xemu/hw/scsi/esp.c

1146 lines
29 KiB
C
Raw Normal View History

/*
* QEMU ESP/NCR53C9x emulation
*
* Copyright (c) 2005-2006 Fabrice Bellard
* Copyright (c) 2012 Herve Poussineau
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include "qemu/osdep.h"
#include "hw/sysbus.h"
#include "migration/vmstate.h"
#include "hw/irq.h"
#include "hw/scsi/esp.h"
#include "trace.h"
#include "qemu/log.h"
#include "qemu/module.h"
/*
* On Sparc32, this is the ESP (NCR53C90) part of chip STP2000 (Master I/O),
* also produced as NCR89C100. See
* http://www.ibiblio.org/pub/historic-linux/early-ports/Sparc/NCR/NCR89C100.txt
* and
* http://www.ibiblio.org/pub/historic-linux/early-ports/Sparc/NCR/NCR53C9X.txt
*
* On Macintosh Quadra it is a NCR53C96.
*/
static void esp_raise_irq(ESPState *s)
{
if (!(s->rregs[ESP_RSTAT] & STAT_INT)) {
s->rregs[ESP_RSTAT] |= STAT_INT;
qemu_irq_raise(s->irq);
trace_esp_raise_irq();
}
}
static void esp_lower_irq(ESPState *s)
{
if (s->rregs[ESP_RSTAT] & STAT_INT) {
s->rregs[ESP_RSTAT] &= ~STAT_INT;
qemu_irq_lower(s->irq);
trace_esp_lower_irq();
}
}
static void esp_raise_drq(ESPState *s)
{
qemu_irq_raise(s->irq_data);
trace_esp_raise_drq();
}
static void esp_lower_drq(ESPState *s)
{
qemu_irq_lower(s->irq_data);
trace_esp_lower_drq();
}
void esp_dma_enable(ESPState *s, int irq, int level)
{
if (level) {
s->dma_enabled = 1;
trace_esp_dma_enable();
if (s->dma_cb) {
s->dma_cb(s);
s->dma_cb = NULL;
}
} else {
trace_esp_dma_disable();
s->dma_enabled = 0;
}
}
void esp_request_cancelled(SCSIRequest *req)
{
ESPState *s = req->hba_private;
if (req == s->current_req) {
scsi_req_unref(s->current_req);
s->current_req = NULL;
s->current_dev = NULL;
}
}
static uint32_t esp_get_tc(ESPState *s)
{
uint32_t dmalen;
dmalen = s->rregs[ESP_TCLO];
dmalen |= s->rregs[ESP_TCMID] << 8;
dmalen |= s->rregs[ESP_TCHI] << 16;
return dmalen;
}
static void esp_set_tc(ESPState *s, uint32_t dmalen)
{
s->rregs[ESP_TCLO] = dmalen;
s->rregs[ESP_TCMID] = dmalen >> 8;
s->rregs[ESP_TCHI] = dmalen >> 16;
}
static uint32_t esp_get_stc(ESPState *s)
{
uint32_t dmalen;
dmalen = s->wregs[ESP_TCLO];
dmalen |= s->wregs[ESP_TCMID] << 8;
dmalen |= s->wregs[ESP_TCHI] << 16;
return dmalen;
}
static void set_pdma(ESPState *s, enum pdma_origin_id origin,
uint32_t index, uint32_t len)
{
s->pdma_origin = origin;
s->pdma_start = index;
s->pdma_cur = index;
s->pdma_len = len;
}
static uint8_t *get_pdma_buf(ESPState *s)
{
switch (s->pdma_origin) {
case TI:
return s->ti_buf;
case CMD:
return s->cmdbuf;
case ASYNC:
return s->async_buf;
}
return NULL;
}
static uint8_t esp_pdma_read(ESPState *s)
{
uint32_t dmalen = esp_get_tc(s);
uint8_t val;
if (dmalen == 0 || s->pdma_len == 0) {
return 0;
}
switch (s->pdma_origin) {
case TI:
val = s->ti_buf[s->pdma_cur++];
break;
case CMD:
val = s->cmdbuf[s->cmdlen++];
s->pdma_cur++;
break;
case ASYNC:
val = s->async_buf[s->pdma_cur++];
break;
default:
g_assert_not_reached();
}
s->pdma_len--;
dmalen--;
esp_set_tc(s, dmalen);
return val;
}
static void esp_pdma_write(ESPState *s, uint8_t val)
{
uint32_t dmalen = esp_get_tc(s);
if (dmalen == 0 || s->pdma_len == 0) {
return;
}
switch (s->pdma_origin) {
case TI:
s->ti_buf[s->pdma_cur++] = val;
break;
case CMD:
s->cmdbuf[s->cmdlen++] = val;
s->pdma_cur++;
break;
case ASYNC:
s->async_buf[s->pdma_cur++] = val;
break;
default:
g_assert_not_reached();
}
s->pdma_len--;
dmalen--;
esp_set_tc(s, dmalen);
}
static int get_cmd_cb(ESPState *s)
{
int target;
target = s->wregs[ESP_WBUSID] & BUSID_DID;
s->ti_size = 0;
s->ti_rptr = 0;
s->ti_wptr = 0;
if (s->current_req) {
/* Started a new command before the old one finished. Cancel it. */
scsi_req_cancel(s->current_req);
s->async_len = 0;
}
s->current_dev = scsi_device_find(&s->bus, 0, target, 0);
if (!s->current_dev) {
/* No such drive */
s->rregs[ESP_RSTAT] = 0;
s->rregs[ESP_RINTR] = INTR_DC;
s->rregs[ESP_RSEQ] = SEQ_0;
esp_raise_irq(s);
return -1;
}
return 0;
}
static uint32_t get_cmd(ESPState *s, uint8_t *buf, uint8_t buflen)
{
uint32_t dmalen;
int target;
target = s->wregs[ESP_WBUSID] & BUSID_DID;
if (s->dma) {
dmalen = esp_get_tc(s);
if (dmalen > buflen) {
return 0;
}
if (s->dma_memory_read) {
s->dma_memory_read(s->dma_opaque, buf, dmalen);
} else {
set_pdma(s, CMD, 0, dmalen);
esp_raise_drq(s);
return 0;
}
} else {
dmalen = s->ti_size;
if (dmalen > TI_BUFSZ) {
return 0;
}
memcpy(buf, s->ti_buf, dmalen);
buf[0] = buf[2] >> 5;
}
trace_esp_get_cmd(dmalen, target);
if (get_cmd_cb(s) < 0) {
return 0;
}
return dmalen;
}
static void do_busid_cmd(ESPState *s, uint8_t *buf, uint8_t busid)
{
int32_t datalen;
int lun;
SCSIDevice *current_lun;
trace_esp_do_busid_cmd(busid);
lun = busid & 7;
current_lun = scsi_device_find(&s->bus, 0, s->current_dev->id, lun);
s->current_req = scsi_req_new(current_lun, 0, lun, buf, s);
datalen = scsi_req_enqueue(s->current_req);
s->ti_size = datalen;
if (datalen != 0) {
s->rregs[ESP_RSTAT] = STAT_TC;
esp_set_tc(s, 0);
if (datalen > 0) {
s->rregs[ESP_RSTAT] |= STAT_DI;
} else {
s->rregs[ESP_RSTAT] |= STAT_DO;
}
scsi_req_continue(s->current_req);
}
s->rregs[ESP_RINTR] = INTR_BS | INTR_FC;
s->rregs[ESP_RSEQ] = SEQ_CD;
esp_raise_irq(s);
}
static void do_cmd(ESPState *s)
{
uint8_t *buf = s->cmdbuf;
uint8_t busid = buf[0];
do_busid_cmd(s, &buf[1], busid);
}
static void satn_pdma_cb(ESPState *s)
{
if (get_cmd_cb(s) < 0) {
return;
}
s->do_cmd = 0;
if (s->cmdlen) {
do_cmd(s);
}
}
static void handle_satn(ESPState *s)
{
if (s->dma && !s->dma_enabled) {
s->dma_cb = handle_satn;
return;
}
s->pdma_cb = satn_pdma_cb;
s->cmdlen = get_cmd(s, s->cmdbuf, sizeof(s->cmdbuf));
if (s->cmdlen) {
do_cmd(s);
} else {
s->do_cmd = 1;
}
}
static void s_without_satn_pdma_cb(ESPState *s)
{
if (get_cmd_cb(s) < 0) {
return;
}
s->do_cmd = 0;
if (s->cmdlen) {
do_busid_cmd(s, get_pdma_buf(s) + s->pdma_start, 0);
}
}
static void handle_s_without_atn(ESPState *s)
{
if (s->dma && !s->dma_enabled) {
s->dma_cb = handle_s_without_atn;
return;
}
s->pdma_cb = s_without_satn_pdma_cb;
s->cmdlen = get_cmd(s, s->cmdbuf, sizeof(s->cmdbuf));
if (s->cmdlen) {
do_busid_cmd(s, s->cmdbuf, 0);
} else {
s->do_cmd = 1;
}
}
static void satn_stop_pdma_cb(ESPState *s)
{
if (get_cmd_cb(s) < 0) {
return;
}
s->do_cmd = 0;
if (s->cmdlen) {
trace_esp_handle_satn_stop(s->cmdlen);
s->do_cmd = 1;
s->rregs[ESP_RSTAT] = STAT_TC | STAT_CD;
s->rregs[ESP_RINTR] = INTR_BS | INTR_FC;
s->rregs[ESP_RSEQ] = SEQ_CD;
esp_raise_irq(s);
}
}
static void handle_satn_stop(ESPState *s)
{
if (s->dma && !s->dma_enabled) {
s->dma_cb = handle_satn_stop;
return;
}
s->pdma_cb = satn_stop_pdma_cb;
s->cmdlen = get_cmd(s, s->cmdbuf, sizeof(s->cmdbuf));
if (s->cmdlen) {
trace_esp_handle_satn_stop(s->cmdlen);
s->do_cmd = 1;
s->rregs[ESP_RSTAT] = STAT_TC | STAT_CD;
s->rregs[ESP_RINTR] = INTR_BS | INTR_FC;
s->rregs[ESP_RSEQ] = SEQ_CD;
esp_raise_irq(s);
} else {
s->do_cmd = 1;
}
}
static void write_response_pdma_cb(ESPState *s)
{
s->rregs[ESP_RSTAT] = STAT_TC | STAT_ST;
s->rregs[ESP_RINTR] = INTR_BS | INTR_FC;
s->rregs[ESP_RSEQ] = SEQ_CD;
esp_raise_irq(s);
}
static void write_response(ESPState *s)
{
trace_esp_write_response(s->status);
s->ti_buf[0] = s->status;
s->ti_buf[1] = 0;
if (s->dma) {
if (s->dma_memory_write) {
s->dma_memory_write(s->dma_opaque, s->ti_buf, 2);
s->rregs[ESP_RSTAT] = STAT_TC | STAT_ST;
s->rregs[ESP_RINTR] = INTR_BS | INTR_FC;
s->rregs[ESP_RSEQ] = SEQ_CD;
} else {
set_pdma(s, TI, 0, 2);
s->pdma_cb = write_response_pdma_cb;
esp_raise_drq(s);
return;
}
} else {
s->ti_size = 2;
s->ti_rptr = 0;
s->ti_wptr = 2;
s->rregs[ESP_RFLAGS] = 2;
}
esp_raise_irq(s);
}
static void esp_dma_done(ESPState *s)
{
s->rregs[ESP_RSTAT] |= STAT_TC;
s->rregs[ESP_RINTR] = INTR_BS;
s->rregs[ESP_RSEQ] = 0;
s->rregs[ESP_RFLAGS] = 0;
esp_set_tc(s, 0);
esp_raise_irq(s);
}
static void do_dma_pdma_cb(ESPState *s)
{
int to_device = ((s->rregs[ESP_RSTAT] & 7) == STAT_DO);
int len = s->pdma_cur - s->pdma_start;
if (s->do_cmd) {
s->ti_size = 0;
s->cmdlen = 0;
s->do_cmd = 0;
do_cmd(s);
return;
}
s->async_buf += len;
s->async_len -= len;
if (to_device) {
s->ti_size += len;
} else {
s->ti_size -= len;
}
if (s->async_len == 0) {
scsi_req_continue(s->current_req);
/*
* If there is still data to be read from the device then
* complete the DMA operation immediately. Otherwise defer
* until the scsi layer has completed.
*/
if (to_device || esp_get_tc(s) != 0 || s->ti_size == 0) {
return;
}
}
/* Partially filled a scsi buffer. Complete immediately. */
esp_dma_done(s);
}
static void esp_do_dma(ESPState *s)
{
uint32_t len;
int to_device = ((s->rregs[ESP_RSTAT] & 7) == STAT_DO);
len = esp_get_tc(s);
if (s->do_cmd) {
/*
* handle_ti_cmd() case: esp_do_dma() is called only from
* handle_ti_cmd() with do_cmd != NULL (see the assert())
*/
trace_esp_do_dma(s->cmdlen, len);
assert(s->cmdlen <= sizeof(s->cmdbuf) &&
len <= sizeof(s->cmdbuf) - s->cmdlen);
if (s->dma_memory_read) {
s->dma_memory_read(s->dma_opaque, &s->cmdbuf[s->cmdlen], len);
} else {
set_pdma(s, CMD, s->cmdlen, len);
s->pdma_cb = do_dma_pdma_cb;
esp_raise_drq(s);
return;
}
trace_esp_handle_ti_cmd(s->cmdlen);
s->ti_size = 0;
s->cmdlen = 0;
s->do_cmd = 0;
do_cmd(s);
return;
}
if (s->async_len == 0) {
/* Defer until data is available. */
return;
}
if (len > s->async_len) {
len = s->async_len;
}
if (to_device) {
if (s->dma_memory_read) {
s->dma_memory_read(s->dma_opaque, s->async_buf, len);
} else {
set_pdma(s, ASYNC, 0, len);
s->pdma_cb = do_dma_pdma_cb;
esp_raise_drq(s);
return;
}
} else {
if (s->dma_memory_write) {
s->dma_memory_write(s->dma_opaque, s->async_buf, len);
} else {
set_pdma(s, ASYNC, 0, len);
s->pdma_cb = do_dma_pdma_cb;
esp_raise_drq(s);
return;
}
}
esp_set_tc(s, esp_get_tc(s) - len);
s->async_buf += len;
s->async_len -= len;
if (to_device) {
s->ti_size += len;
} else {
s->ti_size -= len;
}
if (s->async_len == 0) {
scsi_req_continue(s->current_req);
/*
* If there is still data to be read from the device then
* complete the DMA operation immediately. Otherwise defer
* until the scsi layer has completed.
*/
if (to_device || esp_get_tc(s) != 0 || s->ti_size == 0) {
return;
}
}
/* Partially filled a scsi buffer. Complete immediately. */
esp_dma_done(s);
}
scsi: esp: Defer command completion until previous interrupts have been handled The guest OS reads RSTAT, RSEQ, and RINTR, and expects those registers to reflect a consistent state. However, it is possible that the registers can change after RSTAT was read, but before RINTR is read, when esp_command_complete() is called. Guest OS qemu -------- ---- [handle interrupt] Read RSTAT esp_command_complete() RSTAT = STAT_ST esp_dma_done() RSTAT |= STAT_TC RSEQ = 0 RINTR = INTR_BS Read RSEQ Read RINTR RINTR = 0 RSTAT &= ~STAT_TC RSEQ = SEQ_CD The guest OS would then try to handle INTR_BS combined with an old value of RSTAT. This sometimes resulted in lost events, spurious interrupts, guest OS confusion, and stalled SCSI operations. A typical guest error log (observed with various versions of Linux) looks as follows. scsi host1: Spurious irq, sreg=13. ... scsi host1: Aborting command [84531f10:2a] scsi host1: Current command [f882eea8:35] scsi host1: Queued command [84531f10:2a] scsi host1: Active command [f882eea8:35] scsi host1: Dumping command log scsi host1: ent[15] CMD val[44] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[00] event[0c] scsi host1: ent[16] CMD val[01] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[17] CMD val[43] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[18] EVENT val[0d] sreg[92] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c] ... Defer handling command completion until previous interrupts have been handled to fix the problem. Signed-off-by: Guenter Roeck <linux@roeck-us.net>
2018-11-29 17:17:42 +00:00
static void esp_report_command_complete(ESPState *s, uint32_t status)
{
trace_esp_command_complete();
if (s->ti_size != 0) {
trace_esp_command_complete_unexpected();
}
s->ti_size = 0;
s->async_len = 0;
if (status) {
trace_esp_command_complete_fail();
}
s->status = status;
s->rregs[ESP_RSTAT] = STAT_ST;
esp_dma_done(s);
if (s->current_req) {
scsi_req_unref(s->current_req);
s->current_req = NULL;
s->current_dev = NULL;
}
}
void esp_command_complete(SCSIRequest *req, size_t resid)
scsi: esp: Defer command completion until previous interrupts have been handled The guest OS reads RSTAT, RSEQ, and RINTR, and expects those registers to reflect a consistent state. However, it is possible that the registers can change after RSTAT was read, but before RINTR is read, when esp_command_complete() is called. Guest OS qemu -------- ---- [handle interrupt] Read RSTAT esp_command_complete() RSTAT = STAT_ST esp_dma_done() RSTAT |= STAT_TC RSEQ = 0 RINTR = INTR_BS Read RSEQ Read RINTR RINTR = 0 RSTAT &= ~STAT_TC RSEQ = SEQ_CD The guest OS would then try to handle INTR_BS combined with an old value of RSTAT. This sometimes resulted in lost events, spurious interrupts, guest OS confusion, and stalled SCSI operations. A typical guest error log (observed with various versions of Linux) looks as follows. scsi host1: Spurious irq, sreg=13. ... scsi host1: Aborting command [84531f10:2a] scsi host1: Current command [f882eea8:35] scsi host1: Queued command [84531f10:2a] scsi host1: Active command [f882eea8:35] scsi host1: Dumping command log scsi host1: ent[15] CMD val[44] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[00] event[0c] scsi host1: ent[16] CMD val[01] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[17] CMD val[43] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[18] EVENT val[0d] sreg[92] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c] ... Defer handling command completion until previous interrupts have been handled to fix the problem. Signed-off-by: Guenter Roeck <linux@roeck-us.net>
2018-11-29 17:17:42 +00:00
{
ESPState *s = req->hba_private;
if (s->rregs[ESP_RSTAT] & STAT_INT) {
/*
* Defer handling command complete until the previous
scsi: esp: Defer command completion until previous interrupts have been handled The guest OS reads RSTAT, RSEQ, and RINTR, and expects those registers to reflect a consistent state. However, it is possible that the registers can change after RSTAT was read, but before RINTR is read, when esp_command_complete() is called. Guest OS qemu -------- ---- [handle interrupt] Read RSTAT esp_command_complete() RSTAT = STAT_ST esp_dma_done() RSTAT |= STAT_TC RSEQ = 0 RINTR = INTR_BS Read RSEQ Read RINTR RINTR = 0 RSTAT &= ~STAT_TC RSEQ = SEQ_CD The guest OS would then try to handle INTR_BS combined with an old value of RSTAT. This sometimes resulted in lost events, spurious interrupts, guest OS confusion, and stalled SCSI operations. A typical guest error log (observed with various versions of Linux) looks as follows. scsi host1: Spurious irq, sreg=13. ... scsi host1: Aborting command [84531f10:2a] scsi host1: Current command [f882eea8:35] scsi host1: Queued command [84531f10:2a] scsi host1: Active command [f882eea8:35] scsi host1: Dumping command log scsi host1: ent[15] CMD val[44] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[00] event[0c] scsi host1: ent[16] CMD val[01] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[17] CMD val[43] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[18] EVENT val[0d] sreg[92] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c] ... Defer handling command completion until previous interrupts have been handled to fix the problem. Signed-off-by: Guenter Roeck <linux@roeck-us.net>
2018-11-29 17:17:42 +00:00
* interrupt has been handled.
*/
trace_esp_command_complete_deferred();
s->deferred_status = req->status;
scsi: esp: Defer command completion until previous interrupts have been handled The guest OS reads RSTAT, RSEQ, and RINTR, and expects those registers to reflect a consistent state. However, it is possible that the registers can change after RSTAT was read, but before RINTR is read, when esp_command_complete() is called. Guest OS qemu -------- ---- [handle interrupt] Read RSTAT esp_command_complete() RSTAT = STAT_ST esp_dma_done() RSTAT |= STAT_TC RSEQ = 0 RINTR = INTR_BS Read RSEQ Read RINTR RINTR = 0 RSTAT &= ~STAT_TC RSEQ = SEQ_CD The guest OS would then try to handle INTR_BS combined with an old value of RSTAT. This sometimes resulted in lost events, spurious interrupts, guest OS confusion, and stalled SCSI operations. A typical guest error log (observed with various versions of Linux) looks as follows. scsi host1: Spurious irq, sreg=13. ... scsi host1: Aborting command [84531f10:2a] scsi host1: Current command [f882eea8:35] scsi host1: Queued command [84531f10:2a] scsi host1: Active command [f882eea8:35] scsi host1: Dumping command log scsi host1: ent[15] CMD val[44] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[00] event[0c] scsi host1: ent[16] CMD val[01] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[17] CMD val[43] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[18] EVENT val[0d] sreg[92] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c] ... Defer handling command completion until previous interrupts have been handled to fix the problem. Signed-off-by: Guenter Roeck <linux@roeck-us.net>
2018-11-29 17:17:42 +00:00
s->deferred_complete = true;
return;
}
esp_report_command_complete(s, req->status);
scsi: esp: Defer command completion until previous interrupts have been handled The guest OS reads RSTAT, RSEQ, and RINTR, and expects those registers to reflect a consistent state. However, it is possible that the registers can change after RSTAT was read, but before RINTR is read, when esp_command_complete() is called. Guest OS qemu -------- ---- [handle interrupt] Read RSTAT esp_command_complete() RSTAT = STAT_ST esp_dma_done() RSTAT |= STAT_TC RSEQ = 0 RINTR = INTR_BS Read RSEQ Read RINTR RINTR = 0 RSTAT &= ~STAT_TC RSEQ = SEQ_CD The guest OS would then try to handle INTR_BS combined with an old value of RSTAT. This sometimes resulted in lost events, spurious interrupts, guest OS confusion, and stalled SCSI operations. A typical guest error log (observed with various versions of Linux) looks as follows. scsi host1: Spurious irq, sreg=13. ... scsi host1: Aborting command [84531f10:2a] scsi host1: Current command [f882eea8:35] scsi host1: Queued command [84531f10:2a] scsi host1: Active command [f882eea8:35] scsi host1: Dumping command log scsi host1: ent[15] CMD val[44] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[00] event[0c] scsi host1: ent[16] CMD val[01] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[17] CMD val[43] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[18] EVENT val[0d] sreg[92] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c] ... Defer handling command completion until previous interrupts have been handled to fix the problem. Signed-off-by: Guenter Roeck <linux@roeck-us.net>
2018-11-29 17:17:42 +00:00
}
void esp_transfer_data(SCSIRequest *req, uint32_t len)
{
ESPState *s = req->hba_private;
uint32_t dmalen = esp_get_tc(s);
assert(!s->do_cmd);
trace_esp_transfer_data(dmalen, s->ti_size);
s->async_len = len;
s->async_buf = scsi_req_get_buf(req);
if (dmalen) {
esp_do_dma(s);
} else if (s->ti_size <= 0) {
/*
* If this was the last part of a DMA transfer then the
* completion interrupt is deferred to here.
*/
esp_dma_done(s);
}
}
static void handle_ti(ESPState *s)
{
uint32_t dmalen;
if (s->dma && !s->dma_enabled) {
s->dma_cb = handle_ti;
return;
}
dmalen = esp_get_tc(s);
if (s->dma) {
trace_esp_handle_ti(dmalen);
s->rregs[ESP_RSTAT] &= ~STAT_TC;
esp_do_dma(s);
} else if (s->do_cmd) {
trace_esp_handle_ti_cmd(s->cmdlen);
s->ti_size = 0;
s->cmdlen = 0;
s->do_cmd = 0;
do_cmd(s);
}
}
void esp_hard_reset(ESPState *s)
{
memset(s->rregs, 0, ESP_REGS);
memset(s->wregs, 0, ESP_REGS);
s->tchi_written = 0;
s->ti_size = 0;
s->ti_rptr = 0;
s->ti_wptr = 0;
s->dma = 0;
s->do_cmd = 0;
s->dma_cb = NULL;
s->rregs[ESP_CFG1] = 7;
}
static void esp_soft_reset(ESPState *s)
{
qemu_irq_lower(s->irq);
qemu_irq_lower(s->irq_data);
esp_hard_reset(s);
}
static void parent_esp_reset(ESPState *s, int irq, int level)
{
if (level) {
esp_soft_reset(s);
}
}
uint64_t esp_reg_read(ESPState *s, uint32_t saddr)
{
uint32_t val;
switch (saddr) {
case ESP_FIFO:
if ((s->rregs[ESP_RSTAT] & STAT_PIO_MASK) == 0) {
/* Data out. */
qemu_log_mask(LOG_UNIMP, "esp: PIO data read not implemented\n");
s->rregs[ESP_FIFO] = 0;
} else if (s->ti_rptr < s->ti_wptr) {
s->ti_size--;
s->rregs[ESP_FIFO] = s->ti_buf[s->ti_rptr++];
}
if (s->ti_rptr == s->ti_wptr) {
s->ti_rptr = 0;
s->ti_wptr = 0;
}
val = s->rregs[ESP_FIFO];
break;
case ESP_RINTR:
/*
* Clear sequence step, interrupt register and all status bits
* except TC
*/
val = s->rregs[ESP_RINTR];
s->rregs[ESP_RINTR] = 0;
s->rregs[ESP_RSTAT] &= ~STAT_TC;
s->rregs[ESP_RSEQ] = SEQ_CD;
esp_lower_irq(s);
scsi: esp: Defer command completion until previous interrupts have been handled The guest OS reads RSTAT, RSEQ, and RINTR, and expects those registers to reflect a consistent state. However, it is possible that the registers can change after RSTAT was read, but before RINTR is read, when esp_command_complete() is called. Guest OS qemu -------- ---- [handle interrupt] Read RSTAT esp_command_complete() RSTAT = STAT_ST esp_dma_done() RSTAT |= STAT_TC RSEQ = 0 RINTR = INTR_BS Read RSEQ Read RINTR RINTR = 0 RSTAT &= ~STAT_TC RSEQ = SEQ_CD The guest OS would then try to handle INTR_BS combined with an old value of RSTAT. This sometimes resulted in lost events, spurious interrupts, guest OS confusion, and stalled SCSI operations. A typical guest error log (observed with various versions of Linux) looks as follows. scsi host1: Spurious irq, sreg=13. ... scsi host1: Aborting command [84531f10:2a] scsi host1: Current command [f882eea8:35] scsi host1: Queued command [84531f10:2a] scsi host1: Active command [f882eea8:35] scsi host1: Dumping command log scsi host1: ent[15] CMD val[44] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[00] event[0c] scsi host1: ent[16] CMD val[01] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[17] CMD val[43] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[18] EVENT val[0d] sreg[92] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c] ... Defer handling command completion until previous interrupts have been handled to fix the problem. Signed-off-by: Guenter Roeck <linux@roeck-us.net>
2018-11-29 17:17:42 +00:00
if (s->deferred_complete) {
esp_report_command_complete(s, s->deferred_status);
s->deferred_complete = false;
}
break;
case ESP_TCHI:
/* Return the unique id if the value has never been written */
if (!s->tchi_written) {
val = s->chip_id;
} else {
val = s->rregs[saddr];
}
break;
default:
val = s->rregs[saddr];
break;
}
trace_esp_mem_readb(saddr, val);
return val;
}
void esp_reg_write(ESPState *s, uint32_t saddr, uint64_t val)
{
trace_esp_mem_writeb(saddr, s->wregs[saddr], val);
switch (saddr) {
case ESP_TCHI:
s->tchi_written = true;
/* fall through */
case ESP_TCLO:
case ESP_TCMID:
s->rregs[ESP_RSTAT] &= ~STAT_TC;
break;
case ESP_FIFO:
if (s->do_cmd) {
if (s->cmdlen < ESP_CMDBUF_SZ) {
s->cmdbuf[s->cmdlen++] = val & 0xff;
} else {
trace_esp_error_fifo_overrun();
}
} else if (s->ti_wptr == TI_BUFSZ - 1) {
trace_esp_error_fifo_overrun();
} else {
s->ti_size++;
s->ti_buf[s->ti_wptr++] = val & 0xff;
}
break;
case ESP_CMD:
s->rregs[saddr] = val;
if (val & CMD_DMA) {
s->dma = 1;
/* Reload DMA counter. */
if (esp_get_stc(s) == 0) {
esp_set_tc(s, 0x10000);
} else {
esp_set_tc(s, esp_get_stc(s));
}
} else {
s->dma = 0;
}
switch (val & CMD_CMD) {
case CMD_NOP:
trace_esp_mem_writeb_cmd_nop(val);
break;
case CMD_FLUSH:
trace_esp_mem_writeb_cmd_flush(val);
/*s->ti_size = 0;*/
s->rregs[ESP_RINTR] = INTR_FC;
s->rregs[ESP_RSEQ] = 0;
s->rregs[ESP_RFLAGS] = 0;
break;
case CMD_RESET:
trace_esp_mem_writeb_cmd_reset(val);
esp_soft_reset(s);
break;
case CMD_BUSRESET:
trace_esp_mem_writeb_cmd_bus_reset(val);
s->rregs[ESP_RINTR] = INTR_RST;
if (!(s->wregs[ESP_CFG1] & CFG1_RESREPT)) {
esp_raise_irq(s);
}
break;
case CMD_TI:
trace_esp_mem_writeb_cmd_ti(val);
handle_ti(s);
break;
case CMD_ICCS:
trace_esp_mem_writeb_cmd_iccs(val);
write_response(s);
s->rregs[ESP_RINTR] = INTR_FC;
s->rregs[ESP_RSTAT] |= STAT_MI;
break;
case CMD_MSGACC:
trace_esp_mem_writeb_cmd_msgacc(val);
s->rregs[ESP_RINTR] = INTR_DC;
s->rregs[ESP_RSEQ] = 0;
s->rregs[ESP_RFLAGS] = 0;
esp_raise_irq(s);
break;
case CMD_PAD:
trace_esp_mem_writeb_cmd_pad(val);
s->rregs[ESP_RSTAT] = STAT_TC;
s->rregs[ESP_RINTR] = INTR_FC;
s->rregs[ESP_RSEQ] = 0;
break;
case CMD_SATN:
trace_esp_mem_writeb_cmd_satn(val);
break;
case CMD_RSTATN:
trace_esp_mem_writeb_cmd_rstatn(val);
break;
case CMD_SEL:
trace_esp_mem_writeb_cmd_sel(val);
handle_s_without_atn(s);
break;
case CMD_SELATN:
trace_esp_mem_writeb_cmd_selatn(val);
handle_satn(s);
break;
case CMD_SELATNS:
trace_esp_mem_writeb_cmd_selatns(val);
handle_satn_stop(s);
break;
case CMD_ENSEL:
trace_esp_mem_writeb_cmd_ensel(val);
s->rregs[ESP_RINTR] = 0;
break;
case CMD_DISSEL:
trace_esp_mem_writeb_cmd_dissel(val);
s->rregs[ESP_RINTR] = 0;
esp_raise_irq(s);
break;
default:
trace_esp_error_unhandled_command(val);
break;
}
break;
case ESP_WBUSID ... ESP_WSYNO:
break;
case ESP_CFG1:
case ESP_CFG2: case ESP_CFG3:
case ESP_RES3: case ESP_RES4:
s->rregs[saddr] = val;
break;
case ESP_WCCF ... ESP_WTEST:
break;
default:
trace_esp_error_invalid_write(val, saddr);
return;
}
s->wregs[saddr] = val;
}
static bool esp_mem_accepts(void *opaque, hwaddr addr,
unsigned size, bool is_write,
MemTxAttrs attrs)
{
return (size == 1) || (is_write && size == 4);
}
static bool esp_pdma_needed(void *opaque)
{
ESPState *s = opaque;
return s->dma_memory_read == NULL && s->dma_memory_write == NULL &&
s->dma_enabled;
}
static const VMStateDescription vmstate_esp_pdma = {
.name = "esp/pdma",
.version_id = 2,
.minimum_version_id = 2,
.needed = esp_pdma_needed,
.fields = (VMStateField[]) {
VMSTATE_INT32(pdma_origin, ESPState),
VMSTATE_UINT32(pdma_len, ESPState),
VMSTATE_UINT32(pdma_start, ESPState),
VMSTATE_UINT32(pdma_cur, ESPState),
VMSTATE_END_OF_LIST()
}
};
static bool esp_is_before_version_5(void *opaque, int version_id)
{
ESPState *s = ESP(opaque);
version_id = MIN(version_id, s->mig_version_id);
return version_id < 5;
}
static int esp_pre_save(void *opaque)
{
ESPState *s = ESP(opaque);
s->mig_version_id = vmstate_esp.version_id;
return 0;
}
static int esp_post_load(void *opaque, int version_id)
{
ESPState *s = ESP(opaque);
version_id = MIN(version_id, s->mig_version_id);
if (version_id < 5) {
esp_set_tc(s, s->mig_dma_left);
}
s->mig_version_id = vmstate_esp.version_id;
return 0;
}
const VMStateDescription vmstate_esp = {
.name = "esp",
.version_id = 5,
.minimum_version_id = 3,
.pre_save = esp_pre_save,
.post_load = esp_post_load,
.fields = (VMStateField[]) {
VMSTATE_BUFFER(rregs, ESPState),
VMSTATE_BUFFER(wregs, ESPState),
VMSTATE_INT32(ti_size, ESPState),
VMSTATE_UINT32(ti_rptr, ESPState),
VMSTATE_UINT32(ti_wptr, ESPState),
VMSTATE_BUFFER(ti_buf, ESPState),
VMSTATE_UINT32(status, ESPState),
scsi: esp: Defer command completion until previous interrupts have been handled The guest OS reads RSTAT, RSEQ, and RINTR, and expects those registers to reflect a consistent state. However, it is possible that the registers can change after RSTAT was read, but before RINTR is read, when esp_command_complete() is called. Guest OS qemu -------- ---- [handle interrupt] Read RSTAT esp_command_complete() RSTAT = STAT_ST esp_dma_done() RSTAT |= STAT_TC RSEQ = 0 RINTR = INTR_BS Read RSEQ Read RINTR RINTR = 0 RSTAT &= ~STAT_TC RSEQ = SEQ_CD The guest OS would then try to handle INTR_BS combined with an old value of RSTAT. This sometimes resulted in lost events, spurious interrupts, guest OS confusion, and stalled SCSI operations. A typical guest error log (observed with various versions of Linux) looks as follows. scsi host1: Spurious irq, sreg=13. ... scsi host1: Aborting command [84531f10:2a] scsi host1: Current command [f882eea8:35] scsi host1: Queued command [84531f10:2a] scsi host1: Active command [f882eea8:35] scsi host1: Dumping command log scsi host1: ent[15] CMD val[44] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[00] event[0c] scsi host1: ent[16] CMD val[01] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[17] CMD val[43] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[18] EVENT val[0d] sreg[92] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c] ... Defer handling command completion until previous interrupts have been handled to fix the problem. Signed-off-by: Guenter Roeck <linux@roeck-us.net>
2018-11-29 17:17:42 +00:00
VMSTATE_UINT32(deferred_status, ESPState),
VMSTATE_BOOL(deferred_complete, ESPState),
VMSTATE_UINT32(dma, ESPState),
VMSTATE_PARTIAL_BUFFER(cmdbuf, ESPState, 16),
VMSTATE_BUFFER_START_MIDDLE_V(cmdbuf, ESPState, 16, 4),
VMSTATE_UINT32(cmdlen, ESPState),
VMSTATE_UINT32(do_cmd, ESPState),
VMSTATE_UINT32_TEST(mig_dma_left, ESPState, esp_is_before_version_5),
VMSTATE_END_OF_LIST()
},
.subsections = (const VMStateDescription * []) {
&vmstate_esp_pdma,
NULL
}
};
static void sysbus_esp_mem_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
SysBusESPState *sysbus = opaque;
ESPState *s = ESP(&sysbus->esp);
uint32_t saddr;
saddr = addr >> sysbus->it_shift;
esp_reg_write(s, saddr, val);
}
static uint64_t sysbus_esp_mem_read(void *opaque, hwaddr addr,
unsigned int size)
{
SysBusESPState *sysbus = opaque;
ESPState *s = ESP(&sysbus->esp);
uint32_t saddr;
saddr = addr >> sysbus->it_shift;
return esp_reg_read(s, saddr);
}
static const MemoryRegionOps sysbus_esp_mem_ops = {
.read = sysbus_esp_mem_read,
.write = sysbus_esp_mem_write,
.endianness = DEVICE_NATIVE_ENDIAN,
.valid.accepts = esp_mem_accepts,
};
static void sysbus_esp_pdma_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
SysBusESPState *sysbus = opaque;
ESPState *s = ESP(&sysbus->esp);
trace_esp_pdma_write(size);
switch (size) {
case 1:
esp_pdma_write(s, val);
break;
case 2:
esp_pdma_write(s, val >> 8);
esp_pdma_write(s, val);
break;
}
if (s->pdma_len == 0 && s->pdma_cb) {
esp_lower_drq(s);
s->pdma_cb(s);
s->pdma_cb = NULL;
}
}
static uint64_t sysbus_esp_pdma_read(void *opaque, hwaddr addr,
unsigned int size)
{
SysBusESPState *sysbus = opaque;
ESPState *s = ESP(&sysbus->esp);
uint32_t dmalen = esp_get_tc(s);
uint64_t val = 0;
trace_esp_pdma_read(size);
if (dmalen == 0 || s->pdma_len == 0) {
return 0;
}
switch (size) {
case 1:
val = esp_pdma_read(s);
break;
case 2:
val = esp_pdma_read(s);
val = (val << 8) | esp_pdma_read(s);
break;
}
dmalen = esp_get_tc(s);
if (dmalen == 0 || (s->pdma_len == 0 && s->pdma_cb)) {
esp_lower_drq(s);
s->pdma_cb(s);
s->pdma_cb = NULL;
}
return val;
}
static const MemoryRegionOps sysbus_esp_pdma_ops = {
.read = sysbus_esp_pdma_read,
.write = sysbus_esp_pdma_write,
.endianness = DEVICE_NATIVE_ENDIAN,
.valid.min_access_size = 1,
.valid.max_access_size = 2,
};
static const struct SCSIBusInfo esp_scsi_info = {
.tcq = false,
.max_target = ESP_MAX_DEVS,
.max_lun = 7,
.transfer_data = esp_transfer_data,
.complete = esp_command_complete,
.cancel = esp_request_cancelled
};
static void sysbus_esp_gpio_demux(void *opaque, int irq, int level)
{
SysBusESPState *sysbus = SYSBUS_ESP(opaque);
ESPState *s = ESP(&sysbus->esp);
switch (irq) {
case 0:
parent_esp_reset(s, irq, level);
break;
case 1:
esp_dma_enable(opaque, irq, level);
break;
}
}
static void sysbus_esp_realize(DeviceState *dev, Error **errp)
{
SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
SysBusESPState *sysbus = SYSBUS_ESP(dev);
ESPState *s = ESP(&sysbus->esp);
if (!qdev_realize(DEVICE(s), NULL, errp)) {
return;
}
sysbus_init_irq(sbd, &s->irq);
sysbus_init_irq(sbd, &s->irq_data);
assert(sysbus->it_shift != -1);
s->chip_id = TCHI_FAS100A;
memory_region_init_io(&sysbus->iomem, OBJECT(sysbus), &sysbus_esp_mem_ops,
sysbus, "esp-regs", ESP_REGS << sysbus->it_shift);
sysbus_init_mmio(sbd, &sysbus->iomem);
memory_region_init_io(&sysbus->pdma, OBJECT(sysbus), &sysbus_esp_pdma_ops,
sysbus, "esp-pdma", 2);
sysbus_init_mmio(sbd, &sysbus->pdma);
qdev_init_gpio_in(dev, sysbus_esp_gpio_demux, 2);
scsi_bus_new(&s->bus, sizeof(s->bus), dev, &esp_scsi_info, NULL);
}
static void sysbus_esp_hard_reset(DeviceState *dev)
{
SysBusESPState *sysbus = SYSBUS_ESP(dev);
ESPState *s = ESP(&sysbus->esp);
esp_hard_reset(s);
}
static void sysbus_esp_init(Object *obj)
{
SysBusESPState *sysbus = SYSBUS_ESP(obj);
object_initialize_child(obj, "esp", &sysbus->esp, TYPE_ESP);
}
static const VMStateDescription vmstate_sysbus_esp_scsi = {
.name = "sysbusespscsi",
.version_id = 2,
scsi: esp: Defer command completion until previous interrupts have been handled The guest OS reads RSTAT, RSEQ, and RINTR, and expects those registers to reflect a consistent state. However, it is possible that the registers can change after RSTAT was read, but before RINTR is read, when esp_command_complete() is called. Guest OS qemu -------- ---- [handle interrupt] Read RSTAT esp_command_complete() RSTAT = STAT_ST esp_dma_done() RSTAT |= STAT_TC RSEQ = 0 RINTR = INTR_BS Read RSEQ Read RINTR RINTR = 0 RSTAT &= ~STAT_TC RSEQ = SEQ_CD The guest OS would then try to handle INTR_BS combined with an old value of RSTAT. This sometimes resulted in lost events, spurious interrupts, guest OS confusion, and stalled SCSI operations. A typical guest error log (observed with various versions of Linux) looks as follows. scsi host1: Spurious irq, sreg=13. ... scsi host1: Aborting command [84531f10:2a] scsi host1: Current command [f882eea8:35] scsi host1: Queued command [84531f10:2a] scsi host1: Active command [f882eea8:35] scsi host1: Dumping command log scsi host1: ent[15] CMD val[44] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[00] event[0c] scsi host1: ent[16] CMD val[01] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[17] CMD val[43] sreg[90] seqreg[00] sreg2[00] ireg[20] ss[02] event[0c] scsi host1: ent[18] EVENT val[0d] sreg[92] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c] ... Defer handling command completion until previous interrupts have been handled to fix the problem. Signed-off-by: Guenter Roeck <linux@roeck-us.net>
2018-11-29 17:17:42 +00:00
.minimum_version_id = 1,
.fields = (VMStateField[]) {
VMSTATE_UINT8_V(esp.mig_version_id, SysBusESPState, 2),
VMSTATE_STRUCT(esp, SysBusESPState, 0, vmstate_esp, ESPState),
VMSTATE_END_OF_LIST()
}
};
static void sysbus_esp_class_init(ObjectClass *klass, void *data)
{
DeviceClass *dc = DEVICE_CLASS(klass);
dc->realize = sysbus_esp_realize;
dc->reset = sysbus_esp_hard_reset;
dc->vmsd = &vmstate_sysbus_esp_scsi;
set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
}
static const TypeInfo sysbus_esp_info = {
.name = TYPE_SYSBUS_ESP,
.parent = TYPE_SYS_BUS_DEVICE,
.instance_init = sysbus_esp_init,
.instance_size = sizeof(SysBusESPState),
.class_init = sysbus_esp_class_init,
};
static void esp_class_init(ObjectClass *klass, void *data)
{
DeviceClass *dc = DEVICE_CLASS(klass);
/* internal device for sysbusesp/pciespscsi, not user-creatable */
dc->user_creatable = false;
set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
}
static const TypeInfo esp_info = {
.name = TYPE_ESP,
.parent = TYPE_DEVICE,
.instance_size = sizeof(ESPState),
.class_init = esp_class_init,
};
static void esp_register_types(void)
{
type_register_static(&sysbus_esp_info);
type_register_static(&esp_info);
}
type_init(esp_register_types)