2
0
mirror of https://github.com/edk2-porting/linux-next.git synced 2024-12-22 12:14:01 +08:00
linux-next/drivers/firewire/fw-topology.c
Stefan Richter 792a61021c firewire: fix race of bus reset with request transmission
Reported by Jay Fenlason:  A bus reset tasklet may call
fw_flush_transactions and touch transactions (call their callback which
will free them) while the context which submitted the transaction is
still inserting it into the transmission queue.

A simple solution to this problem is to _not_ "flush" the transactions
because of a bus reset (complete the transcations as 'cancelled').  They
will now simply time out (completed as 'cancelled' by the split-timeout
timer).

Jay Fenlason thought of this fix too but I was quicker to type it out.
:-)

Background:
Contexts which access an instance of struct fw_transaction are:
 1. the submitter, until it inserted the packet which is embedded in the
    transaction into the AT req DMA,
 2. the AsReqTrContext tasklet when the request packet was acked by the
    responder node or transmission to the responder failed,
 3. the AsRspRcvContext tasklet when it found a request which matched
    an incoming response,
 4. the card->flush_timer when it picks up timed-out transactions to
    cancel them,
 5. the bus reset tasklet when it cancels transactions (this access is
    eliminated by this patch),
 6. a process which shuts down an fw_card (unregisters it from fw-core
    when the controller is unbound from fw-ohci) --- although in this
    case there shouldn't really be any transactions anymore because we
    wait until all card users finished their business with the card.

All of these contexts run concurrently (except for the 6th, presumably).
The 1st is safe against the 2nd and 3rd because of the way how a request
packet is carefully submitted to the hardware.  A race between 2nd and
3rd has been fixed a while ago (bug 9617).  The 4th is almost safe
against 1st, 2nd, 3rd;  there are issues with it if huge scheduling
latencies occur, to be fixed separately.  The 5th looks safe against
2nd, 3rd, and 4th but is unsafe against 1st.  Maybe this could be fixed
with an explicit state variable in struct fw_transaction.  But this
would require fw_transaction to be rewritten as only dynamically
allocatable object with reference counting --- not a good solution if we
also can simply kill this 5th accessing context (replace it by the 4th).

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
2008-07-14 13:06:04 +02:00

552 lines
14 KiB
C

/*
* Incremental bus scan, based on bus topology
*
* Copyright (C) 2004-2006 Kristian Hoegsberg <krh@bitplanet.net>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*/
#include <linux/module.h>
#include <linux/wait.h>
#include <linux/errno.h>
#include <asm/bug.h>
#include <asm/system.h>
#include "fw-transaction.h"
#include "fw-topology.h"
#define SELF_ID_PHY_ID(q) (((q) >> 24) & 0x3f)
#define SELF_ID_EXTENDED(q) (((q) >> 23) & 0x01)
#define SELF_ID_LINK_ON(q) (((q) >> 22) & 0x01)
#define SELF_ID_GAP_COUNT(q) (((q) >> 16) & 0x3f)
#define SELF_ID_PHY_SPEED(q) (((q) >> 14) & 0x03)
#define SELF_ID_CONTENDER(q) (((q) >> 11) & 0x01)
#define SELF_ID_PHY_INITIATOR(q) (((q) >> 1) & 0x01)
#define SELF_ID_MORE_PACKETS(q) (((q) >> 0) & 0x01)
#define SELF_ID_EXT_SEQUENCE(q) (((q) >> 20) & 0x07)
static u32 *count_ports(u32 *sid, int *total_port_count, int *child_port_count)
{
u32 q;
int port_type, shift, seq;
*total_port_count = 0;
*child_port_count = 0;
shift = 6;
q = *sid;
seq = 0;
while (1) {
port_type = (q >> shift) & 0x03;
switch (port_type) {
case SELFID_PORT_CHILD:
(*child_port_count)++;
case SELFID_PORT_PARENT:
case SELFID_PORT_NCONN:
(*total_port_count)++;
case SELFID_PORT_NONE:
break;
}
shift -= 2;
if (shift == 0) {
if (!SELF_ID_MORE_PACKETS(q))
return sid + 1;
shift = 16;
sid++;
q = *sid;
/*
* Check that the extra packets actually are
* extended self ID packets and that the
* sequence numbers in the extended self ID
* packets increase as expected.
*/
if (!SELF_ID_EXTENDED(q) ||
seq != SELF_ID_EXT_SEQUENCE(q))
return NULL;
seq++;
}
}
}
static int get_port_type(u32 *sid, int port_index)
{
int index, shift;
index = (port_index + 5) / 8;
shift = 16 - ((port_index + 5) & 7) * 2;
return (sid[index] >> shift) & 0x03;
}
static struct fw_node *fw_node_create(u32 sid, int port_count, int color)
{
struct fw_node *node;
node = kzalloc(sizeof(*node) + port_count * sizeof(node->ports[0]),
GFP_ATOMIC);
if (node == NULL)
return NULL;
node->color = color;
node->node_id = LOCAL_BUS | SELF_ID_PHY_ID(sid);
node->link_on = SELF_ID_LINK_ON(sid);
node->phy_speed = SELF_ID_PHY_SPEED(sid);
node->initiated_reset = SELF_ID_PHY_INITIATOR(sid);
node->port_count = port_count;
atomic_set(&node->ref_count, 1);
INIT_LIST_HEAD(&node->link);
return node;
}
/*
* Compute the maximum hop count for this node and it's children. The
* maximum hop count is the maximum number of connections between any
* two nodes in the subtree rooted at this node. We need this for
* setting the gap count. As we build the tree bottom up in
* build_tree() below, this is fairly easy to do: for each node we
* maintain the max hop count and the max depth, ie the number of hops
* to the furthest leaf. Computing the max hop count breaks down into
* two cases: either the path goes through this node, in which case
* the hop count is the sum of the two biggest child depths plus 2.
* Or it could be the case that the max hop path is entirely
* containted in a child tree, in which case the max hop count is just
* the max hop count of this child.
*/
static void update_hop_count(struct fw_node *node)
{
int depths[2] = { -1, -1 };
int max_child_hops = 0;
int i;
for (i = 0; i < node->port_count; i++) {
if (node->ports[i] == NULL)
continue;
if (node->ports[i]->max_hops > max_child_hops)
max_child_hops = node->ports[i]->max_hops;
if (node->ports[i]->max_depth > depths[0]) {
depths[1] = depths[0];
depths[0] = node->ports[i]->max_depth;
} else if (node->ports[i]->max_depth > depths[1])
depths[1] = node->ports[i]->max_depth;
}
node->max_depth = depths[0] + 1;
node->max_hops = max(max_child_hops, depths[0] + depths[1] + 2);
}
static inline struct fw_node *fw_node(struct list_head *l)
{
return list_entry(l, struct fw_node, link);
}
/**
* build_tree - Build the tree representation of the topology
* @self_ids: array of self IDs to create the tree from
* @self_id_count: the length of the self_ids array
* @local_id: the node ID of the local node
*
* This function builds the tree representation of the topology given
* by the self IDs from the latest bus reset. During the construction
* of the tree, the function checks that the self IDs are valid and
* internally consistent. On succcess this function returns the
* fw_node corresponding to the local card otherwise NULL.
*/
static struct fw_node *build_tree(struct fw_card *card,
u32 *sid, int self_id_count)
{
struct fw_node *node, *child, *local_node, *irm_node;
struct list_head stack, *h;
u32 *next_sid, *end, q;
int i, port_count, child_port_count, phy_id, parent_count, stack_depth;
int gap_count;
bool beta_repeaters_present;
local_node = NULL;
node = NULL;
INIT_LIST_HEAD(&stack);
stack_depth = 0;
end = sid + self_id_count;
phy_id = 0;
irm_node = NULL;
gap_count = SELF_ID_GAP_COUNT(*sid);
beta_repeaters_present = false;
while (sid < end) {
next_sid = count_ports(sid, &port_count, &child_port_count);
if (next_sid == NULL) {
fw_error("Inconsistent extended self IDs.\n");
return NULL;
}
q = *sid;
if (phy_id != SELF_ID_PHY_ID(q)) {
fw_error("PHY ID mismatch in self ID: %d != %d.\n",
phy_id, SELF_ID_PHY_ID(q));
return NULL;
}
if (child_port_count > stack_depth) {
fw_error("Topology stack underflow\n");
return NULL;
}
/*
* Seek back from the top of our stack to find the
* start of the child nodes for this node.
*/
for (i = 0, h = &stack; i < child_port_count; i++)
h = h->prev;
/*
* When the stack is empty, this yields an invalid value,
* but that pointer will never be dereferenced.
*/
child = fw_node(h);
node = fw_node_create(q, port_count, card->color);
if (node == NULL) {
fw_error("Out of memory while building topology.\n");
return NULL;
}
if (phy_id == (card->node_id & 0x3f))
local_node = node;
if (SELF_ID_CONTENDER(q))
irm_node = node;
parent_count = 0;
for (i = 0; i < port_count; i++) {
switch (get_port_type(sid, i)) {
case SELFID_PORT_PARENT:
/*
* Who's your daddy? We dont know the
* parent node at this time, so we
* temporarily abuse node->color for
* remembering the entry in the
* node->ports array where the parent
* node should be. Later, when we
* handle the parent node, we fix up
* the reference.
*/
parent_count++;
node->color = i;
break;
case SELFID_PORT_CHILD:
node->ports[i] = child;
/*
* Fix up parent reference for this
* child node.
*/
child->ports[child->color] = node;
child->color = card->color;
child = fw_node(child->link.next);
break;
}
}
/*
* Check that the node reports exactly one parent
* port, except for the root, which of course should
* have no parents.
*/
if ((next_sid == end && parent_count != 0) ||
(next_sid < end && parent_count != 1)) {
fw_error("Parent port inconsistency for node %d: "
"parent_count=%d\n", phy_id, parent_count);
return NULL;
}
/* Pop the child nodes off the stack and push the new node. */
__list_del(h->prev, &stack);
list_add_tail(&node->link, &stack);
stack_depth += 1 - child_port_count;
if (node->phy_speed == SCODE_BETA &&
parent_count + child_port_count > 1)
beta_repeaters_present = true;
/*
* If PHYs report different gap counts, set an invalid count
* which will force a gap count reconfiguration and a reset.
*/
if (SELF_ID_GAP_COUNT(q) != gap_count)
gap_count = 0;
update_hop_count(node);
sid = next_sid;
phy_id++;
}
card->root_node = node;
card->irm_node = irm_node;
card->gap_count = gap_count;
card->beta_repeaters_present = beta_repeaters_present;
return local_node;
}
typedef void (*fw_node_callback_t)(struct fw_card * card,
struct fw_node * node,
struct fw_node * parent);
static void
for_each_fw_node(struct fw_card *card, struct fw_node *root,
fw_node_callback_t callback)
{
struct list_head list;
struct fw_node *node, *next, *child, *parent;
int i;
INIT_LIST_HEAD(&list);
fw_node_get(root);
list_add_tail(&root->link, &list);
parent = NULL;
list_for_each_entry(node, &list, link) {
node->color = card->color;
for (i = 0; i < node->port_count; i++) {
child = node->ports[i];
if (!child)
continue;
if (child->color == card->color)
parent = child;
else {
fw_node_get(child);
list_add_tail(&child->link, &list);
}
}
callback(card, node, parent);
}
list_for_each_entry_safe(node, next, &list, link)
fw_node_put(node);
}
static void
report_lost_node(struct fw_card *card,
struct fw_node *node, struct fw_node *parent)
{
fw_node_event(card, node, FW_NODE_DESTROYED);
fw_node_put(node);
}
static void
report_found_node(struct fw_card *card,
struct fw_node *node, struct fw_node *parent)
{
int b_path = (node->phy_speed == SCODE_BETA);
if (parent != NULL) {
/* min() macro doesn't work here with gcc 3.4 */
node->max_speed = parent->max_speed < node->phy_speed ?
parent->max_speed : node->phy_speed;
node->b_path = parent->b_path && b_path;
} else {
node->max_speed = node->phy_speed;
node->b_path = b_path;
}
fw_node_event(card, node, FW_NODE_CREATED);
}
void fw_destroy_nodes(struct fw_card *card)
{
unsigned long flags;
spin_lock_irqsave(&card->lock, flags);
card->color++;
if (card->local_node != NULL)
for_each_fw_node(card, card->local_node, report_lost_node);
card->local_node = NULL;
spin_unlock_irqrestore(&card->lock, flags);
}
static void move_tree(struct fw_node *node0, struct fw_node *node1, int port)
{
struct fw_node *tree;
int i;
tree = node1->ports[port];
node0->ports[port] = tree;
for (i = 0; i < tree->port_count; i++) {
if (tree->ports[i] == node1) {
tree->ports[i] = node0;
break;
}
}
}
/**
* update_tree - compare the old topology tree for card with the new
* one specified by root. Queue the nodes and mark them as either
* found, lost or updated. Update the nodes in the card topology tree
* as we go.
*/
static void
update_tree(struct fw_card *card, struct fw_node *root)
{
struct list_head list0, list1;
struct fw_node *node0, *node1;
int i, event;
INIT_LIST_HEAD(&list0);
list_add_tail(&card->local_node->link, &list0);
INIT_LIST_HEAD(&list1);
list_add_tail(&root->link, &list1);
node0 = fw_node(list0.next);
node1 = fw_node(list1.next);
while (&node0->link != &list0) {
WARN_ON(node0->port_count != node1->port_count);
if (node0->link_on && !node1->link_on)
event = FW_NODE_LINK_OFF;
else if (!node0->link_on && node1->link_on)
event = FW_NODE_LINK_ON;
else if (node1->initiated_reset && node1->link_on)
event = FW_NODE_INITIATED_RESET;
else
event = FW_NODE_UPDATED;
node0->node_id = node1->node_id;
node0->color = card->color;
node0->link_on = node1->link_on;
node0->initiated_reset = node1->initiated_reset;
node0->max_hops = node1->max_hops;
node1->color = card->color;
fw_node_event(card, node0, event);
if (card->root_node == node1)
card->root_node = node0;
if (card->irm_node == node1)
card->irm_node = node0;
for (i = 0; i < node0->port_count; i++) {
if (node0->ports[i] && node1->ports[i]) {
/*
* This port didn't change, queue the
* connected node for further
* investigation.
*/
if (node0->ports[i]->color == card->color)
continue;
list_add_tail(&node0->ports[i]->link, &list0);
list_add_tail(&node1->ports[i]->link, &list1);
} else if (node0->ports[i]) {
/*
* The nodes connected here were
* unplugged; unref the lost nodes and
* queue FW_NODE_LOST callbacks for
* them.
*/
for_each_fw_node(card, node0->ports[i],
report_lost_node);
node0->ports[i] = NULL;
} else if (node1->ports[i]) {
/*
* One or more node were connected to
* this port. Move the new nodes into
* the tree and queue FW_NODE_CREATED
* callbacks for them.
*/
move_tree(node0, node1, i);
for_each_fw_node(card, node0->ports[i],
report_found_node);
}
}
node0 = fw_node(node0->link.next);
node1 = fw_node(node1->link.next);
}
}
static void
update_topology_map(struct fw_card *card, u32 *self_ids, int self_id_count)
{
int node_count;
card->topology_map[1]++;
node_count = (card->root_node->node_id & 0x3f) + 1;
card->topology_map[2] = (node_count << 16) | self_id_count;
card->topology_map[0] = (self_id_count + 2) << 16;
memcpy(&card->topology_map[3], self_ids, self_id_count * 4);
fw_compute_block_crc(card->topology_map);
}
void
fw_core_handle_bus_reset(struct fw_card *card,
int node_id, int generation,
int self_id_count, u32 * self_ids)
{
struct fw_node *local_node;
unsigned long flags;
spin_lock_irqsave(&card->lock, flags);
/*
* If the new topology has a different self_id_count the topology
* changed, either nodes were added or removed. In that case we
* reset the IRM reset counter.
*/
if (card->self_id_count != self_id_count)
card->bm_retries = 0;
card->node_id = node_id;
/*
* Update node_id before generation to prevent anybody from using
* a stale node_id together with a current generation.
*/
smp_wmb();
card->generation = generation;
card->reset_jiffies = jiffies;
schedule_delayed_work(&card->work, 0);
local_node = build_tree(card, self_ids, self_id_count);
update_topology_map(card, self_ids, self_id_count);
card->color++;
if (local_node == NULL) {
fw_error("topology build failed\n");
/* FIXME: We need to issue a bus reset in this case. */
} else if (card->local_node == NULL) {
card->local_node = local_node;
for_each_fw_node(card, local_node, report_found_node);
} else {
update_tree(card, local_node);
}
spin_unlock_irqrestore(&card->lock, flags);
}
EXPORT_SYMBOL(fw_core_handle_bus_reset);