From: George Kennedy <george.kennedy@oracle.com>
To: pbonzini@redhat.com, famz@redhat.com, qemu-devel@nongnu.org,
boris.ostrovsky@oracle.com
Subject: [Qemu-devel] [PATCH v5] lsi: Reselection needed to remove pending commands from queue
Date: Wed, 7 Nov 2018 11:19:45 -0500 [thread overview]
Message-ID: <1541607585-15570-1-git-send-email-george.kennedy@oracle.com> (raw)
Under heavy IO (e.g. fio) the queue is not checked frequently enough for
pending commands. As a result some pending commands are timed out by the
linux sym53c8xx driver, which sends SCSI Abort messages for the timed out
commands. The SCSI Abort messages result in linux errors, which show up
on the console and in /var/log/messages.
e.g.
sd 0:0:3:0: [sdd] tag#33 ABORT operation started
scsi target0:0:3: control msgout:
80 20 47 d
sd 0:0:3:0: ABORT operation complete.
scsi target0:0:4: message d sent on bad reselection
When the current command completes, check if there is a pending command
on the queue and if a pending command exists, set a flag indicating
that a WAIT RESELECT command is needed to handle a queued pending command.
When a WAIT RESELECT is needed, intercept and save the current DMA Scripts
Ptr (DSP) contents following a WAIT DISCONNECT Script instruction and load
the DSP instead with the pointer to the Reselection Scripts. When
Reselection has completed, restore the original DSP contents.
Signed-off-by: George Kennedy <george.kennedy@oracle.com>
---
Thanks Paolo,
Sorry for missing your suggested fix. I believe the Script intercept is now
being done as you had suggested.
hw/scsi/lsi53c895a.c | 52 +++++++++++++++++++++++++++++++++++++++++-----------
1 file changed, 41 insertions(+), 11 deletions(-)
diff --git a/hw/scsi/lsi53c895a.c b/hw/scsi/lsi53c895a.c
index d1e6534..7fb8ddc 100644
--- a/hw/scsi/lsi53c895a.c
+++ b/hw/scsi/lsi53c895a.c
@@ -219,6 +219,8 @@ typedef struct {
int command_complete;
QTAILQ_HEAD(, lsi_request) queue;
lsi_request *current;
+ bool want_resel; /* need resel to handle queued completed cmds */
+ uint32_t resel_dsp; /* DMA Scripts Ptr (DSP) of reselection scsi scripts */
uint32_t dsa;
uint32_t temp;
@@ -298,6 +300,18 @@ static inline int lsi_irq_on_rsl(LSIState *s)
return (s->sien0 & LSI_SIST0_RSL) && (s->scid & LSI_SCID_RRE);
}
+static lsi_request *get_pending_req(LSIState *s)
+{
+ lsi_request *p;
+
+ QTAILQ_FOREACH(p, &s->queue, next) {
+ if (p->pending) {
+ return p;
+ }
+ }
+ return NULL;
+}
+
static void lsi_soft_reset(LSIState *s)
{
trace_lsi_reset();
@@ -446,7 +460,6 @@ static void lsi_update_irq(LSIState *s)
{
int level;
static int last_level;
- lsi_request *p;
/* It's unclear whether the DIP/SIP bits should be cleared when the
Interrupt Status Registers are cleared or when istat0 is read.
@@ -477,12 +490,12 @@ static void lsi_update_irq(LSIState *s)
lsi_set_irq(s, level);
if (!level && lsi_irq_on_rsl(s) && !(s->scntl1 & LSI_SCNTL1_CON)) {
+ lsi_request *p;
+
trace_lsi_update_irq_disconnected();
- QTAILQ_FOREACH(p, &s->queue, next) {
- if (p->pending) {
- lsi_reselect(s, p);
- break;
- }
+ p = get_pending_req(s);
+ if (p) {
+ lsi_reselect(s, p);
}
}
}
@@ -759,6 +772,8 @@ static void lsi_command_complete(SCSIRequest *req, uint32_t status, size_t resid
lsi_request_free(s, s->current);
scsi_req_unref(req);
}
+ s->want_resel = get_pending_req(s) ? true : false;
+
lsi_resume_script(s);
}
@@ -1064,11 +1079,13 @@ static void lsi_wait_reselect(LSIState *s)
trace_lsi_wait_reselect();
- QTAILQ_FOREACH(p, &s->queue, next) {
- if (p->pending) {
- lsi_reselect(s, p);
- break;
- }
+ if (s->current) {
+ return;
+ }
+
+ p = get_pending_req(s);
+ if (p) {
+ lsi_reselect(s, p);
}
if (s->current == NULL) {
s->waiting = 1;
@@ -1082,6 +1099,7 @@ static void lsi_execute_script(LSIState *s)
uint32_t addr, addr_high;
int opcode;
int insn_processed = 0;
+ uint32_t save_dsp = 0;
s->istat1 |= LSI_ISTAT1_SRUN;
again:
@@ -1258,9 +1276,21 @@ again:
case 1: /* Disconnect */
trace_lsi_execute_script_io_disconnect();
s->scntl1 &= ~LSI_SCNTL1_CON;
+ if (s->want_resel && s->resel_dsp) {
+ save_dsp = s->dsp;
+ s->dsp = s->resel_dsp;
+ s->want_resel = false;
+ }
break;
case 2: /* Wait Reselect */
+ if (!s->resel_dsp) {
+ s->resel_dsp = s->dsp - 8;
+ }
if (!lsi_irq_on_rsl(s)) {
+ if (save_dsp) {
+ s->dsp = save_dsp;
+ save_dsp = 0;
+ }
lsi_wait_reselect(s);
}
break;
--
1.8.3.1
next reply other threads:[~2018-11-07 16:27 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-11-07 16:19 George Kennedy [this message]
2018-11-07 17:37 ` [Qemu-devel] [PATCH v5] lsi: Reselection needed to remove pending commands from queue Paolo Bonzini
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1541607585-15570-1-git-send-email-george.kennedy@oracle.com \
--to=george.kennedy@oracle.com \
--cc=boris.ostrovsky@oracle.com \
--cc=famz@redhat.com \
--cc=pbonzini@redhat.com \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).