x4f/0x73 [<ffffffff8109e2bc>] spin_dump+0x7c/0xc0 [<ffffffff8109e330>] spin_bug+0x30/0x40 [<ffffffff8109e547>] do_raw_spin_lock+0x127/0x140 [<ffffffff8147bad0>] _raw_spin_lock+0x40/0x50 [<ffffffffa0151fa6>] ? visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa0151fa6>] visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa01683a2>] visornic_xmit+0x302/0x5d0 [visornic] [<ffffffff813b2f30>] dev_hard_start_xmit+0x2e0/0x510 [<ffffffff813b2b75>] ? validate_xmit_skb+0x235/0x310 [<ffffffff813d79e7>] sch_direct_xmit+0xf7/0x1d0 [<ffffffff813b34d3>] __dev_queue_xmit+0x203/0x640 [<ffffffff813b3320>] ? __dev_queue_xmit+0x50/0x640 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff813b3933>] dev_queue_xmit_sk+0x13/0x20 [<ffffffff813f3a5c>] ip_finish_output2+0x22c/0x470 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813f3f6f>] ip_finish_output+0x1df/0x310 [<ffffffff813f4c31>] ip_output+0xb1/0x100 [<ffffffff813f41be>] ip_local_out_sk+0x3e/0x80 [<ffffffff813f4388>] ip_queue_xmit+0x188/0x4a0 [<ffffffff813f4200>] ? ip_local_out_sk+0x80/0x80 [<ffffffff8139fcd6>] ? __alloc_skb+0x86/0x1e0 [<ffffffff8140bd5b>] tcp_transmit_skb+0x4cb/0x9c0 [<ffffffff8139f0dc>] ? __kmalloc_reserve+0x3c/0x90 [<ffffffff8139fcea>] ? __alloc_skb+0x9a/0x1e0 [<ffffffff8140c47d>] tcp_send_ack+0x10d/0x150 [<ffffffff814060ee>] __tcp_ack_snd_check+0x5e/0x90 [<ffffffff81408eb4>] tcp_rcv_established+0x354/0x710 [<ffffffff81412182>] tcp_v4_do_rcv+0x162/0x3f0 [<ffffffff81414412>] tcp_v4_rcv+0xb22/0xb50 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee350>] ip_local_deliver_finish+0xe0/0x2d0 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee72e>] ip_local_deliver+0xae/0xc0 [<ffffffff813edeaf>] ip_rcv_finish+0x14f/0x510 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff813eea13>] ip_rcv+0x2d3/0x3b0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff813ab0f3>] __netif_receive_skb_core+0x663/0xb70 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff810971a9>] ? cpuacct_charge+0x99/0xb0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813ab72c>] ? process_backlog+0xbc/0x150 [<ffffffff813ab78b>] ? process_backlog+0x11b/0x150 [<ffffffff813ab627>] __netif_receive_skb+0x27/0x70 [<ffffffff813ab702>] process_backlog+0x92/0x150 [<ffffffff813afffd>] net_rx_action+0x13d/0x350 [<ffffffff81036b2d>] ? lapic_next_event+0x1d/0x30 [<ffffffff81058694>] __do_softirq+0x104/0x320 [<ffffffff810c0788>] ? hrtimer_interrupt+0xc8/0x1a0 [<ffffffff81074e70>] ? blocking_notifier_chain_cond_register+0x70/0x70 [<ffffffff81058ab9>] irq_exit+0x79/0xa0 [<ffffffff8147ecca>] smp_apic_timer_interrupt+0x4a/0x60 [<ffffffff8147d2c8>] apic_timer_interrupt+0x68/0x70 <EOI> [<ffffffff81271c02>] ? __memcpy+0x12/0x20 [<ffffffffa01517da>] ? visorchannel_write+0x4a/0x80 [visorbus] [<ffffffffa0151eb8>] signalinsert_inner+0x88/0x130 [visorbus] [<ffffffffa0151fb5>] visorchannel_signalinsert+0x55/0x70 [visorbus] [<ffffffffa0166e57>] visornic_rx+0x12e7/0x19d0 [visornic] [<ffffffffa01677c9>] process_incoming_rsps+0x289/0x690 [visornic] [<ffffffff814771c5>] ? preempt_schedule+0x25/0x30 [<ffffffff81001026>] ? ___preempt_schedule+0x12/0x14 [<ffffffff81093080>] ? wait_woken+0x90/0x90 [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffff81073a39>] kthread+0xe9/0x110 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 [<ffffffff8147c89f>] ret_from_fork+0x3f/0x70 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 Fixes: b12fdf7da ('staging: unisys: rework signal remove/insert to avoid sparse lock warnings') Signed-off-by: Tim Sell <Timothy.Sell@unisys.com> Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-07-14 12:43:30 -06:00
spin_lock_irqsave(&channel->remove_lock, flags);
rc = signalremove_inner(channel, queue, msg);
staging: unisys: fix random hangs with network stress in visornic We learned that it was possible for the core networking code to call visornic_xmit() within ISR context, resulting in the need for us to use spin_lock_irqsave() / spin_lock_irqrestore() to lock accesses to our virtual device channels. Without the correct locking added in this patch, random hangs would occur on typical kernels while stressing the netork. When using a kernel with CONFIG_DEBUG_SPINLOCK=y, a stackdump would occur at the time of the hang reporting: BUG: spinlock recursion on CPU#0, vnic_incoming/<pid> (see below for more details) We considered the possibility of adding a protocol between a visordriver and visorbus where the visordriver could specify which type of locking it required for its virtual device channels (essentially indicating whether or not it was possible for the channel to be accessed in ISR context), but decided this extra complexity was NOT needed, and that channel queues should always be accessed with the most-stringent locking. So that is what is implemented in this commit. Below is an example stackdump illustrating the spinlock recursion that is fixed by this commit. Note that we are first in virtnic_rx() writing to the device channel when an APIC timer interrupt occurs. Within the core networking code, net_rx_action() calls process_backlog(), which eventually lands up back up in virtnic_xmit() in the code attempting to also write to the device channel. BUG: spinlock recursion on CPU#0, vnic_incoming/262 lock: 0xffff88002db810c0, .magic: dead4ead, .owner: vnic_incoming/262, .owner_cpu: 0 CPU: 0 PID: 262 Comm: vnic_incoming Tainted: G C 4.2.0-rc1-ARCH+ #56 Hardware name: Dell Inc. PowerEdge T110/ , BIOS 1.23 12/15/2009 ffff8800216ac200 ffff88002c803388 ffffffff81476364 0000000000000106 ffff88002db810c0 ffff88002c8033a8 ffffffff8109e2bc ffff88002db810c0 ffffffff817631d4 ffff88002c8033c8 ffffffff8109e330 ffff88002db810c0 Call Trace: <IRQ> [<ffffffff81476364>] dump_stack+0x4f/0x73 [<ffffffff8109e2bc>] spin_dump+0x7c/0xc0 [<ffffffff8109e330>] spin_bug+0x30/0x40 [<ffffffff8109e547>] do_raw_spin_lock+0x127/0x140 [<ffffffff8147bad0>] _raw_spin_lock+0x40/0x50 [<ffffffffa0151fa6>] ? visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa0151fa6>] visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa01683a2>] visornic_xmit+0x302/0x5d0 [visornic] [<ffffffff813b2f30>] dev_hard_start_xmit+0x2e0/0x510 [<ffffffff813b2b75>] ? validate_xmit_skb+0x235/0x310 [<ffffffff813d79e7>] sch_direct_xmit+0xf7/0x1d0 [<ffffffff813b34d3>] __dev_queue_xmit+0x203/0x640 [<ffffffff813b3320>] ? __dev_queue_xmit+0x50/0x640 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff813b3933>] dev_queue_xmit_sk+0x13/0x20 [<ffffffff813f3a5c>] ip_finish_output2+0x22c/0x470 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813f3f6f>] ip_finish_output+0x1df/0x310 [<ffffffff813f4c31>] ip_output+0xb1/0x100 [<ffffffff813f41be>] ip_local_out_sk+0x3e/0x80 [<ffffffff813f4388>] ip_queue_xmit+0x188/0x4a0 [<ffffffff813f4200>] ? ip_local_out_sk+0x80/0x80 [<ffffffff8139fcd6>] ? __alloc_skb+0x86/0x1e0 [<ffffffff8140bd5b>] tcp_transmit_skb+0x4cb/0x9c0 [<ffffffff8139f0dc>] ? __kmalloc_reserve+0x3c/0x90 [<ffffffff8139fcea>] ? __alloc_skb+0x9a/0x1e0 [<ffffffff8140c47d>] tcp_send_ack+0x10d/0x150 [<ffffffff814060ee>] __tcp_ack_snd_check+0x5e/0x90 [<ffffffff81408eb4>] tcp_rcv_established+0x354/0x710 [<ffffffff81412182>] tcp_v4_do_rcv+0x162/0x3f0 [<ffffffff81414412>] tcp_v4_rcv+0xb22/0xb50 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee350>] ip_local_deliver_finish+0xe0/0x2d0 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee72e>] ip_local_deliver+0xae/0xc0 [<ffffffff813edeaf>] ip_rcv_finish+0x14f/0x510 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff813eea13>] ip_rcv+0x2d3/0x3b0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff813ab0f3>] __netif_receive_skb_core+0x663/0xb70 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff810971a9>] ? cpuacct_charge+0x99/0xb0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813ab72c>] ? process_backlog+0xbc/0x150 [<ffffffff813ab78b>] ? process_backlog+0x11b/0x150 [<ffffffff813ab627>] __netif_receive_skb+0x27/0x70 [<ffffffff813ab702>] process_backlog+0x92/0x150 [<ffffffff813afffd>] net_rx_action+0x13d/0x350 [<ffffffff81036b2d>] ? lapic_next_event+0x1d/0x30 [<ffffffff81058694>] __do_softirq+0x104/0x320 [<ffffffff810c0788>] ? hrtimer_interrupt+0xc8/0x1a0 [<ffffffff81074e70>] ? blocking_notifier_chain_cond_register+0x70/0x70 [<ffffffff81058ab9>] irq_exit+0x79/0xa0 [<ffffffff8147ecca>] smp_apic_timer_interrupt+0x4a/0x60 [<ffffffff8147d2c8>] apic_timer_interrupt+0x68/0x70 <EOI> [<ffffffff81271c02>] ? __memcpy+0x12/0x20 [<ffffffffa01517da>] ? visorchannel_write+0x4a/0x80 [visorbus] [<ffffffffa0151eb8>] signalinsert_inner+0x88/0x130 [visorbus] [<ffffffffa0151fb5>] visorchannel_signalinsert+0x55/0x70 [visorbus] [<ffffffffa0166e57>] visornic_rx+0x12e7/0x19d0 [visornic] [<ffffffffa01677c9>] process_incoming_rsps+0x289/0x690 [visornic] [<ffffffff814771c5>] ? preempt_schedule+0x25/0x30 [<ffffffff81001026>] ? ___preempt_schedule+0x12/0x14 [<ffffffff81093080>] ? wait_woken+0x90/0x90 [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffff81073a39>] kthread+0xe9/0x110 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 [<ffffffff8147c89f>] ret_from_fork+0x3f/0x70 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 Fixes: b12fdf7da ('staging: unisys: rework signal remove/insert to avoid sparse lock warnings') Signed-off-by: Tim Sell <Timothy.Sell@unisys.com> Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-07-14 12:43:30 -06:00
spin_unlock_irqrestore(&channel->remove_lock, flags);
} else {
rc = signalremove_inner(channel, queue, msg);
}
return rc;
}
EXPORT_SYMBOL_GPL(visorchannel_signalremove);
static bool queue_empty(struct visorchannel *channel, u32 queue)
{
struct signal_queue_header sig_hdr;
if (sig_read_header(channel, queue, &sig_hdr))
return true;
return (sig_hdr.head == sig_hdr.tail);
}
/**
* visorchannel_signalempty() - checks if the designated channel/queue contains
* any messages
* @channel: the channel to query
* @queue: the queue in the channel to query
*
* Return: boolean indicating whether any messages in the designated
* channel/queue are present
*/
bool visorchannel_signalempty(struct visorchannel *channel, u32 queue)
{
bool rc;
unsigned long flags;
if (!channel->needs_lock)
return queue_empty(channel, queue);
spin_lock_irqsave(&channel->remove_lock, flags);
rc = queue_empty(channel, queue);
spin_unlock_irqrestore(&channel->remove_lock, flags);
return rc;
}
EXPORT_SYMBOL_GPL(visorchannel_signalempty);
static int signalinsert_inner(struct visorchannel *channel, u32 queue,
void *msg)
{
struct signal_queue_header sig_hdr;
int err;
err = sig_read_header(channel, queue, &sig_hdr);
if (err)
return err;
sig_hdr.head = (sig_hdr.head + 1) % sig_hdr.max_slots;
if (sig_hdr.head == sig_hdr.tail) {
sig_hdr.num_overflows++;
err = SIG_WRITE_FIELD(channel, queue, &sig_hdr, num_overflows);
if (err)
return err;
return -EIO;
}
err = sig_write_data(channel, queue, &sig_hdr, sig_hdr.head, msg);
if (err)
return err;
sig_hdr.num_sent++;
/*
* For each data field in SIGNAL_QUEUE_HEADER that was modified, update
* host memory. Required for channel sync.
*/
mb();
err = SIG_WRITE_FIELD(channel, queue, &sig_hdr, head);
if (err)
return err;
err = SIG_WRITE_FIELD(channel, queue, &sig_hdr, num_sent);
if (err)
return err;
return 0;
}
/*
* visorchannel_create() - creates the struct visorchannel abstraction for a
* data area in memory, but does NOT modify this data
* area
* @physaddr: physical address of start of channel
* @gfp: gfp_t to use when allocating memory for the data struct
* @guid: GUID that identifies channel type;
* @needs_lock: must specify true if you have multiple threads of execution
* that will be calling visorchannel methods of this
* visorchannel at the same time
*
* Return: pointer to visorchannel that was created if successful,
* otherwise NULL
*/
struct visorchannel *visorchannel_create(u64 physaddr, gfp_t gfp,
const guid_t *guid, bool needs_lock)
{
struct visorchannel *channel;
int err;
size_t size = sizeof(struct channel_header);
if (physaddr == 0)
return NULL;
channel = kzalloc(sizeof(*channel), gfp);
if (!channel)
return NULL;
channel->needs_lock = needs_lock;
spin_lock_init(&channel->insert_lock);
spin_lock_init(&channel->remove_lock);
/*
* Video driver constains the efi framebuffer so it will get a conflict
* resource when requesting its full mem region. Since we are only
* using the efi framebuffer for video we can ignore this. Remember that
* we haven't requested it so we don't try to release later on.
*/
channel->requested = request_mem_region(physaddr, size, VISOR_DRV_NAME);
if (!channel->requested && !guid_equal(guid, &visor_video_guid))
/* we only care about errors if this is not the video channel */
goto err_destroy_channel;
channel->mapped = memremap(physaddr, size, MEMREMAP_WB);
if (!channel->mapped) {
release_mem_region(physaddr, size);
goto err_destroy_channel;
}
channel->physaddr = physaddr;
channel->nbytes = size;
err = visorchannel_read(channel, 0, &channel->chan_hdr, size);
if (err)
goto err_destroy_channel;
size = (ulong)channel->chan_hdr.size;
memunmap(channel->mapped);
if (channel->requested)
release_mem_region(channel->physaddr, channel->nbytes);
channel->mapped = NULL;
channel->requested = request_mem_region(channel->physaddr, size,
VISOR_DRV_NAME);
if (!channel->requested && !guid_equal(guid, &visor_video_guid))
/* we only care about errors if this is not the video channel */
goto err_destroy_channel;
channel->mapped = memremap(channel->physaddr, size, MEMREMAP_WB);
if (!channel->mapped) {
release_mem_region(channel->physaddr, size);
goto err_destroy_channel;
}
channel->nbytes = size;
guid_copy(&channel->guid, guid);
return channel;
err_destroy_channel:
visorchannel_destroy(channel);
return NULL;
}
/**
* visorchannel_signalinsert() - inserts a message into the designated
* channel/queue
* @channel: the channel the message will be added to
* @queue: the queue the message will be added to
* @msg: the message to insert
*
* Return: integer error code indicating the status of the insertion
*/
int visorchannel_signalinsert(struct visorchannel *channel, u32 queue,
void *msg)
{
int rc;
staging: unisys: fix random hangs with network stress in visornic We learned that it was possible for the core networking code to call visornic_xmit() within ISR context, resulting in the need for us to use spin_lock_irqsave() / spin_lock_irqrestore() to lock accesses to our virtual device channels. Without the correct locking added in this patch, random hangs would occur on typical kernels while stressing the netork. When using a kernel with CONFIG_DEBUG_SPINLOCK=y, a stackdump would occur at the time of the hang reporting: BUG: spinlock recursion on CPU#0, vnic_incoming/<pid> (see below for more details) We considered the possibility of adding a protocol between a visordriver and visorbus where the visordriver could specify which type of locking it required for its virtual device channels (essentially indicating whether or not it was possible for the channel to be accessed in ISR context), but decided this extra complexity was NOT needed, and that channel queues should always be accessed with the most-stringent locking. So that is what is implemented in this commit. Below is an example stackdump illustrating the spinlock recursion that is fixed by this commit. Note that we are first in virtnic_rx() writing to the device channel when an APIC timer interrupt occurs. Within the core networking code, net_rx_action() calls process_backlog(), which eventually lands up back up in virtnic_xmit() in the code attempting to also write to the device channel. BUG: spinlock recursion on CPU#0, vnic_incoming/262 lock: 0xffff88002db810c0, .magic: dead4ead, .owner: vnic_incoming/262, .owner_cpu: 0 CPU: 0 PID: 262 Comm: vnic_incoming Tainted: G C 4.2.0-rc1-ARCH+ #56 Hardware name: Dell Inc. PowerEdge T110/ , BIOS 1.23 12/15/2009 ffff8800216ac200 ffff88002c803388 ffffffff81476364 0000000000000106 ffff88002db810c0 ffff88002c8033a8 ffffffff8109e2bc ffff88002db810c0 ffffffff817631d4 ffff88002c8033c8 ffffffff8109e330 ffff88002db810c0 Call Trace: <IRQ> [<ffffffff81476364>] dump_stack+0x4f/0x73 [<ffffffff8109e2bc>] spin_dump+0x7c/0xc0 [<ffffffff8109e330>] spin_bug+0x30/0x40 [<ffffffff8109e547>] do_raw_spin_lock+0x127/0x140 [<ffffffff8147bad0>] _raw_spin_lock+0x40/0x50 [<ffffffffa0151fa6>] ? visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa0151fa6>] visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa01683a2>] visornic_xmit+0x302/0x5d0 [visornic] [<ffffffff813b2f30>] dev_hard_start_xmit+0x2e0/0x510 [<ffffffff813b2b75>] ? validate_xmit_skb+0x235/0x310 [<ffffffff813d79e7>] sch_direct_xmit+0xf7/0x1d0 [<ffffffff813b34d3>] __dev_queue_xmit+0x203/0x640 [<ffffffff813b3320>] ? __dev_queue_xmit+0x50/0x640 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff813b3933>] dev_queue_xmit_sk+0x13/0x20 [<ffffffff813f3a5c>] ip_finish_output2+0x22c/0x470 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813f3f6f>] ip_finish_output+0x1df/0x310 [<ffffffff813f4c31>] ip_output+0xb1/0x100 [<ffffffff813f41be>] ip_local_out_sk+0x3e/0x80 [<ffffffff813f4388>] ip_queue_xmit+0x188/0x4a0 [<ffffffff813f4200>] ? ip_local_out_sk+0x80/0x80 [<ffffffff8139fcd6>] ? __alloc_skb+0x86/0x1e0 [<ffffffff8140bd5b>] tcp_transmit_skb+0x4cb/0x9c0 [<ffffffff8139f0dc>] ? __kmalloc_reserve+0x3c/0x90 [<ffffffff8139fcea>] ? __alloc_skb+0x9a/0x1e0 [<ffffffff8140c47d>] tcp_send_ack+0x10d/0x150 [<ffffffff814060ee>] __tcp_ack_snd_check+0x5e/0x90 [<ffffffff81408eb4>] tcp_rcv_established+0x354/0x710 [<ffffffff81412182>] tcp_v4_do_rcv+0x162/0x3f0 [<ffffffff81414412>] tcp_v4_rcv+0xb22/0xb50 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee350>] ip_local_deliver_finish+0xe0/0x2d0 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee72e>] ip_local_deliver+0xae/0xc0 [<ffffffff813edeaf>] ip_rcv_finish+0x14f/0x510 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff813eea13>] ip_rcv+0x2d3/0x3b0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff813ab0f3>] __netif_receive_skb_core+0x663/0xb70 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff810971a9>] ? cpuacct_charge+0x99/0xb0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813ab72c>] ? process_backlog+0xbc/0x150 [<ffffffff813ab78b>] ? process_backlog+0x11b/0x150 [<ffffffff813ab627>] __netif_receive_skb+0x27/0x70 [<ffffffff813ab702>] process_backlog+0x92/0x150 [<ffffffff813afffd>] net_rx_action+0x13d/0x350 [<ffffffff81036b2d>] ? lapic_next_event+0x1d/0x30 [<ffffffff81058694>] __do_softirq+0x104/0x320 [<ffffffff810c0788>] ? hrtimer_interrupt+0xc8/0x1a0 [<ffffffff81074e70>] ? blocking_notifier_chain_cond_register+0x70/0x70 [<ffffffff81058ab9>] irq_exit+0x79/0xa0 [<ffffffff8147ecca>] smp_apic_timer_interrupt+0x4a/0x60 [<ffffffff8147d2c8>] apic_timer_interrupt+0x68/0x70 <EOI> [<ffffffff81271c02>] ? __memcpy+0x12/0x20 [<ffffffffa01517da>] ? visorchannel_write+0x4a/0x80 [visorbus] [<ffffffffa0151eb8>] signalinsert_inner+0x88/0x130 [visorbus] [<ffffffffa0151fb5>] visorchannel_signalinsert+0x55/0x70 [visorbus] [<ffffffffa0166e57>] visornic_rx+0x12e7/0x19d0 [visornic] [<ffffffffa01677c9>] process_incoming_rsps+0x289/0x690 [visornic] [<ffffffff814771c5>] ? preempt_schedule+0x25/0x30 [<ffffffff81001026>] ? ___preempt_schedule+0x12/0x14 [<ffffffff81093080>] ? wait_woken+0x90/0x90 [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffff81073a39>] kthread+0xe9/0x110 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 [<ffffffff8147c89f>] ret_from_fork+0x3f/0x70 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 Fixes: b12fdf7da ('staging: unisys: rework signal remove/insert to avoid sparse lock warnings') Signed-off-by: Tim Sell <Timothy.Sell@unisys.com> Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-07-14 12:43:30 -06:00
unsigned long flags;
if (channel->needs_lock) {
staging: unisys: fix random hangs with network stress in visornic We learned that it was possible for the core networking code to call visornic_xmit() within ISR context, resulting in the need for us to use spin_lock_irqsave() / spin_lock_irqrestore() to lock accesses to our virtual device channels. Without the correct locking added in this patch, random hangs would occur on typical kernels while stressing the netork. When using a kernel with CONFIG_DEBUG_SPINLOCK=y, a stackdump would occur at the time of the hang reporting: BUG: spinlock recursion on CPU#0, vnic_incoming/<pid> (see below for more details) We considered the possibility of adding a protocol between a visordriver and visorbus where the visordriver could specify which type of locking it required for its virtual device channels (essentially indicating whether or not it was possible for the channel to be accessed in ISR context), but decided this extra complexity was NOT needed, and that channel queues should always be accessed with the most-stringent locking. So that is what is implemented in this commit. Below is an example stackdump illustrating the spinlock recursion that is fixed by this commit. Note that we are first in virtnic_rx() writing to the device channel when an APIC timer interrupt occurs. Within the core networking code, net_rx_action() calls process_backlog(), which eventually lands up back up in virtnic_xmit() in the code attempting to also write to the device channel. BUG: spinlock recursion on CPU#0, vnic_incoming/262 lock: 0xffff88002db810c0, .magic: dead4ead, .owner: vnic_incoming/262, .owner_cpu: 0 CPU: 0 PID: 262 Comm: vnic_incoming Tainted: G C 4.2.0-rc1-ARCH+ #56 Hardware name: Dell Inc. PowerEdge T110/ , BIOS 1.23 12/15/2009 ffff8800216ac200 ffff88002c803388 ffffffff81476364 0000000000000106 ffff88002db810c0 ffff88002c8033a8 ffffffff8109e2bc ffff88002db810c0 ffffffff817631d4 ffff88002c8033c8 ffffffff8109e330 ffff88002db810c0 Call Trace: <IRQ> [<ffffffff81476364>] dump_stack+0x4f/0x73 [<ffffffff8109e2bc>] spin_dump+0x7c/0xc0 [<ffffffff8109e330>] spin_bug+0x30/0x40 [<ffffffff8109e547>] do_raw_spin_lock+0x127/0x140 [<ffffffff8147bad0>] _raw_spin_lock+0x40/0x50 [<ffffffffa0151fa6>] ? visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa0151fa6>] visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa01683a2>] visornic_xmit+0x302/0x5d0 [visornic] [<ffffffff813b2f30>] dev_hard_start_xmit+0x2e0/0x510 [<ffffffff813b2b75>] ? validate_xmit_skb+0x235/0x310 [<ffffffff813d79e7>] sch_direct_xmit+0xf7/0x1d0 [<ffffffff813b34d3>] __dev_queue_xmit+0x203/0x640 [<ffffffff813b3320>] ? __dev_queue_xmit+0x50/0x640 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff813b3933>] dev_queue_xmit_sk+0x13/0x20 [<ffffffff813f3a5c>] ip_finish_output2+0x22c/0x470 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813f3f6f>] ip_finish_output+0x1df/0x310 [<ffffffff813f4c31>] ip_output+0xb1/0x100 [<ffffffff813f41be>] ip_local_out_sk+0x3e/0x80 [<ffffffff813f4388>] ip_queue_xmit+0x188/0x4a0 [<ffffffff813f4200>] ? ip_local_out_sk+0x80/0x80 [<ffffffff8139fcd6>] ? __alloc_skb+0x86/0x1e0 [<ffffffff8140bd5b>] tcp_transmit_skb+0x4cb/0x9c0 [<ffffffff8139f0dc>] ? __kmalloc_reserve+0x3c/0x90 [<ffffffff8139fcea>] ? __alloc_skb+0x9a/0x1e0 [<ffffffff8140c47d>] tcp_send_ack+0x10d/0x150 [<ffffffff814060ee>] __tcp_ack_snd_check+0x5e/0x90 [<ffffffff81408eb4>] tcp_rcv_established+0x354/0x710 [<ffffffff81412182>] tcp_v4_do_rcv+0x162/0x3f0 [<ffffffff81414412>] tcp_v4_rcv+0xb22/0xb50 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee350>] ip_local_deliver_finish+0xe0/0x2d0 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee72e>] ip_local_deliver+0xae/0xc0 [<ffffffff813edeaf>] ip_rcv_finish+0x14f/0x510 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff813eea13>] ip_rcv+0x2d3/0x3b0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff813ab0f3>] __netif_receive_skb_core+0x663/0xb70 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff810971a9>] ? cpuacct_charge+0x99/0xb0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813ab72c>] ? process_backlog+0xbc/0x150 [<ffffffff813ab78b>] ? process_backlog+0x11b/0x150 [<ffffffff813ab627>] __netif_receive_skb+0x27/0x70 [<ffffffff813ab702>] process_backlog+0x92/0x150 [<ffffffff813afffd>] net_rx_action+0x13d/0x350 [<ffffffff81036b2d>] ? lapic_next_event+0x1d/0x30 [<ffffffff81058694>] __do_softirq+0x104/0x320 [<ffffffff810c0788>] ? hrtimer_interrupt+0xc8/0x1a0 [<ffffffff81074e70>] ? blocking_notifier_chain_cond_register+0x70/0x70 [<ffffffff81058ab9>] irq_exit+0x79/0xa0 [<ffffffff8147ecca>] smp_apic_timer_interrupt+0x4a/0x60 [<ffffffff8147d2c8>] apic_timer_interrupt+0x68/0x70 <EOI> [<ffffffff81271c02>] ? __memcpy+0x12/0x20 [<ffffffffa01517da>] ? visorchannel_write+0x4a/0x80 [visorbus] [<ffffffffa0151eb8>] signalinsert_inner+0x88/0x130 [visorbus] [<ffffffffa0151fb5>] visorchannel_signalinsert+0x55/0x70 [visorbus] [<ffffffffa0166e57>] visornic_rx+0x12e7/0x19d0 [visornic] [<ffffffffa01677c9>] process_incoming_rsps+0x289/0x690 [visornic] [<ffffffff814771c5>] ? preempt_schedule+0x25/0x30 [<ffffffff81001026>] ? ___preempt_schedule+0x12/0x14 [<ffffffff81093080>] ? wait_woken+0x90/0x90 [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffff81073a39>] kthread+0xe9/0x110 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 [<ffffffff8147c89f>] ret_from_fork+0x3f/0x70 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 Fixes: b12fdf7da ('staging: unisys: rework signal remove/insert to avoid sparse lock warnings') Signed-off-by: Tim Sell <Timothy.Sell@unisys.com> Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-07-14 12:43:30 -06:00
spin_lock_irqsave(&channel->insert_lock, flags);
rc = signalinsert_inner(channel, queue, msg);
staging: unisys: fix random hangs with network stress in visornic We learned that it was possible for the core networking code to call visornic_xmit() within ISR context, resulting in the need for us to use spin_lock_irqsave() / spin_lock_irqrestore() to lock accesses to our virtual device channels. Without the correct locking added in this patch, random hangs would occur on typical kernels while stressing the netork. When using a kernel with CONFIG_DEBUG_SPINLOCK=y, a stackdump would occur at the time of the hang reporting: BUG: spinlock recursion on CPU#0, vnic_incoming/<pid> (see below for more details) We considered the possibility of adding a protocol between a visordriver and visorbus where the visordriver could specify which type of locking it required for its virtual device channels (essentially indicating whether or not it was possible for the channel to be accessed in ISR context), but decided this extra complexity was NOT needed, and that channel queues should always be accessed with the most-stringent locking. So that is what is implemented in this commit. Below is an example stackdump illustrating the spinlock recursion that is fixed by this commit. Note that we are first in virtnic_rx() writing to the device channel when an APIC timer interrupt occurs. Within the core networking code, net_rx_action() calls process_backlog(), which eventually lands up back up in virtnic_xmit() in the code attempting to also write to the device channel. BUG: spinlock recursion on CPU#0, vnic_incoming/262 lock: 0xffff88002db810c0, .magic: dead4ead, .owner: vnic_incoming/262, .owner_cpu: 0 CPU: 0 PID: 262 Comm: vnic_incoming Tainted: G C 4.2.0-rc1-ARCH+ #56 Hardware name: Dell Inc. PowerEdge T110/ , BIOS 1.23 12/15/2009 ffff8800216ac200 ffff88002c803388 ffffffff81476364 0000000000000106 ffff88002db810c0 ffff88002c8033a8 ffffffff8109e2bc ffff88002db810c0 ffffffff817631d4 ffff88002c8033c8 ffffffff8109e330 ffff88002db810c0 Call Trace: <IRQ> [<ffffffff81476364>] dump_stack+0x4f/0x73 [<ffffffff8109e2bc>] spin_dump+0x7c/0xc0 [<ffffffff8109e330>] spin_bug+0x30/0x40 [<ffffffff8109e547>] do_raw_spin_lock+0x127/0x140 [<ffffffff8147bad0>] _raw_spin_lock+0x40/0x50 [<ffffffffa0151fa6>] ? visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa0151fa6>] visorchannel_signalinsert+0x46/0x70 [visorbus] [<ffffffffa01683a2>] visornic_xmit+0x302/0x5d0 [visornic] [<ffffffff813b2f30>] dev_hard_start_xmit+0x2e0/0x510 [<ffffffff813b2b75>] ? validate_xmit_skb+0x235/0x310 [<ffffffff813d79e7>] sch_direct_xmit+0xf7/0x1d0 [<ffffffff813b34d3>] __dev_queue_xmit+0x203/0x640 [<ffffffff813b3320>] ? __dev_queue_xmit+0x50/0x640 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff813b3933>] dev_queue_xmit_sk+0x13/0x20 [<ffffffff813f3a5c>] ip_finish_output2+0x22c/0x470 [<ffffffff813f3f6f>] ? ip_finish_output+0x1df/0x310 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813f3f6f>] ip_finish_output+0x1df/0x310 [<ffffffff813f4c31>] ip_output+0xb1/0x100 [<ffffffff813f41be>] ip_local_out_sk+0x3e/0x80 [<ffffffff813f4388>] ip_queue_xmit+0x188/0x4a0 [<ffffffff813f4200>] ? ip_local_out_sk+0x80/0x80 [<ffffffff8139fcd6>] ? __alloc_skb+0x86/0x1e0 [<ffffffff8140bd5b>] tcp_transmit_skb+0x4cb/0x9c0 [<ffffffff8139f0dc>] ? __kmalloc_reserve+0x3c/0x90 [<ffffffff8139fcea>] ? __alloc_skb+0x9a/0x1e0 [<ffffffff8140c47d>] tcp_send_ack+0x10d/0x150 [<ffffffff814060ee>] __tcp_ack_snd_check+0x5e/0x90 [<ffffffff81408eb4>] tcp_rcv_established+0x354/0x710 [<ffffffff81412182>] tcp_v4_do_rcv+0x162/0x3f0 [<ffffffff81414412>] tcp_v4_rcv+0xb22/0xb50 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee350>] ip_local_deliver_finish+0xe0/0x2d0 [<ffffffff813ee2bc>] ? ip_local_deliver_finish+0x4c/0x2d0 [<ffffffff813ee72e>] ip_local_deliver+0xae/0xc0 [<ffffffff813edeaf>] ip_rcv_finish+0x14f/0x510 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff813eea13>] ip_rcv+0x2d3/0x3b0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff813ab0f3>] __netif_receive_skb_core+0x663/0xb70 [<ffffffff813aab2d>] ? __netif_receive_skb_core+0x9d/0xb70 [<ffffffff810971a9>] ? cpuacct_charge+0x99/0xb0 [<ffffffff81097110>] ? cpuacct_css_alloc+0xb0/0xb0 [<ffffffff810987e0>] ? __lock_is_held+0x50/0x70 [<ffffffff813ab72c>] ? process_backlog+0xbc/0x150 [<ffffffff813ab78b>] ? process_backlog+0x11b/0x150 [<ffffffff813ab627>] __netif_receive_skb+0x27/0x70 [<ffffffff813ab702>] process_backlog+0x92/0x150 [<ffffffff813afffd>] net_rx_action+0x13d/0x350 [<ffffffff81036b2d>] ? lapic_next_event+0x1d/0x30 [<ffffffff81058694>] __do_softirq+0x104/0x320 [<ffffffff810c0788>] ? hrtimer_interrupt+0xc8/0x1a0 [<ffffffff81074e70>] ? blocking_notifier_chain_cond_register+0x70/0x70 [<ffffffff81058ab9>] irq_exit+0x79/0xa0 [<ffffffff8147ecca>] smp_apic_timer_interrupt+0x4a/0x60 [<ffffffff8147d2c8>] apic_timer_interrupt+0x68/0x70 <EOI> [<ffffffff81271c02>] ? __memcpy+0x12/0x20 [<ffffffffa01517da>] ? visorchannel_write+0x4a/0x80 [visorbus] [<ffffffffa0151eb8>] signalinsert_inner+0x88/0x130 [visorbus] [<ffffffffa0151fb5>] visorchannel_signalinsert+0x55/0x70 [visorbus] [<ffffffffa0166e57>] visornic_rx+0x12e7/0x19d0 [visornic] [<ffffffffa01677c9>] process_incoming_rsps+0x289/0x690 [visornic] [<ffffffff814771c5>] ? preempt_schedule+0x25/0x30 [<ffffffff81001026>] ? ___preempt_schedule+0x12/0x14 [<ffffffff81093080>] ? wait_woken+0x90/0x90 [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffffa0167540>] ? visornic_rx+0x19d0/0x19d0 [visornic] [<ffffffff81073a39>] kthread+0xe9/0x110 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 [<ffffffff8147c89f>] ret_from_fork+0x3f/0x70 [<ffffffff81073950>] ? __init_kthread_worker+0x70/0x70 Fixes: b12fdf7da ('staging: unisys: rework signal remove/insert to avoid sparse lock warnings') Signed-off-by: Tim Sell <Timothy.Sell@unisys.com> Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-07-14 12:43:30 -06:00
spin_unlock_irqrestore(&channel->insert_lock, flags);
} else {
rc = signalinsert_inner(channel, queue, msg);
}
return rc;
}
EXPORT_SYMBOL_GPL(visorchannel_signalinsert);