Revision 63cae12bce9861cec309798d34701cf3da20bc71 authored by Peter Zijlstra on 09 December 2016, 13:59:00 UTC, committed by Ingo Molnar on 14 January 2017, 09:56:10 UTC
There is problem with installing an event in a task that is 'stuck' on an offline CPU. Blocked tasks are not dis-assosciated from offlined CPUs, after all, a blocked task doesn't run and doesn't require a CPU etc.. Only on wakeup do we ammend the situation and place the task on a available CPU. If we hit such a task with perf_install_in_context() we'll loop until either that task wakes up or the CPU comes back online, if the task waking depends on the event being installed, we're stuck. While looking into this issue, I also spotted another problem, if we hit a task with perf_install_in_context() that is in the middle of being migrated, that is we observe the old CPU before sending the IPI, but run the IPI (on the old CPU) while the task is already running on the new CPU, things also go sideways. Rework things to rely on task_curr() -- outside of rq->lock -- which is rather tricky. Imagine the following scenario where we're trying to install the first event into our task 't': CPU0 CPU1 CPU2 (current == t) t->perf_event_ctxp[] = ctx; smp_mb(); cpu = task_cpu(t); switch(t, n); migrate(t, 2); switch(p, t); ctx = t->perf_event_ctxp[]; // must not be NULL smp_function_call(cpu, ..); generic_exec_single() func(); spin_lock(ctx->lock); if (task_curr(t)) // false add_event_to_ctx(); spin_unlock(ctx->lock); perf_event_context_sched_in(); spin_lock(ctx->lock); // sees event So its CPU0's store of t->perf_event_ctxp[] that must not go 'missing'. Because if CPU2's load of that variable were to observe NULL, it would not try to schedule the ctx and we'd have a task running without its counter, which would be 'bad'. As long as we observe !NULL, we'll acquire ctx->lock. If we acquire it first and not see the event yet, then CPU0 must observe task_curr() and retry. If the install happens first, then we must see the event on sched-in and all is well. I think we can translate the first part (until the 'must not be NULL') of the scenario to a litmus test like: C C-peterz { } P0(int *x, int *y) { int r1; WRITE_ONCE(*x, 1); smp_mb(); r1 = READ_ONCE(*y); } P1(int *y, int *z) { WRITE_ONCE(*y, 1); smp_store_release(z, 1); } P2(int *x, int *z) { int r1; int r2; r1 = smp_load_acquire(z); smp_mb(); r2 = READ_ONCE(*x); } exists (0:r1=0 /\ 2:r1=1 /\ 2:r2=0) Where: x is perf_event_ctxp[], y is our tasks's CPU, and z is our task being placed on the rq of CPU2. The P0 smp_mb() is the one added by this patch, ordering the store to perf_event_ctxp[] from find_get_context() and the load of task_cpu() in task_function_call(). The smp_store_release/smp_load_acquire model the RCpc locking of the rq->lock and the smp_mb() of P2 is the context switch switching from whatever CPU2 was running to our task 't'. This litmus test evaluates into: Test C-peterz Allowed States 7 0:r1=0; 2:r1=0; 2:r2=0; 0:r1=0; 2:r1=0; 2:r2=1; 0:r1=0; 2:r1=1; 2:r2=1; 0:r1=1; 2:r1=0; 2:r2=0; 0:r1=1; 2:r1=0; 2:r2=1; 0:r1=1; 2:r1=1; 2:r2=0; 0:r1=1; 2:r1=1; 2:r2=1; No Witnesses Positive: 0 Negative: 7 Condition exists (0:r1=0 /\ 2:r1=1 /\ 2:r2=0) Observation C-peterz Never 0 7 Hash=e427f41d9146b2a5445101d3e2fcaa34 And the strong and weak model agree. Reported-by: Mark Rutland <mark.rutland@arm.com> Tested-by: Mark Rutland <mark.rutland@arm.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Stephane Eranian <eranian@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vince Weaver <vincent.weaver@maine.edu> Cc: Will Deacon <will.deacon@arm.com> Cc: jeremy.linton@arm.com Link: http://lkml.kernel.org/r/20161209135900.GU3174@twins.programming.kicks-ass.net Signed-off-by: Ingo Molnar <mingo@kernel.org>
1 parent ad5013d
hwchannel.c
/*
*
* Author Karsten Keil <kkeil@novell.com>
*
* Copyright 2008 by Karsten Keil <kkeil@novell.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*
* 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.
*
*/
#include <linux/gfp.h>
#include <linux/module.h>
#include <linux/mISDNhw.h>
static void
dchannel_bh(struct work_struct *ws)
{
struct dchannel *dch = container_of(ws, struct dchannel, workq);
struct sk_buff *skb;
int err;
if (test_and_clear_bit(FLG_RECVQUEUE, &dch->Flags)) {
while ((skb = skb_dequeue(&dch->rqueue))) {
if (likely(dch->dev.D.peer)) {
err = dch->dev.D.recv(dch->dev.D.peer, skb);
if (err)
dev_kfree_skb(skb);
} else
dev_kfree_skb(skb);
}
}
if (test_and_clear_bit(FLG_PHCHANGE, &dch->Flags)) {
if (dch->phfunc)
dch->phfunc(dch);
}
}
static void
bchannel_bh(struct work_struct *ws)
{
struct bchannel *bch = container_of(ws, struct bchannel, workq);
struct sk_buff *skb;
int err;
if (test_and_clear_bit(FLG_RECVQUEUE, &bch->Flags)) {
while ((skb = skb_dequeue(&bch->rqueue))) {
bch->rcount--;
if (likely(bch->ch.peer)) {
err = bch->ch.recv(bch->ch.peer, skb);
if (err)
dev_kfree_skb(skb);
} else
dev_kfree_skb(skb);
}
}
}
int
mISDN_initdchannel(struct dchannel *ch, int maxlen, void *phf)
{
test_and_set_bit(FLG_HDLC, &ch->Flags);
ch->maxlen = maxlen;
ch->hw = NULL;
ch->rx_skb = NULL;
ch->tx_skb = NULL;
ch->tx_idx = 0;
ch->phfunc = phf;
skb_queue_head_init(&ch->squeue);
skb_queue_head_init(&ch->rqueue);
INIT_LIST_HEAD(&ch->dev.bchannels);
INIT_WORK(&ch->workq, dchannel_bh);
return 0;
}
EXPORT_SYMBOL(mISDN_initdchannel);
int
mISDN_initbchannel(struct bchannel *ch, unsigned short maxlen,
unsigned short minlen)
{
ch->Flags = 0;
ch->minlen = minlen;
ch->next_minlen = minlen;
ch->init_minlen = minlen;
ch->maxlen = maxlen;
ch->next_maxlen = maxlen;
ch->init_maxlen = maxlen;
ch->hw = NULL;
ch->rx_skb = NULL;
ch->tx_skb = NULL;
ch->tx_idx = 0;
skb_queue_head_init(&ch->rqueue);
ch->rcount = 0;
ch->next_skb = NULL;
INIT_WORK(&ch->workq, bchannel_bh);
return 0;
}
EXPORT_SYMBOL(mISDN_initbchannel);
int
mISDN_freedchannel(struct dchannel *ch)
{
if (ch->tx_skb) {
dev_kfree_skb(ch->tx_skb);
ch->tx_skb = NULL;
}
if (ch->rx_skb) {
dev_kfree_skb(ch->rx_skb);
ch->rx_skb = NULL;
}
skb_queue_purge(&ch->squeue);
skb_queue_purge(&ch->rqueue);
flush_work(&ch->workq);
return 0;
}
EXPORT_SYMBOL(mISDN_freedchannel);
void
mISDN_clear_bchannel(struct bchannel *ch)
{
if (ch->tx_skb) {
dev_kfree_skb(ch->tx_skb);
ch->tx_skb = NULL;
}
ch->tx_idx = 0;
if (ch->rx_skb) {
dev_kfree_skb(ch->rx_skb);
ch->rx_skb = NULL;
}
if (ch->next_skb) {
dev_kfree_skb(ch->next_skb);
ch->next_skb = NULL;
}
test_and_clear_bit(FLG_TX_BUSY, &ch->Flags);
test_and_clear_bit(FLG_TX_NEXT, &ch->Flags);
test_and_clear_bit(FLG_ACTIVE, &ch->Flags);
test_and_clear_bit(FLG_FILLEMPTY, &ch->Flags);
test_and_clear_bit(FLG_TX_EMPTY, &ch->Flags);
test_and_clear_bit(FLG_RX_OFF, &ch->Flags);
ch->dropcnt = 0;
ch->minlen = ch->init_minlen;
ch->next_minlen = ch->init_minlen;
ch->maxlen = ch->init_maxlen;
ch->next_maxlen = ch->init_maxlen;
skb_queue_purge(&ch->rqueue);
ch->rcount = 0;
}
EXPORT_SYMBOL(mISDN_clear_bchannel);
void
mISDN_freebchannel(struct bchannel *ch)
{
cancel_work_sync(&ch->workq);
mISDN_clear_bchannel(ch);
}
EXPORT_SYMBOL(mISDN_freebchannel);
int
mISDN_ctrl_bchannel(struct bchannel *bch, struct mISDN_ctrl_req *cq)
{
int ret = 0;
switch (cq->op) {
case MISDN_CTRL_GETOP:
cq->op = MISDN_CTRL_RX_BUFFER | MISDN_CTRL_FILL_EMPTY |
MISDN_CTRL_RX_OFF;
break;
case MISDN_CTRL_FILL_EMPTY:
if (cq->p1) {
memset(bch->fill, cq->p2 & 0xff, MISDN_BCH_FILL_SIZE);
test_and_set_bit(FLG_FILLEMPTY, &bch->Flags);
} else {
test_and_clear_bit(FLG_FILLEMPTY, &bch->Flags);
}
break;
case MISDN_CTRL_RX_OFF:
/* read back dropped byte count */
cq->p2 = bch->dropcnt;
if (cq->p1)
test_and_set_bit(FLG_RX_OFF, &bch->Flags);
else
test_and_clear_bit(FLG_RX_OFF, &bch->Flags);
bch->dropcnt = 0;
break;
case MISDN_CTRL_RX_BUFFER:
if (cq->p2 > MISDN_CTRL_RX_SIZE_IGNORE)
bch->next_maxlen = cq->p2;
if (cq->p1 > MISDN_CTRL_RX_SIZE_IGNORE)
bch->next_minlen = cq->p1;
/* we return the old values */
cq->p1 = bch->minlen;
cq->p2 = bch->maxlen;
break;
default:
pr_info("mISDN unhandled control %x operation\n", cq->op);
ret = -EINVAL;
break;
}
return ret;
}
EXPORT_SYMBOL(mISDN_ctrl_bchannel);
static inline u_int
get_sapi_tei(u_char *p)
{
u_int sapi, tei;
sapi = *p >> 2;
tei = p[1] >> 1;
return sapi | (tei << 8);
}
void
recv_Dchannel(struct dchannel *dch)
{
struct mISDNhead *hh;
if (dch->rx_skb->len < 2) { /* at least 2 for sapi / tei */
dev_kfree_skb(dch->rx_skb);
dch->rx_skb = NULL;
return;
}
hh = mISDN_HEAD_P(dch->rx_skb);
hh->prim = PH_DATA_IND;
hh->id = get_sapi_tei(dch->rx_skb->data);
skb_queue_tail(&dch->rqueue, dch->rx_skb);
dch->rx_skb = NULL;
schedule_event(dch, FLG_RECVQUEUE);
}
EXPORT_SYMBOL(recv_Dchannel);
void
recv_Echannel(struct dchannel *ech, struct dchannel *dch)
{
struct mISDNhead *hh;
if (ech->rx_skb->len < 2) { /* at least 2 for sapi / tei */
dev_kfree_skb(ech->rx_skb);
ech->rx_skb = NULL;
return;
}
hh = mISDN_HEAD_P(ech->rx_skb);
hh->prim = PH_DATA_E_IND;
hh->id = get_sapi_tei(ech->rx_skb->data);
skb_queue_tail(&dch->rqueue, ech->rx_skb);
ech->rx_skb = NULL;
schedule_event(dch, FLG_RECVQUEUE);
}
EXPORT_SYMBOL(recv_Echannel);
void
recv_Bchannel(struct bchannel *bch, unsigned int id, bool force)
{
struct mISDNhead *hh;
/* if allocation did fail upper functions still may call us */
if (unlikely(!bch->rx_skb))
return;
if (unlikely(!bch->rx_skb->len)) {
/* we have no data to send - this may happen after recovery
* from overflow or too small allocation.
* We need to free the buffer here */
dev_kfree_skb(bch->rx_skb);
bch->rx_skb = NULL;
} else {
if (test_bit(FLG_TRANSPARENT, &bch->Flags) &&
(bch->rx_skb->len < bch->minlen) && !force)
return;
hh = mISDN_HEAD_P(bch->rx_skb);
hh->prim = PH_DATA_IND;
hh->id = id;
if (bch->rcount >= 64) {
printk(KERN_WARNING
"B%d receive queue overflow - flushing!\n",
bch->nr);
skb_queue_purge(&bch->rqueue);
}
bch->rcount++;
skb_queue_tail(&bch->rqueue, bch->rx_skb);
bch->rx_skb = NULL;
schedule_event(bch, FLG_RECVQUEUE);
}
}
EXPORT_SYMBOL(recv_Bchannel);
void
recv_Dchannel_skb(struct dchannel *dch, struct sk_buff *skb)
{
skb_queue_tail(&dch->rqueue, skb);
schedule_event(dch, FLG_RECVQUEUE);
}
EXPORT_SYMBOL(recv_Dchannel_skb);
void
recv_Bchannel_skb(struct bchannel *bch, struct sk_buff *skb)
{
if (bch->rcount >= 64) {
printk(KERN_WARNING "B-channel %p receive queue overflow, "
"flushing!\n", bch);
skb_queue_purge(&bch->rqueue);
bch->rcount = 0;
}
bch->rcount++;
skb_queue_tail(&bch->rqueue, skb);
schedule_event(bch, FLG_RECVQUEUE);
}
EXPORT_SYMBOL(recv_Bchannel_skb);
static void
confirm_Dsend(struct dchannel *dch)
{
struct sk_buff *skb;
skb = _alloc_mISDN_skb(PH_DATA_CNF, mISDN_HEAD_ID(dch->tx_skb),
0, NULL, GFP_ATOMIC);
if (!skb) {
printk(KERN_ERR "%s: no skb id %x\n", __func__,
mISDN_HEAD_ID(dch->tx_skb));
return;
}
skb_queue_tail(&dch->rqueue, skb);
schedule_event(dch, FLG_RECVQUEUE);
}
int
get_next_dframe(struct dchannel *dch)
{
dch->tx_idx = 0;
dch->tx_skb = skb_dequeue(&dch->squeue);
if (dch->tx_skb) {
confirm_Dsend(dch);
return 1;
}
dch->tx_skb = NULL;
test_and_clear_bit(FLG_TX_BUSY, &dch->Flags);
return 0;
}
EXPORT_SYMBOL(get_next_dframe);
static void
confirm_Bsend(struct bchannel *bch)
{
struct sk_buff *skb;
if (bch->rcount >= 64) {
printk(KERN_WARNING "B-channel %p receive queue overflow, "
"flushing!\n", bch);
skb_queue_purge(&bch->rqueue);
bch->rcount = 0;
}
skb = _alloc_mISDN_skb(PH_DATA_CNF, mISDN_HEAD_ID(bch->tx_skb),
0, NULL, GFP_ATOMIC);
if (!skb) {
printk(KERN_ERR "%s: no skb id %x\n", __func__,
mISDN_HEAD_ID(bch->tx_skb));
return;
}
bch->rcount++;
skb_queue_tail(&bch->rqueue, skb);
schedule_event(bch, FLG_RECVQUEUE);
}
int
get_next_bframe(struct bchannel *bch)
{
bch->tx_idx = 0;
if (test_bit(FLG_TX_NEXT, &bch->Flags)) {
bch->tx_skb = bch->next_skb;
if (bch->tx_skb) {
bch->next_skb = NULL;
test_and_clear_bit(FLG_TX_NEXT, &bch->Flags);
/* confirm imediately to allow next data */
confirm_Bsend(bch);
return 1;
} else {
test_and_clear_bit(FLG_TX_NEXT, &bch->Flags);
printk(KERN_WARNING "B TX_NEXT without skb\n");
}
}
bch->tx_skb = NULL;
test_and_clear_bit(FLG_TX_BUSY, &bch->Flags);
return 0;
}
EXPORT_SYMBOL(get_next_bframe);
void
queue_ch_frame(struct mISDNchannel *ch, u_int pr, int id, struct sk_buff *skb)
{
struct mISDNhead *hh;
if (!skb) {
_queue_data(ch, pr, id, 0, NULL, GFP_ATOMIC);
} else {
if (ch->peer) {
hh = mISDN_HEAD_P(skb);
hh->prim = pr;
hh->id = id;
if (!ch->recv(ch->peer, skb))
return;
}
dev_kfree_skb(skb);
}
}
EXPORT_SYMBOL(queue_ch_frame);
int
dchannel_senddata(struct dchannel *ch, struct sk_buff *skb)
{
/* check oversize */
if (skb->len <= 0) {
printk(KERN_WARNING "%s: skb too small\n", __func__);
return -EINVAL;
}
if (skb->len > ch->maxlen) {
printk(KERN_WARNING "%s: skb too large(%d/%d)\n",
__func__, skb->len, ch->maxlen);
return -EINVAL;
}
/* HW lock must be obtained */
if (test_and_set_bit(FLG_TX_BUSY, &ch->Flags)) {
skb_queue_tail(&ch->squeue, skb);
return 0;
} else {
/* write to fifo */
ch->tx_skb = skb;
ch->tx_idx = 0;
return 1;
}
}
EXPORT_SYMBOL(dchannel_senddata);
int
bchannel_senddata(struct bchannel *ch, struct sk_buff *skb)
{
/* check oversize */
if (skb->len <= 0) {
printk(KERN_WARNING "%s: skb too small\n", __func__);
return -EINVAL;
}
if (skb->len > ch->maxlen) {
printk(KERN_WARNING "%s: skb too large(%d/%d)\n",
__func__, skb->len, ch->maxlen);
return -EINVAL;
}
/* HW lock must be obtained */
/* check for pending next_skb */
if (ch->next_skb) {
printk(KERN_WARNING
"%s: next_skb exist ERROR (skb->len=%d next_skb->len=%d)\n",
__func__, skb->len, ch->next_skb->len);
return -EBUSY;
}
if (test_and_set_bit(FLG_TX_BUSY, &ch->Flags)) {
test_and_set_bit(FLG_TX_NEXT, &ch->Flags);
ch->next_skb = skb;
return 0;
} else {
/* write to fifo */
ch->tx_skb = skb;
ch->tx_idx = 0;
confirm_Bsend(ch);
return 1;
}
}
EXPORT_SYMBOL(bchannel_senddata);
/* The function allocates a new receive skb on demand with a size for the
* requirements of the current protocol. It returns the tailroom of the
* receive skb or an error.
*/
int
bchannel_get_rxbuf(struct bchannel *bch, int reqlen)
{
int len;
if (bch->rx_skb) {
len = skb_tailroom(bch->rx_skb);
if (len < reqlen) {
pr_warning("B%d no space for %d (only %d) bytes\n",
bch->nr, reqlen, len);
if (test_bit(FLG_TRANSPARENT, &bch->Flags)) {
/* send what we have now and try a new buffer */
recv_Bchannel(bch, 0, true);
} else {
/* on HDLC we have to drop too big frames */
return -EMSGSIZE;
}
} else {
return len;
}
}
/* update current min/max length first */
if (unlikely(bch->maxlen != bch->next_maxlen))
bch->maxlen = bch->next_maxlen;
if (unlikely(bch->minlen != bch->next_minlen))
bch->minlen = bch->next_minlen;
if (unlikely(reqlen > bch->maxlen))
return -EMSGSIZE;
if (test_bit(FLG_TRANSPARENT, &bch->Flags)) {
if (reqlen >= bch->minlen) {
len = reqlen;
} else {
len = 2 * bch->minlen;
if (len > bch->maxlen)
len = bch->maxlen;
}
} else {
/* with HDLC we do not know the length yet */
len = bch->maxlen;
}
bch->rx_skb = mI_alloc_skb(len, GFP_ATOMIC);
if (!bch->rx_skb) {
pr_warning("B%d receive no memory for %d bytes\n",
bch->nr, len);
len = -ENOMEM;
}
return len;
}
EXPORT_SYMBOL(bchannel_get_rxbuf);
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...