freebsd-dev/sys/dev/isp/isp_freebsd.c

4911 lines
138 KiB
C
Raw Normal View History

/*-
* Copyright (c) 1997-2009 by Matthew Jacob
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice immediately at the beginning of the file, without modification,
* this list of conditions, and the following disclaimer.
2000-09-21 20:16:04 +00:00
* 2. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/*
* Platform (FreeBSD) dependent common attachment code for Qlogic adapters.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <dev/isp/isp_freebsd.h>
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
#include <sys/unistd.h>
#include <sys/kthread.h>
#include <sys/conf.h>
#include <sys/module.h>
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
#include <sys/ioccom.h>
#include <dev/isp/isp_ioctl.h>
#include <sys/devicestat.h>
#include <cam/cam_periph.h>
#include <cam/cam_xpt_periph.h>
#if __FreeBSD_version < 800002
#define THREAD_CREATE kthread_create
#else
#define THREAD_CREATE kproc_create
#endif
MODULE_VERSION(isp, 1);
MODULE_DEPEND(isp, cam, 1, 1, 1);
int isp_announced = 0;
int isp_loop_down_limit = 60; /* default loop down limit */
int isp_quickboot_time = 7; /* don't wait more than N secs for loop up */
int isp_gone_device_time = 30; /* grace time before reporting device lost */
static const char prom3[] = "Chan %d [%u] PortID 0x%06x Departed because of %s";
static void isp_freeze_loopdown(ispsoftc_t *, int);
static void isp_loop_changed(ispsoftc_t *isp, int chan);
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
static d_ioctl_t ispioctl;
static void isp_intr_enable(void *);
static void isp_cam_async(void *, uint32_t, struct cam_path *, void *);
static void isp_poll(struct cam_sim *);
static timeout_t isp_watchdog;
static timeout_t isp_gdt;
static task_fn_t isp_gdt_task;
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
static void isp_kthread(void *);
static void isp_action(struct cam_sim *, union ccb *);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
static int isp_timer_count;
static void isp_timer(void *);
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
static struct cdevsw isp_cdevsw = {
.d_version = D_VERSION,
.d_ioctl = ispioctl,
.d_name = "isp",
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
};
static int
isp_role_sysctl(SYSCTL_HANDLER_ARGS)
{
ispsoftc_t *isp = (ispsoftc_t *)arg1;
int chan = arg2;
int error, old, value;
value = FCPARAM(isp, chan)->role;
error = sysctl_handle_int(oidp, &value, 0, req);
if ((error != 0) || (req->newptr == NULL))
return (error);
if (value < ISP_ROLE_NONE || value > ISP_ROLE_BOTH)
return (EINVAL);
ISP_LOCK(isp);
old = FCPARAM(isp, chan)->role;
/* We don't allow target mode switch from here. */
value = (old & ISP_ROLE_TARGET) | (value & ISP_ROLE_INITIATOR);
/* If nothing has changed -- we are done. */
if (value == old) {
ISP_UNLOCK(isp);
return (0);
}
/* Actually change the role. */
error = isp_control(isp, ISPCTL_CHANGE_ROLE, chan, value);
ISP_UNLOCK(isp);
return (error);
}
static int
isp_attach_chan(ispsoftc_t *isp, struct cam_devq *devq, int chan)
{
struct ccb_setasync csa;
struct cam_sim *sim;
struct cam_path *path;
/*
* Construct our SIM entry.
*/
sim = cam_sim_alloc(isp_action, isp_poll, "isp", isp, device_get_unit(isp->isp_dev), &isp->isp_osinfo.lock, isp->isp_maxcmds, isp->isp_maxcmds, devq);
if (sim == NULL) {
return (ENOMEM);
}
ISP_LOCK(isp);
if (xpt_bus_register(sim, isp->isp_dev, chan) != CAM_SUCCESS) {
ISP_UNLOCK(isp);
cam_sim_free(sim, FALSE);
return (EIO);
}
ISP_UNLOCK(isp);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (xpt_create_path(&path, NULL, cam_sim_path(sim), CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD) != CAM_REQ_CMP) {
ISP_LOCK(isp);
xpt_bus_deregister(cam_sim_path(sim));
ISP_UNLOCK(isp);
cam_sim_free(sim, FALSE);
return (ENXIO);
}
xpt_setup_ccb(&csa.ccb_h, path, 5);
csa.ccb_h.func_code = XPT_SASYNC_CB;
csa.event_enable = AC_LOST_DEVICE;
csa.callback = isp_cam_async;
csa.callback_arg = sim;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_LOCK(isp);
xpt_action((union ccb *)&csa);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_UNLOCK(isp);
if (IS_SCSI(isp)) {
struct isp_spi *spi = ISP_SPI_PC(isp, chan);
spi->sim = sim;
spi->path = path;
} else {
fcparam *fcp = FCPARAM(isp, chan);
struct isp_fc *fc = ISP_FC_PC(isp, chan);
struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(isp->isp_osinfo.dev);
struct sysctl_oid *tree = device_get_sysctl_tree(isp->isp_osinfo.dev);
char name[16];
ISP_LOCK(isp);
fc->sim = sim;
fc->path = path;
fc->isp = isp;
fc->ready = 1;
callout_init_mtx(&fc->gdt, &isp->isp_osinfo.lock, 0);
TASK_INIT(&fc->gtask, 1, isp_gdt_task, fc);
isp_loop_changed(isp, chan);
ISP_UNLOCK(isp);
if (THREAD_CREATE(isp_kthread, fc, &fc->kproc, 0, 0, "%s: fc_thrd%d", device_get_nameunit(isp->isp_osinfo.dev), chan)) {
xpt_free_path(fc->path);
ISP_LOCK(isp);
xpt_bus_deregister(cam_sim_path(fc->sim));
ISP_UNLOCK(isp);
cam_sim_free(fc->sim, FALSE);
return (ENOMEM);
}
fc->num_threads += 1;
if (chan > 0) {
snprintf(name, sizeof(name), "chan%d", chan);
tree = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(tree),
OID_AUTO, name, CTLFLAG_RW, 0, "Virtual channel");
}
SYSCTL_ADD_QUAD(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"wwnn", CTLFLAG_RD, &fcp->isp_wwnn,
"World Wide Node Name");
SYSCTL_ADD_QUAD(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"wwpn", CTLFLAG_RD, &fcp->isp_wwpn,
"World Wide Port Name");
SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"loop_down_limit", CTLFLAG_RW, &fc->loop_down_limit, 0,
"Loop Down Limit");
SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"gone_device_time", CTLFLAG_RW, &fc->gone_device_time, 0,
"Gone Device Time");
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
#if defined(ISP_TARGET_MODE) && defined(DEBUG)
SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"inject_lost_data_frame", CTLFLAG_RW, &fc->inject_lost_data_frame, 0,
"Cause a Lost Frame on a Read");
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
#endif
SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"role", CTLTYPE_INT | CTLFLAG_RW, isp, chan,
isp_role_sysctl, "I", "Current role");
SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"speed", CTLFLAG_RD, &fcp->isp_gbspeed, 0,
"Connection speed in gigabits");
SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"linkstate", CTLFLAG_RD, &fcp->isp_linkstate, 0,
"Link state");
SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"fwstate", CTLFLAG_RD, &fcp->isp_fwstate, 0,
"Firmware state");
SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"loopstate", CTLFLAG_RD, &fcp->isp_loopstate, 0,
"Loop state");
SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"topo", CTLFLAG_RD, &fcp->isp_topo, 0,
"Connection topology");
}
return (0);
}
static void
isp_detach_chan(ispsoftc_t *isp, int chan)
{
struct cam_sim *sim;
struct cam_path *path;
struct ccb_setasync csa;
int *num_threads;
ISP_GET_PC(isp, chan, sim, sim);
ISP_GET_PC(isp, chan, path, path);
ISP_GET_PC_ADDR(isp, chan, num_threads, num_threads);
xpt_setup_ccb(&csa.ccb_h, path, 5);
csa.ccb_h.func_code = XPT_SASYNC_CB;
csa.event_enable = 0;
csa.callback = isp_cam_async;
csa.callback_arg = sim;
xpt_action((union ccb *)&csa);
xpt_free_path(path);
xpt_bus_deregister(cam_sim_path(sim));
cam_sim_free(sim, FALSE);
/* Wait for the channel's spawned threads to exit. */
wakeup(isp->isp_osinfo.pc.ptr);
while (*num_threads != 0)
mtx_sleep(isp, &isp->isp_osinfo.lock, PRIBIO, "isp_reap", 100);
}
int
isp_attach(ispsoftc_t *isp)
{
const char *nu = device_get_nameunit(isp->isp_osinfo.dev);
int du = device_get_unit(isp->isp_dev);
int chan;
isp->isp_osinfo.ehook.ich_func = isp_intr_enable;
isp->isp_osinfo.ehook.ich_arg = isp;
/*
* Haha. Set this first, because if we're loaded as a module isp_intr_enable
* will be called right awawy, which will clear isp_osinfo.ehook_active,
* which would be unwise to then set again later.
*/
isp->isp_osinfo.ehook_active = 1;
if (config_intrhook_establish(&isp->isp_osinfo.ehook) != 0) {
isp_prt(isp, ISP_LOGERR, "could not establish interrupt enable hook");
return (-EIO);
}
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
/*
* Create the device queue for our SIM(s).
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
*/
isp->isp_osinfo.devq = cam_simq_alloc(isp->isp_maxcmds);
if (isp->isp_osinfo.devq == NULL) {
config_intrhook_disestablish(&isp->isp_osinfo.ehook);
return (EIO);
}
for (chan = 0; chan < isp->isp_nchan; chan++) {
if (isp_attach_chan(isp, isp->isp_osinfo.devq, chan)) {
goto unwind;
}
}
callout_init_mtx(&isp->isp_osinfo.tmo, &isp->isp_osinfo.lock, 0);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_timer_count = hz >> 2;
callout_reset(&isp->isp_osinfo.tmo, isp_timer_count, isp_timer, isp);
isp->isp_osinfo.timer_active = 1;
isp->isp_osinfo.cdev = make_dev(&isp_cdevsw, du, UID_ROOT, GID_OPERATOR, 0600, "%s", nu);
if (isp->isp_osinfo.cdev) {
isp->isp_osinfo.cdev->si_drv1 = isp;
}
return (0);
unwind:
while (--chan >= 0) {
struct cam_sim *sim;
struct cam_path *path;
ISP_GET_PC(isp, chan, sim, sim);
ISP_GET_PC(isp, chan, path, path);
xpt_free_path(path);
ISP_LOCK(isp);
xpt_bus_deregister(cam_sim_path(sim));
ISP_UNLOCK(isp);
cam_sim_free(sim, FALSE);
}
if (isp->isp_osinfo.ehook_active) {
config_intrhook_disestablish(&isp->isp_osinfo.ehook);
isp->isp_osinfo.ehook_active = 0;
}
if (isp->isp_osinfo.cdev) {
destroy_dev(isp->isp_osinfo.cdev);
isp->isp_osinfo.cdev = NULL;
}
cam_simq_free(isp->isp_osinfo.devq);
isp->isp_osinfo.devq = NULL;
return (-1);
}
int
isp_detach(ispsoftc_t *isp)
{
struct cam_sim *sim;
int chan;
ISP_LOCK(isp);
for (chan = isp->isp_nchan - 1; chan >= 0; chan -= 1) {
ISP_GET_PC(isp, chan, sim, sim);
if (sim->refcount > 2) {
ISP_UNLOCK(isp);
return (EBUSY);
}
}
/* Tell spawned threads that we're exiting. */
isp->isp_osinfo.is_exiting = 1;
if (isp->isp_osinfo.timer_active) {
callout_stop(&isp->isp_osinfo.tmo);
isp->isp_osinfo.timer_active = 0;
}
for (chan = isp->isp_nchan - 1; chan >= 0; chan -= 1)
isp_detach_chan(isp, chan);
ISP_UNLOCK(isp);
if (isp->isp_osinfo.cdev) {
destroy_dev(isp->isp_osinfo.cdev);
isp->isp_osinfo.cdev = NULL;
}
if (isp->isp_osinfo.ehook_active) {
config_intrhook_disestablish(&isp->isp_osinfo.ehook);
isp->isp_osinfo.ehook_active = 0;
}
if (isp->isp_osinfo.devq != NULL) {
cam_simq_free(isp->isp_osinfo.devq);
isp->isp_osinfo.devq = NULL;
}
return (0);
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
}
static void
isp_freeze_loopdown(ispsoftc_t *isp, int chan)
{
if (IS_FC(isp)) {
struct isp_fc *fc = ISP_FC_PC(isp, chan);
if (fc->simqfrozen == 0) {
isp_prt(isp, ISP_LOGDEBUG0,
"Chan %d Freeze simq (loopdown)", chan);
fc->simqfrozen = SIMQFRZ_LOOPDOWN;
#if __FreeBSD_version >= 1000039
xpt_hold_boot();
#endif
xpt_freeze_simq(fc->sim, 1);
} else {
isp_prt(isp, ISP_LOGDEBUG0,
"Chan %d Mark simq frozen (loopdown)", chan);
fc->simqfrozen |= SIMQFRZ_LOOPDOWN;
}
}
}
static void
isp_unfreeze_loopdown(ispsoftc_t *isp, int chan)
{
if (IS_FC(isp)) {
struct isp_fc *fc = ISP_FC_PC(isp, chan);
int wasfrozen = fc->simqfrozen & SIMQFRZ_LOOPDOWN;
fc->simqfrozen &= ~SIMQFRZ_LOOPDOWN;
if (wasfrozen && fc->simqfrozen == 0) {
isp_prt(isp, ISP_LOGDEBUG0,
"Chan %d Release simq", chan);
xpt_release_simq(fc->sim, 1);
#if __FreeBSD_version >= 1000039
xpt_release_boot();
#endif
}
}
}
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
static int
ispioctl(struct cdev *dev, u_long c, caddr_t addr, int flags, struct thread *td)
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
{
ispsoftc_t *isp;
int nr, chan, retval = ENOTTY;
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
isp = dev->si_drv1;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
2006-03-03 07:04:43 +00:00
switch (c) {
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
case ISP_SDBLEV:
{
int olddblev = isp->isp_dblev;
isp->isp_dblev = *(int *)addr;
*(int *)addr = olddblev;
retval = 0;
break;
}
case ISP_GETROLE:
chan = *(int *)addr;
if (chan < 0 || chan >= isp->isp_nchan) {
retval = -ENXIO;
break;
}
if (IS_FC(isp)) {
*(int *)addr = FCPARAM(isp, chan)->role;
} else {
*(int *)addr = ISP_ROLE_INITIATOR;
}
retval = 0;
break;
case ISP_SETROLE:
if (IS_SCSI(isp))
break;
nr = *(int *)addr;
chan = nr >> 8;
if (chan < 0 || chan >= isp->isp_nchan) {
retval = -ENXIO;
break;
}
nr &= 0xff;
if (nr & ~(ISP_ROLE_INITIATOR|ISP_ROLE_TARGET)) {
retval = EINVAL;
break;
}
ISP_LOCK(isp);
*(int *)addr = FCPARAM(isp, chan)->role;
retval = isp_control(isp, ISPCTL_CHANGE_ROLE, chan, nr);
ISP_UNLOCK(isp);
retval = 0;
break;
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
case ISP_RESETHBA:
ISP_LOCK(isp);
isp_reinit(isp, 0);
ISP_UNLOCK(isp);
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
retval = 0;
break;
case ISP_RESCAN:
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
if (IS_FC(isp)) {
chan = *(int *)addr;
if (chan < 0 || chan >= isp->isp_nchan) {
retval = -ENXIO;
break;
}
ISP_LOCK(isp);
if (isp_fc_runstate(isp, chan, 5 * 1000000) != LOOP_READY) {
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
retval = EIO;
} else {
retval = 0;
}
ISP_UNLOCK(isp);
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
}
break;
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
case ISP_FC_LIP:
if (IS_FC(isp)) {
chan = *(int *)addr;
if (chan < 0 || chan >= isp->isp_nchan) {
retval = -ENXIO;
break;
}
ISP_LOCK(isp);
if (isp_control(isp, ISPCTL_SEND_LIP, chan)) {
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
retval = EIO;
} else {
retval = 0;
}
ISP_UNLOCK(isp);
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
}
break;
case ISP_FC_GETDINFO:
{
struct isp_fc_device *ifc = (struct isp_fc_device *) addr;
fcportdb_t *lp;
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
if (IS_SCSI(isp)) {
break;
}
if (ifc->loopid >= MAX_FC_TARG) {
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
retval = EINVAL;
break;
}
lp = &FCPARAM(isp, ifc->chan)->portdb[ifc->loopid];
if (lp->state != FC_PORTDB_STATE_NIL) {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ifc->role = (lp->prli_word3 & SVC3_ROLE_MASK) >> SVC3_ROLE_SHIFT;
ifc->loopid = lp->handle;
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
ifc->portid = lp->portid;
ifc->node_wwn = lp->node_wwn;
ifc->port_wwn = lp->port_wwn;
retval = 0;
} else {
retval = ENODEV;
}
break;
}
case ISP_GET_STATS:
{
isp_stats_t *sp = (isp_stats_t *) addr;
ISP_MEMZERO(sp, sizeof (*sp));
sp->isp_stat_version = ISP_STATS_VERSION;
sp->isp_type = isp->isp_type;
sp->isp_revision = isp->isp_revision;
ISP_LOCK(isp);
sp->isp_stats[ISP_INTCNT] = isp->isp_intcnt;
sp->isp_stats[ISP_INTBOGUS] = isp->isp_intbogus;
sp->isp_stats[ISP_INTMBOXC] = isp->isp_intmboxc;
sp->isp_stats[ISP_INGOASYNC] = isp->isp_intoasync;
sp->isp_stats[ISP_RSLTCCMPLT] = isp->isp_rsltccmplt;
sp->isp_stats[ISP_FPHCCMCPLT] = isp->isp_fphccmplt;
sp->isp_stats[ISP_RSCCHIWAT] = isp->isp_rscchiwater;
sp->isp_stats[ISP_FPCCHIWAT] = isp->isp_fpcchiwater;
ISP_UNLOCK(isp);
retval = 0;
break;
}
case ISP_CLR_STATS:
ISP_LOCK(isp);
isp->isp_intcnt = 0;
isp->isp_intbogus = 0;
isp->isp_intmboxc = 0;
isp->isp_intoasync = 0;
isp->isp_rsltccmplt = 0;
isp->isp_fphccmplt = 0;
isp->isp_rscchiwater = 0;
isp->isp_fpcchiwater = 0;
ISP_UNLOCK(isp);
retval = 0;
break;
case ISP_FC_GETHINFO:
{
struct isp_hba_device *hba = (struct isp_hba_device *) addr;
int chan = hba->fc_channel;
if (chan < 0 || chan >= isp->isp_nchan) {
retval = ENXIO;
break;
}
hba->fc_fw_major = ISP_FW_MAJORX(isp->isp_fwrev);
hba->fc_fw_minor = ISP_FW_MINORX(isp->isp_fwrev);
hba->fc_fw_micro = ISP_FW_MICROX(isp->isp_fwrev);
hba->fc_nchannels = isp->isp_nchan;
if (IS_FC(isp)) {
hba->fc_nports = MAX_FC_TARG;
hba->fc_speed = FCPARAM(isp, hba->fc_channel)->isp_gbspeed;
hba->fc_topology = FCPARAM(isp, chan)->isp_topo + 1;
hba->fc_loopid = FCPARAM(isp, chan)->isp_loopid;
hba->nvram_node_wwn = FCPARAM(isp, chan)->isp_wwnn_nvram;
hba->nvram_port_wwn = FCPARAM(isp, chan)->isp_wwpn_nvram;
hba->active_node_wwn = FCPARAM(isp, chan)->isp_wwnn;
hba->active_port_wwn = FCPARAM(isp, chan)->isp_wwpn;
} else {
hba->fc_nports = MAX_TARGETS;
hba->fc_speed = 0;
hba->fc_topology = 0;
hba->nvram_node_wwn = 0ull;
hba->nvram_port_wwn = 0ull;
hba->active_node_wwn = 0ull;
hba->active_port_wwn = 0ull;
}
retval = 0;
break;
}
case ISP_TSK_MGMT:
{
int needmarker;
struct isp_fc_tsk_mgmt *fct = (struct isp_fc_tsk_mgmt *) addr;
uint16_t nphdl;
mbreg_t mbs;
if (IS_SCSI(isp)) {
break;
}
chan = fct->chan;
if (chan < 0 || chan >= isp->isp_nchan) {
retval = -ENXIO;
break;
}
needmarker = retval = 0;
nphdl = fct->loopid;
ISP_LOCK(isp);
if (IS_24XX(isp)) {
void *reqp;
uint8_t resp[QENTRY_LEN];
isp24xx_tmf_t tmf;
isp24xx_statusreq_t sp;
fcparam *fcp = FCPARAM(isp, chan);
fcportdb_t *lp;
int i;
for (i = 0; i < MAX_FC_TARG; i++) {
lp = &fcp->portdb[i];
if (lp->handle == nphdl) {
break;
}
}
if (i == MAX_FC_TARG) {
retval = ENXIO;
ISP_UNLOCK(isp);
break;
}
ISP_MEMZERO(&tmf, sizeof(tmf));
tmf.tmf_header.rqs_entry_type = RQSTYPE_TSK_MGMT;
tmf.tmf_header.rqs_entry_count = 1;
tmf.tmf_nphdl = lp->handle;
tmf.tmf_delay = 2;
tmf.tmf_timeout = 4;
tmf.tmf_tidlo = lp->portid;
tmf.tmf_tidhi = lp->portid >> 16;
tmf.tmf_vpidx = ISP_GET_VPIDX(isp, chan);
tmf.tmf_lun[1] = fct->lun & 0xff;
if (fct->lun >= 256) {
tmf.tmf_lun[0] = 0x40 | (fct->lun >> 8);
}
switch (fct->action) {
case IPT_CLEAR_ACA:
tmf.tmf_flags = ISP24XX_TMF_CLEAR_ACA;
break;
case IPT_TARGET_RESET:
tmf.tmf_flags = ISP24XX_TMF_TARGET_RESET;
needmarker = 1;
break;
case IPT_LUN_RESET:
tmf.tmf_flags = ISP24XX_TMF_LUN_RESET;
needmarker = 1;
break;
case IPT_CLEAR_TASK_SET:
tmf.tmf_flags = ISP24XX_TMF_CLEAR_TASK_SET;
needmarker = 1;
break;
case IPT_ABORT_TASK_SET:
tmf.tmf_flags = ISP24XX_TMF_ABORT_TASK_SET;
needmarker = 1;
break;
default:
retval = EINVAL;
break;
}
if (retval) {
ISP_UNLOCK(isp);
break;
}
/* Prepare space for response in memory */
memset(resp, 0xff, sizeof(resp));
tmf.tmf_handle = isp_allocate_handle(isp, resp,
ISP_HANDLE_CTRL);
if (tmf.tmf_handle == 0) {
isp_prt(isp, ISP_LOGERR,
"%s: TMF of Chan %d out of handles",
__func__, chan);
ISP_UNLOCK(isp);
retval = ENOMEM;
break;
}
/* Send request and wait for response. */
reqp = isp_getrqentry(isp);
if (reqp == NULL) {
isp_prt(isp, ISP_LOGERR,
"%s: TMF of Chan %d out of rqent",
__func__, chan);
isp_destroy_handle(isp, tmf.tmf_handle);
ISP_UNLOCK(isp);
retval = EIO;
break;
}
isp_put_24xx_tmf(isp, &tmf, (isp24xx_tmf_t *)reqp);
if (isp->isp_dblev & ISP_LOGDEBUG1)
isp_print_bytes(isp, "IOCB TMF", QENTRY_LEN, reqp);
ISP_SYNC_REQUEST(isp);
if (msleep(resp, &isp->isp_lock, 0, "TMF", 5*hz) == EWOULDBLOCK) {
isp_prt(isp, ISP_LOGERR,
"%s: TMF of Chan %d timed out",
__func__, chan);
isp_destroy_handle(isp, tmf.tmf_handle);
ISP_UNLOCK(isp);
retval = EIO;
break;
}
if (isp->isp_dblev & ISP_LOGDEBUG1)
isp_print_bytes(isp, "IOCB TMF response", QENTRY_LEN, resp);
isp_get_24xx_response(isp, (isp24xx_statusreq_t *)resp, &sp);
if (sp.req_completion_status != 0)
retval = EIO;
else if (needmarker)
fcp->sendmarker = 1;
} else {
MBSINIT(&mbs, 0, MBLOGALL, 0);
if (ISP_CAP_2KLOGIN(isp) == 0) {
nphdl <<= 8;
}
switch (fct->action) {
case IPT_CLEAR_ACA:
mbs.param[0] = MBOX_CLEAR_ACA;
mbs.param[1] = nphdl;
mbs.param[2] = fct->lun;
break;
case IPT_TARGET_RESET:
mbs.param[0] = MBOX_TARGET_RESET;
mbs.param[1] = nphdl;
needmarker = 1;
break;
case IPT_LUN_RESET:
mbs.param[0] = MBOX_LUN_RESET;
mbs.param[1] = nphdl;
mbs.param[2] = fct->lun;
needmarker = 1;
break;
case IPT_CLEAR_TASK_SET:
mbs.param[0] = MBOX_CLEAR_TASK_SET;
mbs.param[1] = nphdl;
mbs.param[2] = fct->lun;
needmarker = 1;
break;
case IPT_ABORT_TASK_SET:
mbs.param[0] = MBOX_ABORT_TASK_SET;
mbs.param[1] = nphdl;
mbs.param[2] = fct->lun;
needmarker = 1;
break;
default:
retval = EINVAL;
break;
}
if (retval == 0) {
if (needmarker) {
FCPARAM(isp, chan)->sendmarker = 1;
}
retval = isp_control(isp, ISPCTL_RUN_MBOXCMD, &mbs);
if (retval) {
retval = EIO;
}
}
}
ISP_UNLOCK(isp);
break;
}
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
default:
break;
}
return (retval);
}
static void
isp_intr_enable(void *arg)
{
int chan;
ispsoftc_t *isp = arg;
ISP_LOCK(isp);
if (IS_FC(isp)) {
for (chan = 0; chan < isp->isp_nchan; chan++) {
if (FCPARAM(isp, chan)->role != ISP_ROLE_NONE) {
ISP_ENABLE_INTS(isp);
break;
}
}
} else {
ISP_ENABLE_INTS(isp);
}
isp->isp_osinfo.ehook_active = 0;
ISP_UNLOCK(isp);
/* Release our hook so that the boot can continue. */
config_intrhook_disestablish(&isp->isp_osinfo.ehook);
}
/*
* Local Inlines
*/
static ISP_INLINE int isp_get_pcmd(ispsoftc_t *, union ccb *);
static ISP_INLINE void isp_free_pcmd(ispsoftc_t *, union ccb *);
static ISP_INLINE int
isp_get_pcmd(ispsoftc_t *isp, union ccb *ccb)
{
ISP_PCMD(ccb) = isp->isp_osinfo.pcmd_free;
if (ISP_PCMD(ccb) == NULL) {
return (-1);
}
isp->isp_osinfo.pcmd_free = ((struct isp_pcmd *)ISP_PCMD(ccb))->next;
return (0);
}
static ISP_INLINE void
isp_free_pcmd(ispsoftc_t *isp, union ccb *ccb)
{
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (ISP_PCMD(ccb)) {
#ifdef ISP_TARGET_MODE
PISP_PCMD(ccb)->datalen = 0;
PISP_PCMD(ccb)->totslen = 0;
PISP_PCMD(ccb)->cumslen = 0;
PISP_PCMD(ccb)->crn = 0;
#endif
PISP_PCMD(ccb)->next = isp->isp_osinfo.pcmd_free;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp->isp_osinfo.pcmd_free = ISP_PCMD(ccb);
ISP_PCMD(ccb) = NULL;
}
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
/*
* Put the target mode functions here, because some are inlines
*/
#ifdef ISP_TARGET_MODE
static ISP_INLINE int is_lun_enabled(ispsoftc_t *, int, lun_id_t);
static ISP_INLINE tstate_t *get_lun_statep(ispsoftc_t *, int, lun_id_t);
static ISP_INLINE tstate_t *get_lun_statep_from_tag(ispsoftc_t *, int, uint32_t);
static ISP_INLINE void rls_lun_statep(ispsoftc_t *, tstate_t *);
static ISP_INLINE inot_private_data_t *get_ntp_from_tagdata(ispsoftc_t *, uint32_t, uint32_t, tstate_t **);
static ISP_INLINE atio_private_data_t *isp_get_atpd(ispsoftc_t *, tstate_t *, uint32_t);
static ISP_INLINE atio_private_data_t *isp_find_atpd(ispsoftc_t *, tstate_t *, uint32_t);
static ISP_INLINE void isp_put_atpd(ispsoftc_t *, tstate_t *, atio_private_data_t *);
static ISP_INLINE inot_private_data_t *isp_get_ntpd(ispsoftc_t *, tstate_t *);
static ISP_INLINE inot_private_data_t *isp_find_ntpd(ispsoftc_t *, tstate_t *, uint32_t, uint32_t);
static ISP_INLINE void isp_put_ntpd(ispsoftc_t *, tstate_t *, inot_private_data_t *);
static cam_status create_lun_state(ispsoftc_t *, int, struct cam_path *, tstate_t **);
static void destroy_lun_state(ispsoftc_t *, tstate_t *);
static void isp_enable_lun(ispsoftc_t *, union ccb *);
static void isp_disable_lun(ispsoftc_t *, union ccb *);
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
static timeout_t isp_refire_putback_atio;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
static timeout_t isp_refire_notify_ack;
static void isp_complete_ctio(union ccb *);
static void isp_target_putback_atio(union ccb *);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
enum Start_Ctio_How { FROM_CAM, FROM_TIMER, FROM_SRR, FROM_CTIO_DONE };
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
static void isp_target_start_ctio(ispsoftc_t *, union ccb *, enum Start_Ctio_How);
static void isp_handle_platform_atio2(ispsoftc_t *, at2_entry_t *);
static void isp_handle_platform_atio7(ispsoftc_t *, at7_entry_t *);
static void isp_handle_platform_ctio(ispsoftc_t *, void *);
static void isp_handle_platform_notify_fc(ispsoftc_t *, in_fcentry_t *);
static void isp_handle_platform_notify_24xx(ispsoftc_t *, in_fcentry_24xx_t *);
static int isp_handle_platform_target_notify_ack(ispsoftc_t *, isp_notify_t *, uint32_t rsp);
static void isp_handle_platform_target_tmf(ispsoftc_t *, isp_notify_t *);
static void isp_target_mark_aborted(ispsoftc_t *, union ccb *);
static void isp_target_mark_aborted_early(ispsoftc_t *, tstate_t *, uint32_t);
static ISP_INLINE int
is_lun_enabled(ispsoftc_t *isp, int bus, lun_id_t lun)
{
tstate_t *tptr;
struct tslist *lhp;
ISP_GET_PC_ADDR(isp, bus, lun_hash[LUN_HASH_FUNC(lun)], lhp);
SLIST_FOREACH(tptr, lhp, next) {
if (tptr->ts_lun == lun) {
return (1);
}
}
return (0);
}
static void
dump_tstates(ispsoftc_t *isp, int bus)
{
int i, j;
struct tslist *lhp;
tstate_t *tptr = NULL;
if (bus >= isp->isp_nchan) {
return;
}
for (i = 0; i < LUN_HASH_SIZE; i++) {
ISP_GET_PC_ADDR(isp, bus, lun_hash[i], lhp);
j = 0;
SLIST_FOREACH(tptr, lhp, next) {
xpt_print(tptr->owner, "[%d, %d] atio_cnt=%d inot_cnt=%d\n", i, j, tptr->atio_count, tptr->inot_count);
j++;
}
}
}
static ISP_INLINE tstate_t *
get_lun_statep(ispsoftc_t *isp, int bus, lun_id_t lun)
{
tstate_t *tptr = NULL;
struct tslist *lhp;
if (bus < isp->isp_nchan) {
ISP_GET_PC_ADDR(isp, bus, lun_hash[LUN_HASH_FUNC(lun)], lhp);
SLIST_FOREACH(tptr, lhp, next) {
if (tptr->ts_lun == lun) {
tptr->hold++;
return (tptr);
}
}
}
return (NULL);
}
static ISP_INLINE tstate_t *
get_lun_statep_from_tag(ispsoftc_t *isp, int bus, uint32_t tagval)
{
tstate_t *tptr = NULL;
atio_private_data_t *atp;
struct tslist *lhp;
int i;
if (bus < isp->isp_nchan && tagval != 0) {
for (i = 0; i < LUN_HASH_SIZE; i++) {
ISP_GET_PC_ADDR(isp, bus, lun_hash[i], lhp);
SLIST_FOREACH(tptr, lhp, next) {
atp = isp_find_atpd(isp, tptr, tagval);
if (atp) {
tptr->hold++;
return (tptr);
}
}
}
}
return (NULL);
}
static ISP_INLINE inot_private_data_t *
get_ntp_from_tagdata(ispsoftc_t *isp, uint32_t tag_id, uint32_t seq_id, tstate_t **rslt)
{
inot_private_data_t *ntp;
tstate_t *tptr;
struct tslist *lhp;
int bus, i;
for (bus = 0; bus < isp->isp_nchan; bus++) {
for (i = 0; i < LUN_HASH_SIZE; i++) {
ISP_GET_PC_ADDR(isp, bus, lun_hash[i], lhp);
SLIST_FOREACH(tptr, lhp, next) {
ntp = isp_find_ntpd(isp, tptr, tag_id, seq_id);
if (ntp) {
*rslt = tptr;
tptr->hold++;
return (ntp);
}
}
}
}
return (NULL);
}
static ISP_INLINE void
rls_lun_statep(ispsoftc_t *isp, tstate_t *tptr)
{
KASSERT((tptr->hold), ("tptr not held"));
tptr->hold--;
}
static void
isp_tmcmd_restart(ispsoftc_t *isp)
{
inot_private_data_t *ntp;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
inot_private_data_t *restart_queue;
tstate_t *tptr;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
union ccb *ccb;
struct tslist *lhp;
int bus, i;
for (bus = 0; bus < isp->isp_nchan; bus++) {
for (i = 0; i < LUN_HASH_SIZE; i++) {
ISP_GET_PC_ADDR(isp, bus, lun_hash[i], lhp);
SLIST_FOREACH(tptr, lhp, next) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if ((restart_queue = tptr->restart_queue) != NULL)
tptr->restart_queue = NULL;
while (restart_queue) {
ntp = restart_queue;
restart_queue = ntp->rd.nt.nt_hba;
if (IS_24XX(isp)) {
isp_prt(isp, ISP_LOGTDEBUG0, "%s: restarting resrc deprived %x", __func__, ((at7_entry_t *)ntp->rd.data)->at_rxid);
isp_handle_platform_atio7(isp, (at7_entry_t *) ntp->rd.data);
} else {
isp_prt(isp, ISP_LOGTDEBUG0, "%s: restarting resrc deprived %x", __func__, ((at2_entry_t *)ntp->rd.data)->at_rxid);
isp_handle_platform_atio2(isp, (at2_entry_t *) ntp->rd.data);
}
isp_put_ntpd(isp, tptr, ntp);
if (tptr->restart_queue && restart_queue != NULL) {
ntp = tptr->restart_queue;
tptr->restart_queue = restart_queue;
while (restart_queue->rd.nt.nt_hba) {
restart_queue = restart_queue->rd.nt.nt_hba;
}
restart_queue->rd.nt.nt_hba = ntp;
break;
}
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
/*
* We only need to do this once per tptr
*/
if (!TAILQ_EMPTY(&tptr->waitq)) {
ccb = (union ccb *)TAILQ_LAST(&tptr->waitq, isp_ccbq);
TAILQ_REMOVE(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
isp_target_start_ctio(isp, ccb, FROM_TIMER);
}
}
}
}
}
static ISP_INLINE atio_private_data_t *
isp_get_atpd(ispsoftc_t *isp, tstate_t *tptr, uint32_t tag)
{
atio_private_data_t *atp;
atp = LIST_FIRST(&tptr->atfree);
if (atp) {
LIST_REMOVE(atp, next);
atp->tag = tag;
LIST_INSERT_HEAD(&tptr->atused[ATPDPHASH(tag)], atp, next);
}
return (atp);
}
static ISP_INLINE atio_private_data_t *
isp_find_atpd(ispsoftc_t *isp, tstate_t *tptr, uint32_t tag)
{
atio_private_data_t *atp;
LIST_FOREACH(atp, &tptr->atused[ATPDPHASH(tag)], next) {
if (atp->tag == tag)
return (atp);
}
return (NULL);
}
static ISP_INLINE void
isp_put_atpd(ispsoftc_t *isp, tstate_t *tptr, atio_private_data_t *atp)
{
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (atp->ests) {
isp_put_ecmd(isp, atp->ests);
}
LIST_REMOVE(atp, next);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
memset(atp, 0, sizeof (*atp));
LIST_INSERT_HEAD(&tptr->atfree, atp, next);
}
static void
isp_dump_atpd(ispsoftc_t *isp, tstate_t *tptr)
{
atio_private_data_t *atp;
const char *states[8] = { "Free", "ATIO", "CAM", "CTIO", "LAST_CTIO", "PDON", "?6", "7" };
for (atp = tptr->atpool; atp < &tptr->atpool[ATPDPSIZE]; atp++) {
xpt_print(tptr->owner, "ATP: [0x%x] origdlen %u bytes_xfrd %u lun %x nphdl 0x%04x s_id 0x%06x d_id 0x%06x oxid 0x%04x state %s\n",
atp->tag, atp->orig_datalen, atp->bytes_xfered, atp->lun, atp->nphdl, atp->sid, atp->portid, atp->oxid, states[atp->state & 0x7]);
}
}
static ISP_INLINE inot_private_data_t *
isp_get_ntpd(ispsoftc_t *isp, tstate_t *tptr)
{
inot_private_data_t *ntp;
ntp = tptr->ntfree;
if (ntp) {
tptr->ntfree = ntp->next;
}
return (ntp);
}
static ISP_INLINE inot_private_data_t *
isp_find_ntpd(ispsoftc_t *isp, tstate_t *tptr, uint32_t tag_id, uint32_t seq_id)
{
inot_private_data_t *ntp;
for (ntp = tptr->ntpool; ntp < &tptr->ntpool[ATPDPSIZE]; ntp++) {
if (ntp->rd.tag_id == tag_id && ntp->rd.seq_id == seq_id) {
return (ntp);
}
}
return (NULL);
}
static ISP_INLINE void
isp_put_ntpd(ispsoftc_t *isp, tstate_t *tptr, inot_private_data_t *ntp)
{
ntp->rd.tag_id = ntp->rd.seq_id = 0;
ntp->next = tptr->ntfree;
tptr->ntfree = ntp;
}
static cam_status
create_lun_state(ispsoftc_t *isp, int bus, struct cam_path *path, tstate_t **rslt)
{
cam_status status;
lun_id_t lun;
struct tslist *lhp;
tstate_t *tptr;
int i;
lun = xpt_path_lun_id(path);
if (lun != CAM_LUN_WILDCARD) {
if (ISP_MAX_LUNS(isp) > 0 && lun >= ISP_MAX_LUNS(isp)) {
return (CAM_LUN_INVALID);
}
}
if (is_lun_enabled(isp, bus, lun)) {
return (CAM_LUN_ALRDY_ENA);
}
tptr = malloc(sizeof (tstate_t), M_DEVBUF, M_NOWAIT|M_ZERO);
if (tptr == NULL) {
return (CAM_RESRC_UNAVAIL);
}
tptr->ts_lun = lun;
status = xpt_create_path(&tptr->owner, NULL, xpt_path_path_id(path), xpt_path_target_id(path), lun);
if (status != CAM_REQ_CMP) {
free(tptr, M_DEVBUF);
return (status);
}
SLIST_INIT(&tptr->atios);
SLIST_INIT(&tptr->inots);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
TAILQ_INIT(&tptr->waitq);
LIST_INIT(&tptr->atfree);
for (i = ATPDPSIZE-1; i >= 0; i--)
LIST_INSERT_HEAD(&tptr->atfree, &tptr->atpool[i], next);
for (i = 0; i < ATPDPHASHSIZE; i++)
LIST_INIT(&tptr->atused[i]);
for (i = 0; i < ATPDPSIZE-1; i++)
tptr->ntpool[i].next = &tptr->ntpool[i+1];
tptr->ntfree = tptr->ntpool;
tptr->hold = 1;
ISP_GET_PC_ADDR(isp, bus, lun_hash[LUN_HASH_FUNC(lun)], lhp);
SLIST_INSERT_HEAD(lhp, tptr, next);
*rslt = tptr;
ISP_PATH_PRT(isp, ISP_LOGTDEBUG0, path, "created tstate\n");
return (CAM_REQ_CMP);
}
static ISP_INLINE void
destroy_lun_state(ispsoftc_t *isp, tstate_t *tptr)
{
union ccb *ccb;
struct tslist *lhp;
KASSERT((tptr->hold != 0), ("tptr is not held"));
KASSERT((tptr->hold == 1), ("tptr still held (%d)", tptr->hold));
do {
ccb = (union ccb *)SLIST_FIRST(&tptr->atios);
if (ccb) {
SLIST_REMOVE_HEAD(&tptr->atios, sim_links.sle);
ccb->ccb_h.status = CAM_REQ_ABORTED;
xpt_done(ccb);
}
} while (ccb);
do {
ccb = (union ccb *)SLIST_FIRST(&tptr->inots);
if (ccb) {
SLIST_REMOVE_HEAD(&tptr->inots, sim_links.sle);
ccb->ccb_h.status = CAM_REQ_ABORTED;
xpt_done(ccb);
}
} while (ccb);
ISP_GET_PC_ADDR(isp, cam_sim_bus(xpt_path_sim(tptr->owner)), lun_hash[LUN_HASH_FUNC(tptr->ts_lun)], lhp);
SLIST_REMOVE(lhp, tptr, tstate, next);
ISP_PATH_PRT(isp, ISP_LOGTDEBUG0, tptr->owner, "destroyed tstate\n");
xpt_free_path(tptr->owner);
free(tptr, M_DEVBUF);
}
static void
isp_enable_lun(ispsoftc_t *isp, union ccb *ccb)
{
tstate_t *tptr;
int bus;
target_id_t target;
lun_id_t lun;
if (!IS_FC(isp) || !ISP_CAP_TMODE(isp) || !ISP_CAP_SCCFW(isp)) {
xpt_print(ccb->ccb_h.path, "Target mode is not supported\n");
ccb->ccb_h.status = CAM_FUNC_NOTAVAIL;
xpt_done(ccb);
return;
}
/*
* We only support either target and lun both wildcard
* or target and lun both non-wildcard.
*/
bus = XS_CHANNEL(ccb);
target = ccb->ccb_h.target_id;
lun = ccb->ccb_h.target_lun;
ISP_PATH_PRT(isp, ISP_LOGTDEBUG0|ISP_LOGCONFIG, ccb->ccb_h.path,
"enabling lun %jx\n", (uintmax_t)lun);
if ((target == CAM_TARGET_WILDCARD) != (lun == CAM_LUN_WILDCARD)) {
ccb->ccb_h.status = CAM_LUN_INVALID;
xpt_done(ccb);
return;
}
/* Create the state pointer. It should not already exist. */
tptr = get_lun_statep(isp, bus, lun);
if (tptr) {
ccb->ccb_h.status = CAM_LUN_ALRDY_ENA;
xpt_done(ccb);
return;
}
ccb->ccb_h.status = create_lun_state(isp, bus, ccb->ccb_h.path, &tptr);
if (ccb->ccb_h.status != CAM_REQ_CMP) {
xpt_done(ccb);
return;
}
rls_lun_statep(isp, tptr);
ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
}
static void
isp_disable_lun(ispsoftc_t *isp, union ccb *ccb)
{
tstate_t *tptr = NULL;
int bus;
target_id_t target;
lun_id_t lun;
bus = XS_CHANNEL(ccb);
target = ccb->ccb_h.target_id;
lun = ccb->ccb_h.target_lun;
ISP_PATH_PRT(isp, ISP_LOGTDEBUG0|ISP_LOGCONFIG, ccb->ccb_h.path,
"disabling lun %jx\n", (uintmax_t)lun);
if ((target == CAM_TARGET_WILDCARD) != (lun == CAM_LUN_WILDCARD)) {
ccb->ccb_h.status = CAM_LUN_INVALID;
xpt_done(ccb);
return;
}
/* Find the state pointer. */
if ((tptr = get_lun_statep(isp, bus, lun)) == NULL) {
ccb->ccb_h.status = CAM_PATH_INVALID;
xpt_done(ccb);
return;
}
destroy_lun_state(isp, tptr);
ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
}
static void
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_target_start_ctio(ispsoftc_t *isp, union ccb *ccb, enum Start_Ctio_How how)
{
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
int fctape, sendstatus, resid;
tstate_t *tptr;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
fcparam *fcp;
atio_private_data_t *atp;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
struct ccb_scsiio *cso;
uint32_t dmaresult, handle, xfrlen, sense_length, tmp;
uint8_t local[QENTRY_LEN];
tptr = get_lun_statep(isp, XS_CHANNEL(ccb), XS_LUN(ccb));
if (tptr == NULL) {
tptr = get_lun_statep(isp, XS_CHANNEL(ccb), CAM_LUN_WILDCARD);
if (tptr == NULL) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGERR, "%s: [0x%x] cannot find tstate pointer", __func__, ccb->csio.tag_id);
ccb->ccb_h.status = CAM_DEV_NOT_THERE;
xpt_done(ccb);
return;
}
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGTDEBUG0, "%s: ENTRY[0x%x] how %u xfrlen %u sendstatus %d sense_len %u", __func__, ccb->csio.tag_id, how, ccb->csio.dxfer_len,
(ccb->ccb_h.flags & CAM_SEND_STATUS) != 0, ((ccb->ccb_h.flags & CAM_SEND_SENSE)? ccb->csio.sense_len : 0));
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
switch (how) {
case FROM_TIMER:
case FROM_CAM:
/*
* Insert at the tail of the list, if any, waiting CTIO CCBs
*/
TAILQ_INSERT_TAIL(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
break;
case FROM_SRR:
case FROM_CTIO_DONE:
TAILQ_INSERT_HEAD(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
break;
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
while (TAILQ_FIRST(&tptr->waitq) != NULL) {
ccb = (union ccb *) TAILQ_FIRST(&tptr->waitq);
TAILQ_REMOVE(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
cso = &ccb->csio;
xfrlen = cso->dxfer_len;
if (xfrlen == 0) {
if ((ccb->ccb_h.flags & CAM_SEND_STATUS) == 0) {
ISP_PATH_PRT(isp, ISP_LOGERR, ccb->ccb_h.path, "a data transfer length of zero but no status to send is wrong\n");
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
continue;
}
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
atp = isp_find_atpd(isp, tptr, cso->tag_id);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (atp == NULL) {
isp_prt(isp, ISP_LOGERR, "%s: [0x%x] cannot find private data adjunct in %s", __func__, cso->tag_id, __func__);
isp_dump_atpd(isp, tptr);
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
xpt_done(ccb);
continue;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
/*
* Is this command a dead duck?
*/
if (atp->dead) {
isp_prt(isp, ISP_LOGERR, "%s: [0x%x] not sending a CTIO for a dead command", __func__, cso->tag_id);
ccb->ccb_h.status = CAM_REQ_ABORTED;
xpt_done(ccb);
continue;
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
/*
* Check to make sure we're still in target mode.
*/
fcp = FCPARAM(isp, XS_CHANNEL(ccb));
if ((fcp->role & ISP_ROLE_TARGET) == 0) {
isp_prt(isp, ISP_LOGERR, "%s: [0x%x] stopping sending a CTIO because we're no longer in target mode", __func__, cso->tag_id);
ccb->ccb_h.status = CAM_PROVIDE_FAIL;
xpt_done(ccb);
continue;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
/*
* We're only handling ATPD_CCB_OUTSTANDING outstanding CCB at a time (one of which
* could be split into two CTIOs to split data and status).
*/
if (atp->ctcnt >= ATPD_CCB_OUTSTANDING) {
isp_prt(isp, ISP_LOGTINFO, "[0x%x] handling only %d CCBs at a time (flags for this ccb: 0x%x)", cso->tag_id, ATPD_CCB_OUTSTANDING, ccb->ccb_h.flags);
TAILQ_INSERT_HEAD(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
break;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
/*
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
* Does the initiator expect FC-Tape style responses?
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
*/
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if ((atp->word3 & PRLI_WD3_RETRY) && fcp->fctape_enabled) {
fctape = 1;
} else {
fctape = 0;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
/*
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
* If we already did the data xfer portion of a CTIO that sends data
* and status, don't do it again and do the status portion now.
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
*/
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (atp->sendst) {
isp_prt(isp, ISP_LOGTDEBUG0, "[0x%x] now sending synthesized status orig_dl=%u xfered=%u bit=%u",
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
cso->tag_id, atp->orig_datalen, atp->bytes_xfered, atp->bytes_in_transit);
xfrlen = 0; /* we already did the data transfer */
atp->sendst = 0;
}
if (ccb->ccb_h.flags & CAM_SEND_STATUS) {
sendstatus = 1;
} else {
sendstatus = 0;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (ccb->ccb_h.flags & CAM_SEND_SENSE) {
KASSERT((sendstatus != 0), ("how can you have CAM_SEND_SENSE w/o CAM_SEND_STATUS?"));
/*
* Sense length is not the entire sense data structure size. Periph
* drivers don't seem to be setting sense_len to reflect the actual
* size. We'll peek inside to get the right amount.
*/
sense_length = cso->sense_len;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
/*
* This 'cannot' happen
*/
if (sense_length > (XCMD_SIZE - MIN_FCP_RESPONSE_SIZE)) {
sense_length = XCMD_SIZE - MIN_FCP_RESPONSE_SIZE;
}
} else {
sense_length = 0;
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
memset(local, 0, QENTRY_LEN);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
/*
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
* Check for overflow
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
*/
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
tmp = atp->bytes_xfered + atp->bytes_in_transit + xfrlen;
if (tmp > atp->orig_datalen) {
isp_prt(isp, ISP_LOGERR, "%s: [0x%x] data overflow by %u bytes", __func__, cso->tag_id, tmp - atp->orig_datalen);
ccb->ccb_h.status = CAM_DATA_RUN_ERR;
xpt_done(ccb);
continue;
}
if (IS_24XX(isp)) {
ct7_entry_t *cto = (ct7_entry_t *) local;
cto->ct_header.rqs_entry_type = RQSTYPE_CTIO7;
cto->ct_header.rqs_entry_count = 1;
cto->ct_header.rqs_seqno |= ATPD_SEQ_NOTIFY_CAM;
ATPD_SET_SEQNO(cto, atp);
cto->ct_nphdl = atp->nphdl;
cto->ct_rxid = atp->tag;
cto->ct_iid_lo = atp->portid;
cto->ct_iid_hi = atp->portid >> 16;
cto->ct_oxid = atp->oxid;
cto->ct_vpidx = ISP_GET_VPIDX(isp, XS_CHANNEL(ccb));
cto->ct_timeout = (XS_TIME(ccb) + 999) / 1000;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
cto->ct_flags = atp->tattr << CT7_TASK_ATTR_SHIFT;
/*
* Mode 1, status, no data. Only possible when we are sending status, have
2012-09-17 02:50:16 +00:00
* no data to transfer, and any sense data can fit into a ct7_entry_t.
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
*
* Mode 2, status, no data. We have to use this in the case that
* the sense data won't fit into a ct7_entry_t.
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
*
*/
if (sendstatus && xfrlen == 0) {
cto->ct_flags |= CT7_SENDSTATUS | CT7_NO_DATA;
resid = atp->orig_datalen - atp->bytes_xfered - atp->bytes_in_transit;
if (sense_length <= MAXRESPLEN_24XX) {
if (resid < 0) {
cto->ct_resid = -resid;
} else if (resid > 0) {
cto->ct_resid = resid;
}
cto->ct_flags |= CT7_FLAG_MODE1;
cto->ct_scsi_status = cso->scsi_status;
if (resid < 0) {
cto->ct_scsi_status |= (FCP_RESID_OVERFLOW << 8);
} else if (resid > 0) {
cto->ct_scsi_status |= (FCP_RESID_UNDERFLOW << 8);
}
if (fctape) {
cto->ct_flags |= CT7_CONFIRM|CT7_EXPLCT_CONF;
}
if (sense_length) {
cto->ct_scsi_status |= (FCP_SNSLEN_VALID << 8);
cto->rsp.m1.ct_resplen = cto->ct_senselen = sense_length;
memcpy(cto->rsp.m1.ct_resp, &cso->sense_data, sense_length);
}
} else {
bus_addr_t addr;
char buf[XCMD_SIZE];
fcp_rsp_iu_t *rp;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (atp->ests == NULL) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
atp->ests = isp_get_ecmd(isp);
if (atp->ests == NULL) {
TAILQ_INSERT_HEAD(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
break;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
memset(buf, 0, sizeof (buf));
rp = (fcp_rsp_iu_t *)buf;
if (fctape) {
cto->ct_flags |= CT7_CONFIRM|CT7_EXPLCT_CONF;
rp->fcp_rsp_bits |= FCP_CONF_REQ;
}
cto->ct_flags |= CT7_FLAG_MODE2;
rp->fcp_rsp_scsi_status = cso->scsi_status;
if (resid < 0) {
rp->fcp_rsp_resid = -resid;
rp->fcp_rsp_bits |= FCP_RESID_OVERFLOW;
} else if (resid > 0) {
rp->fcp_rsp_resid = resid;
rp->fcp_rsp_bits |= FCP_RESID_UNDERFLOW;
}
if (sense_length) {
rp->fcp_rsp_snslen = sense_length;
cto->ct_senselen = sense_length;
rp->fcp_rsp_bits |= FCP_SNSLEN_VALID;
isp_put_fcp_rsp_iu(isp, rp, atp->ests);
memcpy(((fcp_rsp_iu_t *)atp->ests)->fcp_rsp_extra, &cso->sense_data, sense_length);
} else {
isp_put_fcp_rsp_iu(isp, rp, atp->ests);
}
if (isp->isp_dblev & ISP_LOGTDEBUG1) {
isp_print_bytes(isp, "FCP Response Frame After Swizzling", MIN_FCP_RESPONSE_SIZE + sense_length, atp->ests);
}
addr = isp->isp_osinfo.ecmd_dma;
addr += ((((isp_ecmd_t *)atp->ests) - isp->isp_osinfo.ecmd_base) * XCMD_SIZE);
isp_prt(isp, ISP_LOGTDEBUG0, "%s: ests base %p vaddr %p ecmd_dma %jx addr %jx len %u", __func__, isp->isp_osinfo.ecmd_base, atp->ests,
(uintmax_t) isp->isp_osinfo.ecmd_dma, (uintmax_t)addr, MIN_FCP_RESPONSE_SIZE + sense_length);
cto->rsp.m2.ct_datalen = MIN_FCP_RESPONSE_SIZE + sense_length;
cto->rsp.m2.ct_fcp_rsp_iudata.ds_base = DMA_LO32(addr);
cto->rsp.m2.ct_fcp_rsp_iudata.ds_basehi = DMA_HI32(addr);
cto->rsp.m2.ct_fcp_rsp_iudata.ds_count = MIN_FCP_RESPONSE_SIZE + sense_length;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
if (sense_length) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGTDEBUG0, "%s: CTIO7[0x%x] seq %u nc %d CDB0=%x sstatus=0x%x flags=0x%x resid=%d slen %u sense: %x %x/%x/%x", __func__,
cto->ct_rxid, ATPD_GET_SEQNO(cto), ATPD_GET_NCAM(cto), atp->cdb0, cto->ct_scsi_status, cto->ct_flags, cto->ct_resid, sense_length,
cso->sense_data.error_code, cso->sense_data.sense_buf[1], cso->sense_data.sense_buf[11], cso->sense_data.sense_buf[12]);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
} else {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "%s: CTIO7[0x%x] seq %u nc %d CDB0=%x sstatus=0x%x flags=0x%x resid=%d", __func__,
cto->ct_rxid, ATPD_GET_SEQNO(cto), ATPD_GET_NCAM(cto), atp->cdb0, cto->ct_scsi_status, cto->ct_flags, cto->ct_resid);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
atp->state = ATPD_STATE_LAST_CTIO;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
/*
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
* Mode 0 data transfers, *possibly* with status.
*/
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (xfrlen != 0) {
cto->ct_flags |= CT7_FLAG_MODE0;
if ((cso->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
cto->ct_flags |= CT7_DATA_IN;
} else {
cto->ct_flags |= CT7_DATA_OUT;
}
cto->rsp.m0.reloff = atp->bytes_xfered + atp->bytes_in_transit;
cto->rsp.m0.ct_xfrlen = xfrlen;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
#ifdef DEBUG
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (ISP_FC_PC(isp, XS_CHANNEL(ccb))->inject_lost_data_frame && xfrlen > ISP_FC_PC(isp, XS_CHANNEL(ccb))->inject_lost_data_frame) {
isp_prt(isp, ISP_LOGWARN, "%s: truncating data frame with xfrlen %d to %d", __func__, xfrlen, xfrlen - (xfrlen >> 2));
ISP_FC_PC(isp, XS_CHANNEL(ccb))->inject_lost_data_frame = 0;
cto->rsp.m0.ct_xfrlen -= xfrlen >> 2;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
#endif
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (sendstatus) {
resid = atp->orig_datalen - atp->bytes_xfered - xfrlen;
if (cso->scsi_status == SCSI_STATUS_OK && resid == 0 /* && fctape == 0 */) {
cto->ct_flags |= CT7_SENDSTATUS;
atp->state = ATPD_STATE_LAST_CTIO;
if (fctape) {
cto->ct_flags |= CT7_CONFIRM|CT7_EXPLCT_CONF;
}
} else {
atp->sendst = 1; /* send status later */
cto->ct_header.rqs_seqno &= ~ATPD_SEQ_NOTIFY_CAM;
atp->state = ATPD_STATE_CTIO;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
} else {
atp->state = ATPD_STATE_CTIO;
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGTDEBUG0, "%s: CTIO7[0x%x] seq %u nc %d CDB0=%x sstatus=0x%x flags=0x%x xfrlen=%u off=%u", __func__,
cto->ct_rxid, ATPD_GET_SEQNO(cto), ATPD_GET_NCAM(cto), atp->cdb0, cto->ct_scsi_status, cto->ct_flags, xfrlen, atp->bytes_xfered);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
} else {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
ct2_entry_t *cto = (ct2_entry_t *) local;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (isp->isp_osinfo.sixtyfourbit)
cto->ct_header.rqs_entry_type = RQSTYPE_CTIO3;
else
cto->ct_header.rqs_entry_type = RQSTYPE_CTIO2;
cto->ct_header.rqs_entry_count = 1;
cto->ct_header.rqs_seqno |= ATPD_SEQ_NOTIFY_CAM;
ATPD_SET_SEQNO(cto, atp);
if (ISP_CAP_2KLOGIN(isp)) {
((ct2e_entry_t *)cto)->ct_iid = atp->nphdl;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
} else {
cto->ct_iid = atp->nphdl;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (ISP_CAP_SCCFW(isp) == 0) {
cto->ct_lun = ccb->ccb_h.target_lun;
}
}
cto->ct_timeout = (XS_TIME(ccb) + 999) / 1000;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
cto->ct_rxid = cso->tag_id;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
/*
* Mode 1, status, no data. Only possible when we are sending status, have
* no data to transfer, and the sense length can fit in the ct7_entry.
*
* Mode 2, status, no data. We have to use this in the case the response
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
* length won't fit into a ct2_entry_t.
*
* We'll fill out this structure with information as if this were a
* Mode 1. The hardware layer will create the Mode 2 FCP RSP IU as
* needed based upon this.
*/
if (sendstatus && xfrlen == 0) {
cto->ct_flags |= CT2_SENDSTATUS | CT2_NO_DATA;
resid = atp->orig_datalen - atp->bytes_xfered - atp->bytes_in_transit;
if (sense_length <= MAXRESPLEN) {
if (resid < 0) {
cto->ct_resid = -resid;
} else if (resid > 0) {
cto->ct_resid = resid;
}
cto->ct_flags |= CT2_FLAG_MODE1;
cto->rsp.m1.ct_scsi_status = cso->scsi_status;
if (resid < 0) {
cto->rsp.m1.ct_scsi_status |= CT2_DATA_OVER;
} else if (resid > 0) {
cto->rsp.m1.ct_scsi_status |= CT2_DATA_UNDER;
}
if (fctape) {
cto->ct_flags |= CT2_CONFIRM;
}
if (sense_length) {
cto->rsp.m1.ct_scsi_status |= CT2_SNSLEN_VALID;
cto->rsp.m1.ct_resplen = cto->rsp.m1.ct_senselen = sense_length;
memcpy(cto->rsp.m1.ct_resp, &cso->sense_data, sense_length);
}
} else {
bus_addr_t addr;
char buf[XCMD_SIZE];
fcp_rsp_iu_t *rp;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (atp->ests == NULL) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
atp->ests = isp_get_ecmd(isp);
if (atp->ests == NULL) {
TAILQ_INSERT_HEAD(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
break;
}
}
memset(buf, 0, sizeof (buf));
rp = (fcp_rsp_iu_t *)buf;
if (fctape) {
cto->ct_flags |= CT2_CONFIRM;
rp->fcp_rsp_bits |= FCP_CONF_REQ;
}
cto->ct_flags |= CT2_FLAG_MODE2;
rp->fcp_rsp_scsi_status = cso->scsi_status;
if (resid < 0) {
rp->fcp_rsp_resid = -resid;
rp->fcp_rsp_bits |= FCP_RESID_OVERFLOW;
} else if (resid > 0) {
rp->fcp_rsp_resid = resid;
rp->fcp_rsp_bits |= FCP_RESID_UNDERFLOW;
}
if (sense_length) {
rp->fcp_rsp_snslen = sense_length;
rp->fcp_rsp_bits |= FCP_SNSLEN_VALID;
isp_put_fcp_rsp_iu(isp, rp, atp->ests);
memcpy(((fcp_rsp_iu_t *)atp->ests)->fcp_rsp_extra, &cso->sense_data, sense_length);
} else {
isp_put_fcp_rsp_iu(isp, rp, atp->ests);
}
if (isp->isp_dblev & ISP_LOGTDEBUG1) {
isp_print_bytes(isp, "FCP Response Frame After Swizzling", MIN_FCP_RESPONSE_SIZE + sense_length, atp->ests);
}
addr = isp->isp_osinfo.ecmd_dma;
addr += ((((isp_ecmd_t *)atp->ests) - isp->isp_osinfo.ecmd_base) * XCMD_SIZE);
isp_prt(isp, ISP_LOGTDEBUG0, "%s: ests base %p vaddr %p ecmd_dma %jx addr %jx len %u", __func__, isp->isp_osinfo.ecmd_base, atp->ests,
(uintmax_t) isp->isp_osinfo.ecmd_dma, (uintmax_t)addr, MIN_FCP_RESPONSE_SIZE + sense_length);
cto->rsp.m2.ct_datalen = MIN_FCP_RESPONSE_SIZE + sense_length;
if (isp->isp_osinfo.sixtyfourbit) {
cto->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_base = DMA_LO32(addr);
cto->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_basehi = DMA_HI32(addr);
cto->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_count = MIN_FCP_RESPONSE_SIZE + sense_length;
} else {
cto->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_base = DMA_LO32(addr);
cto->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_count = MIN_FCP_RESPONSE_SIZE + sense_length;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
}
if (sense_length) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGTDEBUG0, "%s: CTIO2[0x%x] seq %u nc %d CDB0=%x sstatus=0x%x flags=0x%x resid=%d sense: %x %x/%x/%x", __func__,
cto->ct_rxid, ATPD_GET_SEQNO(cto), ATPD_GET_NCAM(cto), atp->cdb0, cso->scsi_status, cto->ct_flags, cto->ct_resid,
cso->sense_data.error_code, cso->sense_data.sense_buf[1], cso->sense_data.sense_buf[11], cso->sense_data.sense_buf[12]);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
} else {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGTDEBUG0, "%s: CTIO2[0x%x] seq %u nc %d CDB0=%x sstatus=0x%x flags=0x%x resid=%d", __func__, cto->ct_rxid,
ATPD_GET_SEQNO(cto), ATPD_GET_NCAM(cto), atp->cdb0, cso->scsi_status, cto->ct_flags, cto->ct_resid);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
atp->state = ATPD_STATE_LAST_CTIO;
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (xfrlen != 0) {
cto->ct_flags |= CT2_FLAG_MODE0;
if ((cso->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
cto->ct_flags |= CT2_DATA_IN;
} else {
cto->ct_flags |= CT2_DATA_OUT;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
cto->ct_reloff = atp->bytes_xfered + atp->bytes_in_transit;
cto->rsp.m0.ct_xfrlen = xfrlen;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (sendstatus) {
resid = atp->orig_datalen - atp->bytes_xfered - xfrlen;
if (cso->scsi_status == SCSI_STATUS_OK && resid == 0 /*&& fctape == 0*/) {
cto->ct_flags |= CT2_SENDSTATUS;
atp->state = ATPD_STATE_LAST_CTIO;
if (fctape) {
cto->ct_flags |= CT2_CONFIRM;
}
} else {
atp->sendst = 1; /* send status later */
cto->ct_header.rqs_seqno &= ~ATPD_SEQ_NOTIFY_CAM;
atp->state = ATPD_STATE_CTIO;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
} else {
atp->state = ATPD_STATE_CTIO;
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
}
isp_prt(isp, ISP_LOGTDEBUG0, "%s: CTIO2[%x] seq %u nc %d CDB0=%x scsi status %x flags %x resid %d xfrlen %u offset %u", __func__, cto->ct_rxid,
ATPD_GET_SEQNO(cto), ATPD_GET_NCAM(cto), atp->cdb0, cso->scsi_status, cto->ct_flags, cto->ct_resid, cso->dxfer_len, atp->bytes_xfered);
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (isp_get_pcmd(isp, ccb)) {
ISP_PATH_PRT(isp, ISP_LOGWARN, ccb->ccb_h.path, "out of PCMDs\n");
TAILQ_INSERT_HEAD(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
break;
}
handle = isp_allocate_handle(isp, ccb, ISP_HANDLE_TARGET);
if (handle == 0) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
ISP_PATH_PRT(isp, ISP_LOGWARN, ccb->ccb_h.path, "No XFLIST pointers for %s\n", __func__);
TAILQ_INSERT_HEAD(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
isp_free_pcmd(isp, ccb);
break;
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
atp->bytes_in_transit += xfrlen;
PISP_PCMD(ccb)->datalen = xfrlen;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
/*
* Call the dma setup routines for this entry (and any subsequent
* CTIOs) if there's data to move, and then tell the f/w it's got
* new things to play with. As with isp_start's usage of DMA setup,
* any swizzling is done in the machine dependent layer. Because
* of this, we put the request onto the queue area first in native
* format.
*/
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
if (IS_24XX(isp)) {
ct7_entry_t *cto = (ct7_entry_t *) local;
cto->ct_syshandle = handle;
} else {
ct2_entry_t *cto = (ct2_entry_t *) local;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
cto->ct_syshandle = handle;
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
dmaresult = ISP_DMASETUP(isp, cso, (ispreq_t *) local);
if (dmaresult != CMD_QUEUED) {
isp_destroy_handle(isp, handle);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_free_pcmd(isp, ccb);
if (dmaresult == CMD_EAGAIN) {
TAILQ_INSERT_HEAD(&tptr->waitq, &ccb->ccb_h, periph_links.tqe);
break;
}
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
xpt_done(ccb);
continue;
}
isp->isp_nactive++;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
ccb->ccb_h.status = CAM_REQ_INPROG | CAM_SIM_QUEUED;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (xfrlen) {
ccb->ccb_h.spriv_field0 = atp->bytes_xfered;
} else {
ccb->ccb_h.spriv_field0 = ~0;
}
atp->ctcnt++;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
atp->seqno++;
}
rls_lun_statep(isp, tptr);
}
static void
isp_refire_putback_atio(void *arg)
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
{
union ccb *ccb = arg;
ISP_ASSERT_LOCKED((ispsoftc_t *)XS_ISP(ccb));
isp_target_putback_atio(ccb);
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
static void
isp_refire_notify_ack(void *arg)
{
isp_tna_t *tp = arg;
ispsoftc_t *isp = tp->isp;
ISP_ASSERT_LOCKED(isp);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (isp_notify_ack(isp, tp->not)) {
callout_schedule(&tp->timer, 5);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
} else {
free(tp, M_DEVBUF);
}
}
static void
isp_target_putback_atio(union ccb *ccb)
{
ispsoftc_t *isp;
struct ccb_scsiio *cso;
void *qe;
at2_entry_t local, *at = &local;
isp = XS_ISP(ccb);
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
qe = isp_getrqentry(isp);
if (qe == NULL) {
xpt_print(ccb->ccb_h.path,
"%s: Request Queue Overflow\n", __func__);
callout_reset(&PISP_PCMD(ccb)->wdog, 10,
isp_refire_putback_atio, ccb);
return;
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
}
memset(qe, 0, QENTRY_LEN);
cso = &ccb->csio;
ISP_MEMZERO(at, sizeof (at2_entry_t));
at->at_header.rqs_entry_type = RQSTYPE_ATIO2;
at->at_header.rqs_entry_count = 1;
if (ISP_CAP_SCCFW(isp)) {
at->at_scclun = (uint16_t) ccb->ccb_h.target_lun;
#if __FreeBSD_version < 1000700
if (at->at_scclun >= 256)
at->at_scclun |= 0x4000;
#endif
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
} else {
at->at_lun = (uint8_t) ccb->ccb_h.target_lun;
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
}
at->at_status = CT_OK;
at->at_rxid = cso->tag_id;
at->at_iid = cso->ccb_h.target_id;
isp_put_atio2(isp, at, qe);
ISP_TDQE(isp, "isp_target_putback_atio", isp->isp_reqidx, qe);
ISP_SYNC_REQUEST(isp);
isp_complete_ctio(ccb);
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
}
static void
isp_complete_ctio(union ccb *ccb)
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
{
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if ((ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_INPROG) {
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
xpt_done(ccb);
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
}
}
static void
isp_handle_platform_atio2(ispsoftc_t *isp, at2_entry_t *aep)
{
fcparam *fcp;
lun_id_t lun;
fcportdb_t *lp;
tstate_t *tptr;
struct ccb_accept_tio *atiop;
uint16_t nphdl;
atio_private_data_t *atp;
inot_private_data_t *ntp;
/*
* The firmware status (except for the QLTM_SVALID bit)
* indicates why this ATIO was sent to us.
*
* If QLTM_SVALID is set, the firmware has recommended Sense Data.
*/
if ((aep->at_status & ~QLTM_SVALID) != AT_CDB) {
isp_prt(isp, ISP_LOGWARN, "bogus atio (0x%x) leaked to platform", aep->at_status);
isp_endcmd(isp, aep, SCSI_STATUS_BUSY, 0);
return;
}
fcp = FCPARAM(isp, 0);
if (ISP_CAP_SCCFW(isp)) {
lun = aep->at_scclun;
#if __FreeBSD_version < 1000700
lun &= 0x3fff;
#endif
} else {
lun = aep->at_lun;
}
if (ISP_CAP_2KLOGIN(isp)) {
nphdl = ((at2e_entry_t *)aep)->at_iid;
} else {
nphdl = aep->at_iid;
}
tptr = get_lun_statep(isp, 0, lun);
if (tptr == NULL) {
tptr = get_lun_statep(isp, 0, CAM_LUN_WILDCARD);
if (tptr == NULL) {
isp_prt(isp, ISP_LOGWARN, "%s: [0x%x] no state pointer for lun %jx or wildcard", __func__, aep->at_rxid, (uintmax_t)lun);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (lun == 0) {
isp_endcmd(isp, aep, SCSI_STATUS_BUSY, 0);
} else {
isp_endcmd(isp, aep, SCSI_STATUS_CHECK_COND | ECMD_SVALID | (0x5 << 12) | (0x25 << 16), 0);
}
return;
}
}
/*
* Start any commands pending resources first.
*/
if (tptr->restart_queue) {
inot_private_data_t *restart_queue = tptr->restart_queue;
tptr->restart_queue = NULL;
while (restart_queue) {
ntp = restart_queue;
restart_queue = ntp->rd.nt.nt_hba;
isp_prt(isp, ISP_LOGTDEBUG0, "%s: restarting resrc deprived %x", __func__, ((at2_entry_t *)ntp->rd.data)->at_rxid);
isp_handle_platform_atio2(isp, (at2_entry_t *) ntp->rd.data);
isp_put_ntpd(isp, tptr, ntp);
/*
* If a recursion caused the restart queue to start to fill again,
* stop and splice the new list on top of the old list and restore
* it and go to noresrc.
*/
if (tptr->restart_queue) {
ntp = tptr->restart_queue;
tptr->restart_queue = restart_queue;
while (restart_queue->rd.nt.nt_hba) {
restart_queue = restart_queue->rd.nt.nt_hba;
}
restart_queue->rd.nt.nt_hba = ntp;
goto noresrc;
}
}
}
atiop = (struct ccb_accept_tio *) SLIST_FIRST(&tptr->atios);
if (atiop == NULL) {
goto noresrc;
}
atp = isp_get_atpd(isp, tptr, aep->at_rxid);
if (atp == NULL) {
goto noresrc;
}
atp->state = ATPD_STATE_ATIO;
SLIST_REMOVE_HEAD(&tptr->atios, sim_links.sle);
tptr->atio_count--;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_prt(isp, ISP_LOGTDEBUG2, "Take FREE ATIO count now %d", tptr->atio_count);
atiop->ccb_h.target_id = fcp->isp_loopid;
atiop->ccb_h.target_lun = lun;
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
/*
* We don't get 'suggested' sense data as we do with SCSI cards.
*/
atiop->sense_len = 0;
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
/*
* If we're not in the port database, add ourselves.
*/
if (IS_2100(isp))
atiop->init_id = nphdl;
else {
if ((isp_find_pdb_by_handle(isp, 0, nphdl, &lp) == 0 ||
lp->state == FC_PORTDB_STATE_ZOMBIE)) {
uint64_t wwpn =
(((uint64_t) aep->at_wwpn[0]) << 48) |
(((uint64_t) aep->at_wwpn[1]) << 32) |
(((uint64_t) aep->at_wwpn[2]) << 16) |
(((uint64_t) aep->at_wwpn[3]) << 0);
isp_add_wwn_entry(isp, 0, wwpn, INI_NONE,
nphdl, PORT_ANY, 0);
if (fcp->isp_loopstate > LOOP_LTEST_DONE)
fcp->isp_loopstate = LOOP_LTEST_DONE;
isp_async(isp, ISPASYNC_CHANGE_NOTIFY, 0,
ISPASYNC_CHANGE_PDB, nphdl, 0x06, 0xff);
isp_find_pdb_by_handle(isp, 0, nphdl, &lp);
}
atiop->init_id = FC_PORTDB_TGT(isp, 0, lp);
}
atiop->cdb_len = ATIO2_CDBLEN;
ISP_MEMCPY(atiop->cdb_io.cdb_bytes, aep->at_cdb, ATIO2_CDBLEN);
atiop->ccb_h.status = CAM_CDB_RECVD;
atiop->tag_id = atp->tag;
switch (aep->at_taskflags & ATIO2_TC_ATTR_MASK) {
case ATIO2_TC_ATTR_SIMPLEQ:
atiop->ccb_h.flags |= CAM_TAG_ACTION_VALID;
atiop->tag_action = MSG_SIMPLE_Q_TAG;
break;
case ATIO2_TC_ATTR_HEADOFQ:
atiop->ccb_h.flags |= CAM_TAG_ACTION_VALID;
atiop->tag_action = MSG_HEAD_OF_Q_TAG;
break;
case ATIO2_TC_ATTR_ORDERED:
atiop->ccb_h.flags |= CAM_TAG_ACTION_VALID;
atiop->tag_action = MSG_ORDERED_Q_TAG;
break;
case ATIO2_TC_ATTR_ACAQ: /* ?? */
case ATIO2_TC_ATTR_UNTAGGED:
default:
atiop->tag_action = 0;
break;
}
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
atp->orig_datalen = aep->at_datalen;
atp->bytes_xfered = 0;
atp->lun = lun;
atp->nphdl = nphdl;
atp->sid = PORT_ANY;
atp->oxid = aep->at_oxid;
atp->cdb0 = aep->at_cdb[0];
atp->tattr = aep->at_taskflags & ATIO2_TC_ATTR_MASK;
atp->state = ATPD_STATE_CAM;
xpt_done((union ccb *)atiop);
isp_prt(isp, ISP_LOGTDEBUG0, "ATIO2[0x%x] CDB=0x%x lun %jx datalen %u", aep->at_rxid, atp->cdb0, (uintmax_t)lun, atp->orig_datalen);
rls_lun_statep(isp, tptr);
return;
noresrc:
ntp = isp_get_ntpd(isp, tptr);
if (ntp == NULL) {
rls_lun_statep(isp, tptr);
isp_endcmd(isp, aep, SCSI_STATUS_BUSY, 0);
return;
}
memcpy(ntp->rd.data, aep, QENTRY_LEN);
ntp->rd.nt.nt_hba = tptr->restart_queue;
tptr->restart_queue = ntp;
rls_lun_statep(isp, tptr);
}
static void
isp_handle_platform_atio7(ispsoftc_t *isp, at7_entry_t *aep)
{
int cdbxlen;
lun_id_t lun;
uint16_t chan, nphdl = NIL_HANDLE;
uint32_t did, sid;
fcportdb_t *lp;
tstate_t *tptr;
struct ccb_accept_tio *atiop;
atio_private_data_t *atp = NULL;
atio_private_data_t *oatp;
inot_private_data_t *ntp;
did = (aep->at_hdr.d_id[0] << 16) | (aep->at_hdr.d_id[1] << 8) | aep->at_hdr.d_id[2];
sid = (aep->at_hdr.s_id[0] << 16) | (aep->at_hdr.s_id[1] << 8) | aep->at_hdr.s_id[2];
#if __FreeBSD_version >= 1000700
lun = CAM_EXTLUN_BYTE_SWIZZLE(be64dec(aep->at_cmnd.fcp_cmnd_lun));
#else
lun = (aep->at_cmnd.fcp_cmnd_lun[0] & 0x3f << 8) |
aep->at_cmnd.fcp_cmnd_lun[1];
#endif
/*
* Find the N-port handle, and Virtual Port Index for this command.
*
* If we can't, we're somewhat in trouble because we can't actually respond w/o that information.
* We also, as a matter of course, need to know the WWN of the initiator too.
*/
if (ISP_CAP_MULTI_ID(isp) && isp->isp_nchan > 1) {
/*
* Find the right channel based upon D_ID
*/
isp_find_chan_by_did(isp, did, &chan);
if (chan == ISP_NOCHAN) {
NANOTIME_T now;
/*
* If we don't recognizer our own D_DID, terminate the exchange, unless we're within 2 seconds of startup
* It's a bit tricky here as we need to stash this command *somewhere*.
*/
GET_NANOTIME(&now);
if (NANOTIME_SUB(&now, &isp->isp_init_time) > 2000000000ULL) {
isp_prt(isp, ISP_LOGWARN, "%s: [RX_ID 0x%x] D_ID %x not found on any channel- dropping", __func__, aep->at_rxid, did);
isp_endcmd(isp, aep, NIL_HANDLE, ISP_NOCHAN, ECMD_TERMINATE, 0);
return;
}
tptr = get_lun_statep(isp, 0, 0);
if (tptr == NULL) {
tptr = get_lun_statep(isp, 0, CAM_LUN_WILDCARD);
if (tptr == NULL) {
isp_prt(isp, ISP_LOGWARN, "%s: [RX_ID 0x%x] D_ID %x not found on any channel and no tptr- dropping", __func__, aep->at_rxid, did);
isp_endcmd(isp, aep, NIL_HANDLE, ISP_NOCHAN, ECMD_TERMINATE, 0);
return;
}
}
isp_prt(isp, ISP_LOGWARN, "%s: [RX_ID 0x%x] D_ID %x not found on any channel- deferring", __func__, aep->at_rxid, did);
goto noresrc;
}
isp_prt(isp, ISP_LOGTDEBUG0, "%s: [RX_ID 0x%x] D_ID 0x%06x found on Chan %d for S_ID 0x%06x", __func__, aep->at_rxid, did, chan, sid);
} else {
chan = 0;
}
/*
* Find the PDB entry for this initiator
*/
if (isp_find_pdb_by_portid(isp, chan, sid, &lp) == 0) {
/*
* If we're not in the port database terminate the exchange.
*/
isp_prt(isp, ISP_LOGTINFO, "%s: [RX_ID 0x%x] D_ID 0x%06x found on Chan %d for S_ID 0x%06x wasn't in PDB already",
__func__, aep->at_rxid, did, chan, sid);
isp_dump_portdb(isp, chan);
isp_endcmd(isp, aep, NIL_HANDLE, chan, ECMD_TERMINATE, 0);
return;
}
nphdl = lp->handle;
/*
* Get the tstate pointer
*/
tptr = get_lun_statep(isp, chan, lun);
if (tptr == NULL) {
tptr = get_lun_statep(isp, chan, CAM_LUN_WILDCARD);
if (tptr == NULL) {
isp_prt(isp, ISP_LOGWARN,
"%s: [0x%x] no state pointer for lun %jx or wildcard",
__func__, aep->at_rxid, (uintmax_t)lun);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (lun == 0) {
isp_endcmd(isp, aep, nphdl, chan, SCSI_STATUS_BUSY, 0);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
} else {
isp_endcmd(isp, aep, nphdl, chan, SCSI_STATUS_CHECK_COND | ECMD_SVALID | (0x5 << 12) | (0x25 << 16), 0);
}
return;
}
}
/*
* Start any commands pending resources first.
*/
if (tptr->restart_queue) {
inot_private_data_t *restart_queue = tptr->restart_queue;
tptr->restart_queue = NULL;
while (restart_queue) {
ntp = restart_queue;
restart_queue = ntp->rd.nt.nt_hba;
isp_prt(isp, ISP_LOGTDEBUG0, "%s: restarting resrc deprived %x", __func__, ((at7_entry_t *)ntp->rd.data)->at_rxid);
isp_handle_platform_atio7(isp, (at7_entry_t *) ntp->rd.data);
isp_put_ntpd(isp, tptr, ntp);
/*
* If a recursion caused the restart queue to start to fill again,
* stop and splice the new list on top of the old list and restore
* it and go to noresrc.
*/
if (tptr->restart_queue) {
isp_prt(isp, ISP_LOGTDEBUG0, "%s: restart queue refilling", __func__);
if (restart_queue) {
ntp = tptr->restart_queue;
tptr->restart_queue = restart_queue;
while (restart_queue->rd.nt.nt_hba) {
restart_queue = restart_queue->rd.nt.nt_hba;
}
restart_queue->rd.nt.nt_hba = ntp;
}
goto noresrc;
}
}
}
/*
* If the f/w is out of resources, just send a BUSY status back.
*/
if (aep->at_rxid == AT7_NORESRC_RXID) {
rls_lun_statep(isp, tptr);
isp_endcmd(isp, aep, nphdl, chan, SCSI_BUSY, 0);
return;
}
/*
* If we're out of resources, just send a BUSY status back.
*/
atiop = (struct ccb_accept_tio *) SLIST_FIRST(&tptr->atios);
if (atiop == NULL) {
isp_prt(isp, ISP_LOGTDEBUG0, "[0x%x] out of atios", aep->at_rxid);
goto noresrc;
}
oatp = isp_find_atpd(isp, tptr, aep->at_rxid);
if (oatp) {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_prt(isp, ISP_LOGTDEBUG0, "[0x%x] tag wraparound in isp_handle_platforms_atio7 (N-Port Handle 0x%04x S_ID 0x%04x OX_ID 0x%04x) oatp state %d",
aep->at_rxid, nphdl, sid, aep->at_hdr.ox_id, oatp->state);
/*
* It's not a "no resource" condition- but we can treat it like one
*/
goto noresrc;
}
atp = isp_get_atpd(isp, tptr, aep->at_rxid);
if (atp == NULL) {
isp_prt(isp, ISP_LOGTDEBUG0, "[0x%x] out of atps", aep->at_rxid);
goto noresrc;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
atp->word3 = lp->prli_word3;
atp->state = ATPD_STATE_ATIO;
SLIST_REMOVE_HEAD(&tptr->atios, sim_links.sle);
tptr->atio_count--;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_PATH_PRT(isp, ISP_LOGTDEBUG2, atiop->ccb_h.path, "Take FREE ATIO count now %d\n", tptr->atio_count);
atiop->init_id = FC_PORTDB_TGT(isp, chan, lp);
atiop->ccb_h.target_id = FCPARAM(isp, chan)->isp_loopid;
atiop->ccb_h.target_lun = lun;
atiop->sense_len = 0;
cdbxlen = aep->at_cmnd.fcp_cmnd_alen_datadir >> FCP_CMND_ADDTL_CDBLEN_SHIFT;
if (cdbxlen) {
isp_prt(isp, ISP_LOGWARN, "additional CDBLEN ignored");
}
cdbxlen = sizeof (aep->at_cmnd.cdb_dl.sf.fcp_cmnd_cdb);
ISP_MEMCPY(atiop->cdb_io.cdb_bytes, aep->at_cmnd.cdb_dl.sf.fcp_cmnd_cdb, cdbxlen);
atiop->cdb_len = cdbxlen;
atiop->ccb_h.status = CAM_CDB_RECVD;
atiop->tag_id = atp->tag;
switch (aep->at_cmnd.fcp_cmnd_task_attribute & FCP_CMND_TASK_ATTR_MASK) {
case FCP_CMND_TASK_ATTR_SIMPLE:
atiop->ccb_h.flags |= CAM_TAG_ACTION_VALID;
atiop->tag_action = MSG_SIMPLE_Q_TAG;
break;
case FCP_CMND_TASK_ATTR_HEAD:
atiop->ccb_h.flags |= CAM_TAG_ACTION_VALID;
atiop->tag_action = MSG_HEAD_OF_Q_TAG;
break;
case FCP_CMND_TASK_ATTR_ORDERED:
atiop->ccb_h.flags |= CAM_TAG_ACTION_VALID;
atiop->tag_action = MSG_ORDERED_Q_TAG;
break;
default:
/* FALLTHROUGH */
case FCP_CMND_TASK_ATTR_ACA:
case FCP_CMND_TASK_ATTR_UNTAGGED:
atiop->tag_action = 0;
break;
}
atp->orig_datalen = aep->at_cmnd.cdb_dl.sf.fcp_cmnd_dl;
atp->bytes_xfered = 0;
atp->lun = lun;
atp->nphdl = nphdl;
atp->portid = sid;
atp->oxid = aep->at_hdr.ox_id;
atp->rxid = aep->at_hdr.rx_id;
atp->cdb0 = atiop->cdb_io.cdb_bytes[0];
atp->tattr = aep->at_cmnd.fcp_cmnd_task_attribute & FCP_CMND_TASK_ATTR_MASK;
atp->state = ATPD_STATE_CAM;
isp_prt(isp, ISP_LOGTDEBUG0, "ATIO7[0x%x] CDB=0x%x lun %jx datalen %u",
aep->at_rxid, atp->cdb0, (uintmax_t)lun, atp->orig_datalen);
xpt_done((union ccb *)atiop);
rls_lun_statep(isp, tptr);
return;
noresrc:
if (atp) {
isp_put_atpd(isp, tptr, atp);
}
ntp = isp_get_ntpd(isp, tptr);
if (ntp == NULL) {
rls_lun_statep(isp, tptr);
isp_endcmd(isp, aep, nphdl, chan, SCSI_STATUS_BUSY, 0);
return;
}
memcpy(ntp->rd.data, aep, QENTRY_LEN);
ntp->rd.nt.nt_hba = tptr->restart_queue;
tptr->restart_queue = ntp;
rls_lun_statep(isp, tptr);
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
/*
* Handle starting an SRR (sequence retransmit request)
* We get here when we've gotten the immediate notify
* and the return of all outstanding CTIOs for this
* transaction.
*/
static void
isp_handle_srr_start(ispsoftc_t *isp, tstate_t *tptr, atio_private_data_t *atp)
{
in_fcentry_24xx_t *inot;
uint32_t srr_off, ccb_off, ccb_len, ccb_end;
union ccb *ccb;
inot = (in_fcentry_24xx_t *)atp->srr;
srr_off = inot->in_srr_reloff_lo | (inot->in_srr_reloff_hi << 16);
ccb = atp->srr_ccb;
atp->srr_ccb = NULL;
atp->nsrr++;
if (ccb == NULL) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGWARN, "SRR[0x%x] null ccb", atp->tag);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
goto fail;
}
ccb_off = ccb->ccb_h.spriv_field0;
ccb_len = ccb->csio.dxfer_len;
ccb_end = (ccb_off == ~0)? ~0 : ccb_off + ccb_len;
switch (inot->in_srr_iu) {
case R_CTL_INFO_SOLICITED_DATA:
/*
* We have to restart a FCP_DATA data out transaction
*/
atp->sendst = 0;
atp->bytes_xfered = srr_off;
if (ccb_len == 0) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGWARN, "SRR[0x%x] SRR offset 0x%x but current CCB doesn't transfer data", atp->tag, srr_off);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
goto mdp;
}
if (srr_off < ccb_off || ccb_off > srr_off + ccb_len) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGWARN, "SRR[0x%x] SRR offset 0x%x not covered by current CCB data range [0x%x..0x%x]", atp->tag, srr_off, ccb_off, ccb_end);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
goto mdp;
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGWARN, "SRR[0x%x] SRR offset 0x%x covered by current CCB data range [0x%x..0x%x]", atp->tag, srr_off, ccb_off, ccb_end);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
break;
case R_CTL_INFO_COMMAND_STATUS:
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGTINFO, "SRR[0x%x] Got an FCP RSP SRR- resending status", atp->tag);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
atp->sendst = 1;
/*
* We have to restart a FCP_RSP IU transaction
*/
break;
case R_CTL_INFO_DATA_DESCRIPTOR:
/*
* We have to restart an FCP DATA in transaction
*/
isp_prt(isp, ISP_LOGWARN, "Got an FCP DATA IN SRR- dropping");
goto fail;
default:
isp_prt(isp, ISP_LOGWARN, "Got an unknown information (%x) SRR- dropping", inot->in_srr_iu);
goto fail;
}
/*
* We can't do anything until this is acked, so we might as well start it now.
* We aren't going to do the usual asynchronous ack issue because we need
* to make sure this gets on the wire first.
*/
if (isp_notify_ack(isp, inot)) {
isp_prt(isp, ISP_LOGWARN, "could not push positive ack for SRR- you lose");
goto fail;
}
isp_target_start_ctio(isp, ccb, FROM_SRR);
return;
fail:
inot->in_reserved = 1;
isp_async(isp, ISPASYNC_TARGET_NOTIFY_ACK, inot);
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status |= CAM_REQ_CMP_ERR;
isp_complete_ctio(ccb);
return;
mdp:
if (isp_notify_ack(isp, inot)) {
isp_prt(isp, ISP_LOGWARN, "could not push positive ack for SRR- you lose");
goto fail;
}
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status = CAM_MESSAGE_RECV;
/*
* This is not a strict interpretation of MDP, but it's close
*/
ccb->csio.msg_ptr = &ccb->csio.sense_data.sense_buf[SSD_FULL_SIZE - 16];
ccb->csio.msg_len = 7;
ccb->csio.msg_ptr[0] = MSG_EXTENDED;
ccb->csio.msg_ptr[1] = 5;
ccb->csio.msg_ptr[2] = 0; /* modify data pointer */
ccb->csio.msg_ptr[3] = srr_off >> 24;
ccb->csio.msg_ptr[4] = srr_off >> 16;
ccb->csio.msg_ptr[5] = srr_off >> 8;
ccb->csio.msg_ptr[6] = srr_off;
isp_complete_ctio(ccb);
}
static void
isp_handle_srr_notify(ispsoftc_t *isp, void *inot_raw)
{
tstate_t *tptr;
in_fcentry_24xx_t *inot = inot_raw;
atio_private_data_t *atp;
uint32_t tag = inot->in_rxid;
uint32_t bus = inot->in_vpidx;
if (!IS_24XX(isp)) {
isp_async(isp, ISPASYNC_TARGET_NOTIFY_ACK, inot_raw);
return;
}
tptr = get_lun_statep_from_tag(isp, bus, tag);
if (tptr == NULL) {
isp_prt(isp, ISP_LOGERR, "%s: cannot find tptr for tag %x in SRR Notify", __func__, tag);
isp_async(isp, ISPASYNC_TARGET_NOTIFY_ACK, inot);
return;
}
atp = isp_find_atpd(isp, tptr, tag);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (atp == NULL) {
rls_lun_statep(isp, tptr);
isp_prt(isp, ISP_LOGERR, "%s: cannot find adjunct for %x in SRR Notify", __func__, tag);
isp_async(isp, ISPASYNC_TARGET_NOTIFY_ACK, inot);
return;
}
atp->srr_notify_rcvd = 1;
memcpy(atp->srr, inot, sizeof (atp->srr));
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGTINFO /* ISP_LOGTDEBUG0 */, "SRR[0x%x] inot->in_rxid flags 0x%x srr_iu=%x reloff 0x%x", inot->in_rxid, inot->in_flags, inot->in_srr_iu,
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
inot->in_srr_reloff_lo | (inot->in_srr_reloff_hi << 16));
if (atp->srr_ccb)
isp_handle_srr_start(isp, tptr, atp);
rls_lun_statep(isp, tptr);
}
static void
isp_handle_platform_ctio(ispsoftc_t *isp, void *arg)
{
union ccb *ccb;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
int sentstatus = 0, ok = 0, notify_cam = 0, resid = 0, failure = 0;
tstate_t *tptr = NULL;
atio_private_data_t *atp = NULL;
int bus;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
uint32_t handle, moved_data = 0, data_requested;
handle = ((ct2_entry_t *)arg)->ct_syshandle;
ccb = isp_find_xs(isp, handle);
if (ccb == NULL) {
isp_print_bytes(isp, "null ccb in isp_handle_platform_ctio", QENTRY_LEN, arg);
return;
}
isp_destroy_handle(isp, handle);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
data_requested = PISP_PCMD(ccb)->datalen;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_free_pcmd(isp, ccb);
if (isp->isp_nactive) {
isp->isp_nactive--;
}
bus = XS_CHANNEL(ccb);
tptr = get_lun_statep(isp, bus, XS_LUN(ccb));
if (tptr == NULL) {
tptr = get_lun_statep(isp, bus, CAM_LUN_WILDCARD);
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (tptr == NULL) {
isp_prt(isp, ISP_LOGERR, "%s: cannot find tptr for tag %x after I/O", __func__, ccb->csio.tag_id);
return;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (IS_24XX(isp)) {
atp = isp_find_atpd(isp, tptr, ((ct7_entry_t *)arg)->ct_rxid);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
} else {
atp = isp_find_atpd(isp, tptr, ((ct2_entry_t *)arg)->ct_rxid);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
if (atp == NULL) {
/*
* XXX: isp_clear_commands() generates fake CTIO with zero
* ct_rxid value, filling only ct_syshandle. Workaround
* that using tag_id from the CCB, pointed by ct_syshandle.
*/
atp = isp_find_atpd(isp, tptr, ccb->csio.tag_id);
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (atp == NULL) {
rls_lun_statep(isp, tptr);
isp_prt(isp, ISP_LOGERR, "%s: cannot find adjunct for %x after I/O", __func__, ccb->csio.tag_id);
return;
}
KASSERT((atp->ctcnt > 0), ("ctio count not greater than zero"));
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
atp->bytes_in_transit -= data_requested;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
atp->ctcnt -= 1;
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
if (IS_24XX(isp)) {
ct7_entry_t *ct = arg;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (ct->ct_nphdl == CT7_SRR) {
atp->srr_ccb = ccb;
if (atp->srr_notify_rcvd)
isp_handle_srr_start(isp, tptr, atp);
rls_lun_statep(isp, tptr);
return;
}
if (ct->ct_nphdl == CT_HBA_RESET) {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
failure = CAM_UNREC_HBA_ERROR;
} else {
sentstatus = ct->ct_flags & CT7_SENDSTATUS;
ok = (ct->ct_nphdl == CT7_OK);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
notify_cam = (ct->ct_header.rqs_seqno & ATPD_SEQ_NOTIFY_CAM) != 0;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if ((ct->ct_flags & CT7_DATAMASK) != CT7_NO_DATA) {
resid = ct->ct_resid;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
moved_data = data_requested - resid;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ok? ISP_LOGTDEBUG0 : ISP_LOGWARN, "%s: CTIO7[%x] seq %u nc %d sts 0x%x flg 0x%x sns %d resid %d %s", __func__, ct->ct_rxid, ATPD_GET_SEQNO(ct),
notify_cam, ct->ct_nphdl, ct->ct_flags, (ccb->ccb_h.status & CAM_SENT_SENSE) != 0, resid, sentstatus? "FIN" : "MID");
} else {
ct2_entry_t *ct = arg;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (ct->ct_status == CT_SRR) {
atp->srr_ccb = ccb;
if (atp->srr_notify_rcvd)
isp_handle_srr_start(isp, tptr, atp);
rls_lun_statep(isp, tptr);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_target_putback_atio(ccb);
return;
}
if (ct->ct_status == CT_HBA_RESET) {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
failure = CAM_UNREC_HBA_ERROR;
} else {
sentstatus = ct->ct_flags & CT2_SENDSTATUS;
ok = (ct->ct_status & ~QLTM_SVALID) == CT_OK;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
notify_cam = (ct->ct_header.rqs_seqno & ATPD_SEQ_NOTIFY_CAM) != 0;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if ((ct->ct_flags & CT2_DATAMASK) != CT2_NO_DATA) {
resid = ct->ct_resid;
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
moved_data = data_requested - resid;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ok? ISP_LOGTDEBUG0 : ISP_LOGWARN, "%s: CTIO2[%x] seq %u nc %d sts 0x%x flg 0x%x sns %d resid %d %s", __func__, ct->ct_rxid, ATPD_GET_SEQNO(ct),
notify_cam, ct->ct_status, ct->ct_flags, (ccb->ccb_h.status & CAM_SENT_SENSE) != 0, resid, sentstatus? "FIN" : "MID");
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (ok) {
if (moved_data) {
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
atp->bytes_xfered += moved_data;
ccb->csio.resid = atp->orig_datalen - atp->bytes_xfered - atp->bytes_in_transit;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
if (sentstatus && (ccb->ccb_h.flags & CAM_SEND_SENSE)) {
ccb->ccb_h.status |= CAM_SENT_SENSE;
}
ccb->ccb_h.status |= CAM_REQ_CMP;
} else {
notify_cam = 1;
if (failure == CAM_UNREC_HBA_ERROR)
ccb->ccb_h.status |= CAM_UNREC_HBA_ERROR;
else
ccb->ccb_h.status |= CAM_REQ_CMP_ERR;
}
atp->state = ATPD_STATE_PDON;
rls_lun_statep(isp, tptr);
/*
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
* We never *not* notify CAM when there has been any error (ok == 0),
* so we never need to do an ATIO putback if we're not notifying CAM.
*/
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
isp_prt(isp, ISP_LOGTDEBUG0, "%s CTIO[0x%x] done (ok=%d nc=%d nowsendstatus=%d ccb ss=%d)",
(sentstatus)? " FINAL " : "MIDTERM ", atp->tag, ok, notify_cam, atp->sendst, (ccb->ccb_h.flags & CAM_SEND_STATUS) != 0);
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
if (notify_cam == 0) {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (atp->sendst) {
isp_target_start_ctio(isp, ccb, FROM_CTIO_DONE);
}
return;
Add a isp_target_putback_atio- we aren't using CCINCR at this time, so we need a function that tells the Qlogic f/w that a target mode command is done, so increase the resource count for that lun. Add in a timeout function to kick the putback again if we fail to do it the first time (we may not have the request queue space for ATIO push). Split the function isp_handle_platform_ctio into two parts so that the timeout function for the ATIO push or isp_handle_platform_ctio can inform CAM that the requested CTIO(s) are now done. Clean up (cough) residual handling. What we need for Fibre Channel is to preserve the at_datalen field from the original incoming ATIO so we can calculate a 'true' residual. Unfortunately, we're not guaranteed to get that back from CAM. We'll *try* to find it hiding in the periph_priv field (layering violation)- but if an ATIO was passed in from user land- forget it. This means that we'll probably get residuals wrong for Fibre Channel commands we're completing with an error. It's too late to 4.1 release to fix this- too bad. Luckily the only device we'd really care about this occurring on is a tape device and they're still so rare as FC attached devices that this can be considered an untested combination anyway. Remove all CCINCR usage (resource autoreplenish). When we've proved to ourself that things are working properly, we can add it back in. Make sure we propage 'suggested' sense data from the incoming ATIO into the created system ATIO- and set sense_len appropriately. Correctly propagate tag values. Fall back to the model of generating (well, the functions in isp_pci.c do the work) multiple CTIOs based upon what we get from XPT. Instead of being able to pair Qlogic generated ATIOs with CAM ATIOs, and then to pair CAM CTIOs with Qlogic CTIOs, we have to take the CTIO passed to us from XPT, and if it implies that we have to generate extra Qlogic CTIOs, so be it. This means that we have to wait until the last CTIO in a sequence we generated completes before calling xpt_done. Executive summary- target mode actually now pretty much works well enough to tell folks about.
2000-07-18 06:58:28 +00:00
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
/*
* We're telling CAM we're done with this CTIO transaction.
*
* 24XX cards never need an ATIO put back.
*
* Other cards need one put back only on error.
* In the latter case, a timeout will re-fire
* and try again in case we didn't have
* queue resources to do so at first. In any case,
* once the putback is done we do the completion
* call.
*/
if (ok || IS_24XX(isp)) {
isp_complete_ctio(ccb);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
} else {
isp_target_putback_atio(ccb);
}
}
static void
isp_handle_platform_notify_fc(ispsoftc_t *isp, in_fcentry_t *inp)
{
int needack = 1;
switch (inp->in_status) {
case IN_PORT_LOGOUT:
/*
* XXX: Need to delete this initiator's WWN from the database
* XXX: Need to send this LOGOUT upstream
*/
isp_prt(isp, ISP_LOGWARN, "port logout of S_ID 0x%x", inp->in_iid);
break;
case IN_PORT_CHANGED:
isp_prt(isp, ISP_LOGWARN, "port changed for S_ID 0x%x", inp->in_iid);
break;
case IN_GLOBAL_LOGO:
isp_del_all_wwn_entries(isp, 0);
isp_prt(isp, ISP_LOGINFO, "all ports logged out");
break;
case IN_ABORT_TASK:
{
tstate_t *tptr;
uint16_t nphdl, lun;
uint32_t sid;
uint64_t wwn;
atio_private_data_t *atp;
fcportdb_t *lp;
struct ccb_immediate_notify *inot = NULL;
if (ISP_CAP_SCCFW(isp)) {
lun = inp->in_scclun;
#if __FreeBSD_version < 1000700
lun &= 0x3fff;
#endif
} else {
lun = inp->in_lun;
}
if (ISP_CAP_2KLOGIN(isp)) {
nphdl = ((in_fcentry_e_t *)inp)->in_iid;
} else {
nphdl = inp->in_iid;
}
if (isp_find_pdb_by_handle(isp, 0, nphdl, &lp)) {
wwn = lp->port_wwn;
sid = lp->portid;
} else {
wwn = INI_ANY;
sid = PORT_ANY;
}
tptr = get_lun_statep(isp, 0, lun);
if (tptr == NULL) {
tptr = get_lun_statep(isp, 0, CAM_LUN_WILDCARD);
if (tptr == NULL) {
isp_prt(isp, ISP_LOGWARN, "ABORT TASK for lun %x, but no tstate", lun);
return;
}
}
atp = isp_find_atpd(isp, tptr, inp->in_seqid);
if (atp) {
inot = (struct ccb_immediate_notify *) SLIST_FIRST(&tptr->inots);
isp_prt(isp, ISP_LOGTDEBUG0, "ABORT TASK RX_ID %x WWN 0x%016llx state %d", inp->in_seqid, (unsigned long long) wwn, atp->state);
if (inot) {
tptr->inot_count--;
SLIST_REMOVE_HEAD(&tptr->inots, sim_links.sle);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_PATH_PRT(isp, ISP_LOGTDEBUG2, inot->ccb_h.path, "%s: Take FREE INOT count now %d\n", __func__, tptr->inot_count);
} else {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_PATH_PRT(isp, ISP_LOGWARN, tptr->owner, "out of INOT structures\n");
}
} else {
ISP_PATH_PRT(isp, ISP_LOGWARN, tptr->owner, "abort task RX_ID %x from wwn 0x%016llx, state unknown\n", inp->in_seqid, wwn);
}
if (inot) {
isp_notify_t tmp, *nt = &tmp;
ISP_MEMZERO(nt, sizeof (isp_notify_t));
nt->nt_hba = isp;
nt->nt_tgt = FCPARAM(isp, 0)->isp_wwpn;
nt->nt_wwn = wwn;
nt->nt_nphdl = nphdl;
nt->nt_sid = sid;
nt->nt_did = PORT_ANY;
nt->nt_lun = lun;
nt->nt_need_ack = 1;
nt->nt_channel = 0;
nt->nt_ncode = NT_ABORT_TASK;
nt->nt_lreserved = inot;
isp_handle_platform_target_tmf(isp, nt);
needack = 0;
}
rls_lun_statep(isp, tptr);
break;
}
default:
break;
}
if (needack) {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_async(isp, ISPASYNC_TARGET_NOTIFY_ACK, inp);
}
}
static void
isp_handle_platform_notify_24xx(ispsoftc_t *isp, in_fcentry_24xx_t *inot)
{
uint16_t nphdl;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
uint16_t prli_options = 0;
uint32_t portid;
fcportdb_t *lp;
char *msg = NULL;
uint8_t *ptr = (uint8_t *)inot;
uint64_t wwpn = INI_NONE, wwnn = INI_NONE;
nphdl = inot->in_nphdl;
if (nphdl != NIL_HANDLE) {
portid = inot->in_portid_hi << 16 | inot->in_portid_lo;
} else {
portid = PORT_ANY;
}
switch (inot->in_status) {
case IN24XX_ELS_RCVD:
{
char buf[16];
int chan = ISP_GET_VPIDX(isp, inot->in_vpidx);
/*
* Note that we're just getting notification that an ELS was received
* (possibly with some associated information sent upstream). This is
* *not* the same as being given the ELS frame to accept or reject.
*/
switch (inot->in_status_subcode) {
case LOGO:
msg = "LOGO";
wwpn = be64dec(&ptr[IN24XX_PLOGI_WWPN_OFF]);
isp_del_wwn_entry(isp, chan, wwpn, nphdl, portid);
break;
case PRLO:
msg = "PRLO";
break;
case PLOGI:
msg = "PLOGI";
wwnn = be64dec(&ptr[IN24XX_PLOGI_WWNN_OFF]);
wwpn = be64dec(&ptr[IN24XX_PLOGI_WWPN_OFF]);
isp_add_wwn_entry(isp, chan, wwpn, wwnn,
nphdl, portid, prli_options);
break;
case PRLI:
msg = "PRLI";
prli_options = inot->in_prli_options;
if (inot->in_flags & IN24XX_FLAG_PN_NN_VALID)
wwnn = be64dec(&ptr[IN24XX_PRLI_WWNN_OFF]);
wwpn = be64dec(&ptr[IN24XX_PRLI_WWPN_OFF]);
isp_add_wwn_entry(isp, chan, wwpn, wwnn,
nphdl, portid, prli_options);
break;
case PDISC:
msg = "PDISC";
break;
case ADISC:
msg = "ADISC";
break;
default:
ISP_SNPRINTF(buf, sizeof (buf), "ELS 0x%x", inot->in_status_subcode);
msg = buf;
break;
}
if (inot->in_flags & IN24XX_FLAG_PUREX_IOCB) {
isp_prt(isp, ISP_LOGERR, "%s Chan %d ELS N-port handle %x PortID 0x%06x marked as needing a PUREX response", msg, chan, nphdl, portid);
break;
}
isp_prt(isp, ISP_LOGTDEBUG0, "%s Chan %d ELS N-port handle %x PortID 0x%06x RX_ID 0x%x OX_ID 0x%x", msg, chan, nphdl, portid,
inot->in_rxid, inot->in_oxid);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_async(isp, ISPASYNC_TARGET_NOTIFY_ACK, inot);
break;
}
case IN24XX_PORT_LOGOUT:
msg = "PORT LOGOUT";
if (isp_find_pdb_by_handle(isp, ISP_GET_VPIDX(isp, inot->in_vpidx), nphdl, &lp)) {
isp_del_wwn_entry(isp, ISP_GET_VPIDX(isp, inot->in_vpidx), lp->port_wwn, nphdl, lp->portid);
}
/* FALLTHROUGH */
case IN24XX_PORT_CHANGED:
if (msg == NULL)
msg = "PORT CHANGED";
/* FALLTHROUGH */
case IN24XX_LIP_RESET:
if (msg == NULL)
msg = "LIP RESET";
isp_prt(isp, ISP_LOGINFO, "Chan %d %s (sub-status 0x%x) for N-port handle 0x%x", ISP_GET_VPIDX(isp, inot->in_vpidx), msg, inot->in_status_subcode, nphdl);
/*
* All subcodes here are irrelevant. What is relevant
* is that we need to terminate all active commands from
* this initiator (known by N-port handle).
*/
/* XXX IMPLEMENT XXX */
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_async(isp, ISPASYNC_TARGET_NOTIFY_ACK, inot);
break;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
case IN24XX_SRR_RCVD:
#ifdef ISP_TARGET_MODE
isp_handle_srr_notify(isp, inot);
break;
#else
if (msg == NULL)
msg = "SRR RCVD";
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
/* FALLTHROUGH */
#endif
case IN24XX_LINK_RESET:
if (msg == NULL)
msg = "LINK RESET";
case IN24XX_LINK_FAILED:
if (msg == NULL)
msg = "LINK FAILED";
default:
isp_prt(isp, ISP_LOGWARN, "Chan %d %s", ISP_GET_VPIDX(isp, inot->in_vpidx), msg);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_async(isp, ISPASYNC_TARGET_NOTIFY_ACK, inot);
break;
}
}
static int
isp_handle_platform_target_notify_ack(ispsoftc_t *isp, isp_notify_t *mp, uint32_t rsp)
{
if (isp->isp_state != ISP_RUNSTATE) {
isp_prt(isp, ISP_LOGTINFO, "Notify Code 0x%x (qevalid=%d) acked- h/w not ready (dropping)", mp->nt_ncode, mp->nt_lreserved != NULL);
return (0);
}
/*
* This case is for a Task Management Function, which shows up as an ATIO7 entry.
*/
if (IS_24XX(isp) && mp->nt_lreserved && ((isphdr_t *)mp->nt_lreserved)->rqs_entry_type == RQSTYPE_ATIO) {
ct7_entry_t local, *cto = &local;
at7_entry_t *aep = (at7_entry_t *)mp->nt_lreserved;
fcportdb_t *lp;
uint32_t sid;
uint16_t nphdl;
sid = (aep->at_hdr.s_id[0] << 16) | (aep->at_hdr.s_id[1] << 8) | aep->at_hdr.s_id[2];
if (isp_find_pdb_by_portid(isp, mp->nt_channel, sid, &lp)) {
nphdl = lp->handle;
} else {
nphdl = NIL_HANDLE;
}
ISP_MEMZERO(&local, sizeof (local));
cto->ct_header.rqs_entry_type = RQSTYPE_CTIO7;
cto->ct_header.rqs_entry_count = 1;
cto->ct_nphdl = nphdl;
cto->ct_rxid = aep->at_rxid;
cto->ct_vpidx = mp->nt_channel;
cto->ct_iid_lo = sid;
cto->ct_iid_hi = sid >> 16;
cto->ct_oxid = aep->at_hdr.ox_id;
cto->ct_flags = CT7_SENDSTATUS|CT7_NOACK|CT7_NO_DATA|CT7_FLAG_MODE1;
cto->ct_flags |= (aep->at_ta_len >> 12) << CT7_TASK_ATTR_SHIFT;
if (rsp != 0) {
cto->ct_scsi_status |= (FCP_RSPLEN_VALID << 8);
cto->rsp.m1.ct_resplen = 4;
ISP_MEMZERO(cto->rsp.m1.ct_resp, sizeof (cto->rsp.m1.ct_resp));
cto->rsp.m1.ct_resp[0] = rsp & 0xff;
cto->rsp.m1.ct_resp[1] = (rsp >> 8) & 0xff;
cto->rsp.m1.ct_resp[2] = (rsp >> 16) & 0xff;
cto->rsp.m1.ct_resp[3] = (rsp >> 24) & 0xff;
}
return (isp_target_put_entry(isp, &local));
}
/*
* This case is for a responding to an ABTS frame
*/
if (IS_24XX(isp) && mp->nt_lreserved && ((isphdr_t *)mp->nt_lreserved)->rqs_entry_type == RQSTYPE_ABTS_RCVD) {
/*
* Overload nt_need_ack here to mark whether we've terminated the associated command.
*/
if (mp->nt_need_ack) {
uint8_t storage[QENTRY_LEN];
ct7_entry_t *cto = (ct7_entry_t *) storage;
abts_t *abts = (abts_t *)mp->nt_lreserved;
ISP_MEMZERO(cto, sizeof (ct7_entry_t));
isp_prt(isp, ISP_LOGTDEBUG0, "%s: [%x] terminating after ABTS received", __func__, abts->abts_rxid_task);
cto->ct_header.rqs_entry_type = RQSTYPE_CTIO7;
cto->ct_header.rqs_entry_count = 1;
cto->ct_nphdl = mp->nt_nphdl;
cto->ct_rxid = abts->abts_rxid_task;
cto->ct_iid_lo = mp->nt_sid;
cto->ct_iid_hi = mp->nt_sid >> 16;
cto->ct_oxid = abts->abts_ox_id;
cto->ct_vpidx = mp->nt_channel;
cto->ct_flags = CT7_NOACK|CT7_TERMINATE;
if (isp_target_put_entry(isp, cto)) {
return (ENOMEM);
}
mp->nt_need_ack = 0;
}
if (isp_acknak_abts(isp, mp->nt_lreserved, 0) == ENOMEM) {
return (ENOMEM);
} else {
return (0);
}
}
/*
* Handle logout cases here
*/
if (mp->nt_ncode == NT_GLOBAL_LOGOUT) {
isp_del_all_wwn_entries(isp, mp->nt_channel);
}
if (mp->nt_ncode == NT_LOGOUT) {
if (!IS_2100(isp) && IS_FC(isp)) {
isp_del_wwn_entries(isp, mp);
}
}
/*
* General purpose acknowledgement
*/
if (mp->nt_need_ack) {
isp_prt(isp, ISP_LOGTINFO, "Notify Code 0x%x (qevalid=%d) being acked", mp->nt_ncode, mp->nt_lreserved != NULL);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
/*
* Don't need to use the guaranteed send because the caller can retry
*/
return (isp_notify_ack(isp, mp->nt_lreserved));
}
return (0);
}
/*
* Handle task management functions.
*
* We show up here with a notify structure filled out.
*
* The nt_lreserved tag points to the original queue entry
*/
static void
isp_handle_platform_target_tmf(ispsoftc_t *isp, isp_notify_t *notify)
{
tstate_t *tptr;
fcportdb_t *lp;
struct ccb_immediate_notify *inot;
inot_private_data_t *ntp = NULL;
lun_id_t lun;
isp_prt(isp, ISP_LOGTDEBUG0, "%s: code 0x%x sid 0x%x tagval 0x%016llx chan %d lun 0x%x", __func__, notify->nt_ncode,
notify->nt_sid, (unsigned long long) notify->nt_tagval, notify->nt_channel, notify->nt_lun);
/*
* NB: This assignment is necessary because of tricky type conversion.
* XXX: This is tricky and I need to check this. If the lun isn't known
* XXX: for the task management function, it does not of necessity follow
* XXX: that it should go up stream to the wildcard listener.
*/
if (notify->nt_lun == LUN_ANY) {
lun = CAM_LUN_WILDCARD;
} else {
lun = notify->nt_lun;
}
tptr = get_lun_statep(isp, notify->nt_channel, lun);
if (tptr == NULL) {
tptr = get_lun_statep(isp, notify->nt_channel, CAM_LUN_WILDCARD);
if (tptr == NULL) {
isp_prt(isp, ISP_LOGWARN, "%s: no state pointer found for chan %d lun %#jx", __func__, notify->nt_channel, (uintmax_t)lun);
goto bad;
}
}
inot = (struct ccb_immediate_notify *) SLIST_FIRST(&tptr->inots);
if (inot == NULL) {
isp_prt(isp, ISP_LOGWARN, "%s: out of immediate notify structures for chan %d lun %#jx", __func__, notify->nt_channel, (uintmax_t)lun);
goto bad;
}
if (isp_find_pdb_by_portid(isp, notify->nt_channel, notify->nt_sid, &lp) == 0 &&
isp_find_pdb_by_handle(isp, notify->nt_channel, notify->nt_nphdl, &lp) == 0) {
inot->initiator_id = CAM_TARGET_WILDCARD;
} else {
inot->initiator_id = FC_PORTDB_TGT(isp, notify->nt_channel, lp);
}
inot->seq_id = notify->nt_tagval;
inot->tag_id = notify->nt_tagval >> 32;
switch (notify->nt_ncode) {
case NT_ABORT_TASK:
isp_target_mark_aborted_early(isp, tptr, inot->tag_id);
inot->arg = MSG_ABORT_TASK;
break;
case NT_ABORT_TASK_SET:
isp_target_mark_aborted_early(isp, tptr, TAG_ANY);
inot->arg = MSG_ABORT_TASK_SET;
break;
case NT_CLEAR_ACA:
inot->arg = MSG_CLEAR_ACA;
break;
case NT_CLEAR_TASK_SET:
inot->arg = MSG_CLEAR_TASK_SET;
break;
case NT_LUN_RESET:
inot->arg = MSG_LOGICAL_UNIT_RESET;
break;
case NT_TARGET_RESET:
inot->arg = MSG_TARGET_RESET;
break;
case NT_QUERY_TASK_SET:
inot->arg = MSG_QUERY_TASK_SET;
break;
case NT_QUERY_ASYNC_EVENT:
inot->arg = MSG_QUERY_ASYNC_EVENT;
break;
default:
isp_prt(isp, ISP_LOGWARN, "%s: unknown TMF code 0x%x for chan %d lun %#jx", __func__, notify->nt_ncode, notify->nt_channel, (uintmax_t)lun);
goto bad;
}
ntp = isp_get_ntpd(isp, tptr);
if (ntp == NULL) {
isp_prt(isp, ISP_LOGWARN, "%s: out of inotify private structures", __func__);
goto bad;
}
ISP_MEMCPY(&ntp->rd.nt, notify, sizeof (isp_notify_t));
if (notify->nt_lreserved) {
ISP_MEMCPY(&ntp->rd.data, notify->nt_lreserved, QENTRY_LEN);
ntp->rd.nt.nt_lreserved = &ntp->rd.data;
}
ntp->rd.seq_id = notify->nt_tagval;
ntp->rd.tag_id = notify->nt_tagval >> 32;
tptr->inot_count--;
SLIST_REMOVE_HEAD(&tptr->inots, sim_links.sle);
rls_lun_statep(isp, tptr);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_PATH_PRT(isp, ISP_LOGTDEBUG2, inot->ccb_h.path, "%s: Take FREE INOT count now %d\n", __func__, tptr->inot_count);
inot->ccb_h.status = CAM_MESSAGE_RECV;
xpt_done((union ccb *)inot);
return;
bad:
if (tptr) {
rls_lun_statep(isp, tptr);
}
if (notify->nt_need_ack && notify->nt_lreserved) {
if (((isphdr_t *)notify->nt_lreserved)->rqs_entry_type == RQSTYPE_ABTS_RCVD) {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (isp_acknak_abts(isp, notify->nt_lreserved, ENOMEM)) {
isp_prt(isp, ISP_LOGWARN, "you lose- unable to send an ACKNAK");
}
} else {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_async(isp, ISPASYNC_TARGET_NOTIFY_ACK, notify->nt_lreserved);
}
}
}
/*
* Find the associated private data and mark it as dead so
* we don't try to work on it any further.
*/
static void
isp_target_mark_aborted(ispsoftc_t *isp, union ccb *ccb)
{
tstate_t *tptr;
atio_private_data_t *atp;
union ccb *accb = ccb->cab.abort_ccb;
tptr = get_lun_statep(isp, XS_CHANNEL(accb), XS_LUN(accb));
if (tptr == NULL) {
tptr = get_lun_statep(isp, XS_CHANNEL(accb), CAM_LUN_WILDCARD);
if (tptr == NULL) {
ccb->ccb_h.status = CAM_REQ_INVALID;
return;
}
}
atp = isp_find_atpd(isp, tptr, accb->atio.tag_id);
if (atp == NULL) {
ccb->ccb_h.status = CAM_REQ_INVALID;
} else {
atp->dead = 1;
ccb->ccb_h.status = CAM_REQ_CMP;
}
rls_lun_statep(isp, tptr);
}
static void
isp_target_mark_aborted_early(ispsoftc_t *isp, tstate_t *tptr, uint32_t tag_id)
{
atio_private_data_t *atp;
inot_private_data_t *restart_queue = tptr->restart_queue;
/*
* First, clean any commands pending restart
*/
tptr->restart_queue = NULL;
while (restart_queue) {
uint32_t this_tag_id;
inot_private_data_t *ntp = restart_queue;
restart_queue = ntp->rd.nt.nt_hba;
if (IS_24XX(isp)) {
this_tag_id = ((at7_entry_t *)ntp->rd.data)->at_rxid;
} else {
this_tag_id = ((at2_entry_t *)ntp->rd.data)->at_rxid;
}
if ((uint64_t)tag_id == TAG_ANY || tag_id == this_tag_id) {
isp_put_ntpd(isp, tptr, ntp);
} else {
ntp->rd.nt.nt_hba = tptr->restart_queue;
tptr->restart_queue = ntp;
}
}
/*
* Now mark other ones dead as well.
*/
for (atp = tptr->atpool; atp < &tptr->atpool[ATPDPSIZE]; atp++) {
if ((uint64_t)tag_id == TAG_ANY || atp->tag == tag_id) {
atp->dead = 1;
}
}
}
#endif
static void
isp_cam_async(void *cbarg, uint32_t code, struct cam_path *path, void *arg)
{
struct cam_sim *sim;
int bus, tgt;
ispsoftc_t *isp;
sim = (struct cam_sim *)cbarg;
isp = (ispsoftc_t *) cam_sim_softc(sim);
bus = cam_sim_bus(sim);
tgt = xpt_path_target_id(path);
switch (code) {
case AC_LOST_DEVICE:
if (IS_SCSI(isp)) {
uint16_t oflags, nflags;
sdparam *sdp = SDPARAM(isp, bus);
if (tgt >= 0) {
nflags = sdp->isp_devparam[tgt].nvrm_flags;
nflags &= DPARM_SAFE_DFLT;
if (isp->isp_loaded_fw) {
nflags |= DPARM_NARROW | DPARM_ASYNC;
}
oflags = sdp->isp_devparam[tgt].goal_flags;
sdp->isp_devparam[tgt].goal_flags = nflags;
sdp->isp_devparam[tgt].dev_update = 1;
sdp->update = 1;
(void) isp_control(isp, ISPCTL_UPDATE_PARAMS, bus);
sdp->isp_devparam[tgt].goal_flags = oflags;
}
}
break;
default:
isp_prt(isp, ISP_LOGWARN, "isp_cam_async: Code 0x%x", code);
break;
}
}
static void
isp_poll(struct cam_sim *sim)
{
ispsoftc_t *isp = cam_sim_softc(sim);
uint16_t isr, sema, info;
if (ISP_READ_ISR(isp, &isr, &sema, &info))
isp_intr(isp, isr, sema, info);
}
static void
isp_watchdog(void *arg)
{
struct ccb_scsiio *xs = arg;
ispsoftc_t *isp;
uint32_t ohandle = ISP_HANDLE_FREE, handle;
isp = XS_ISP(xs);
handle = isp_find_handle(isp, xs);
/*
* Hand crank the interrupt code just to be sure the command isn't stuck somewhere.
*/
if (handle != ISP_HANDLE_FREE) {
uint16_t isr, sema, info;
if (ISP_READ_ISR(isp, &isr, &sema, &info) != 0)
isp_intr(isp, isr, sema, info);
ohandle = handle;
handle = isp_find_handle(isp, xs);
}
if (handle != ISP_HANDLE_FREE) {
/*
* Try and make sure the command is really dead before
* we release the handle (and DMA resources) for reuse.
*
* If we are successful in aborting the command then
* we're done here because we'll get the command returned
* back separately.
*/
if (isp_control(isp, ISPCTL_ABORT_CMD, xs) == 0) {
return;
}
/*
* Note that after calling the above, the command may in
* fact have been completed.
*/
xs = isp_find_xs(isp, handle);
/*
* If the command no longer exists, then we won't
* be able to find the xs again with this handle.
*/
if (xs == NULL) {
return;
}
/*
* After this point, the command is really dead.
*/
if (XS_XFRLEN(xs)) {
ISP_DMAFREE(isp, xs, handle);
}
isp_destroy_handle(isp, handle);
isp_prt(isp, ISP_LOGERR, "%s: timeout for handle 0x%x", __func__, handle);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
xs->ccb_h.status &= ~CAM_STATUS_MASK;
xs->ccb_h.status |= CAM_CMD_TIMEOUT;
isp_prt_endcmd(isp, xs);
isp_done(xs);
} else {
if (ohandle != ISP_HANDLE_FREE) {
isp_prt(isp, ISP_LOGWARN, "%s: timeout for handle 0x%x, recovered during interrupt", __func__, ohandle);
} else {
isp_prt(isp, ISP_LOGWARN, "%s: timeout for handle already free", __func__);
}
}
}
static void
Close a race in the isp(4) driver that caused devices to disappear and not automatically come back if they were gone for a short period of time. The isp(4) driver has a 30 second gone device timer that gets activated whenever a device goes away. If the device comes back before the timer expires, we don't send a notification to CAM that it has gone away. If, however, there is a command sent to the device while it is gone and before it comes back, the isp(4) driver sends the command back with CAM_SEL_TIMEOUT status. CAM responds to the CAM_SEL_TIMEOUT status by removing the device. In the case where a device comes back within the 30 second gone device timer window, though, we weren't telling CAM the device came back. So, fix this by tracking whether we have told CAM the device is gone, and if we have, send a rescan if it comes back within the 30 second window. ispvar.h: In the fcportdb_t structure, add a new bitfield, reported_gone. This gets set whenever we return a command with CAM_SEL_TIMEOUT status on a Fibre Channel device. isp_freebsd.c: In isp_done(), if we're sending CAM_SEL_TIMEOUT for for a command sent to a FC device, set the reported_gone bit. In isp_async(), in the ISPASYNC_DEV_STAYED case, rescan the device in question if it is mapped to a target ID and has been reported gone. In isp_make_here(), take a port database entry argument, and clear the reported_gone bit when we send a rescan to CAM. In isp_make_gone(), take a port database entry as an argument, and set the reported_gone bit when we send an async event telling CAM consumers that the device is gone. Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:51:12 +00:00
isp_make_here(ispsoftc_t *isp, fcportdb_t *fcp, int chan, int tgt)
{
union ccb *ccb;
struct isp_fc *fc = ISP_FC_PC(isp, chan);
/*
* Allocate a CCB, create a wildcard path for this target and schedule a rescan.
*/
ccb = xpt_alloc_ccb_nowait();
if (ccb == NULL) {
isp_prt(isp, ISP_LOGWARN, "Chan %d unable to alloc CCB for rescan", chan);
return;
}
if (xpt_create_path(&ccb->ccb_h.path, NULL, cam_sim_path(fc->sim),
tgt, CAM_LUN_WILDCARD) != CAM_REQ_CMP) {
isp_prt(isp, ISP_LOGWARN, "unable to create path for rescan");
xpt_free_ccb(ccb);
return;
}
xpt_rescan(ccb);
}
static void
Close a race in the isp(4) driver that caused devices to disappear and not automatically come back if they were gone for a short period of time. The isp(4) driver has a 30 second gone device timer that gets activated whenever a device goes away. If the device comes back before the timer expires, we don't send a notification to CAM that it has gone away. If, however, there is a command sent to the device while it is gone and before it comes back, the isp(4) driver sends the command back with CAM_SEL_TIMEOUT status. CAM responds to the CAM_SEL_TIMEOUT status by removing the device. In the case where a device comes back within the 30 second gone device timer window, though, we weren't telling CAM the device came back. So, fix this by tracking whether we have told CAM the device is gone, and if we have, send a rescan if it comes back within the 30 second window. ispvar.h: In the fcportdb_t structure, add a new bitfield, reported_gone. This gets set whenever we return a command with CAM_SEL_TIMEOUT status on a Fibre Channel device. isp_freebsd.c: In isp_done(), if we're sending CAM_SEL_TIMEOUT for for a command sent to a FC device, set the reported_gone bit. In isp_async(), in the ISPASYNC_DEV_STAYED case, rescan the device in question if it is mapped to a target ID and has been reported gone. In isp_make_here(), take a port database entry argument, and clear the reported_gone bit when we send a rescan to CAM. In isp_make_gone(), take a port database entry as an argument, and set the reported_gone bit when we send an async event telling CAM consumers that the device is gone. Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:51:12 +00:00
isp_make_gone(ispsoftc_t *isp, fcportdb_t *fcp, int chan, int tgt)
{
struct cam_path *tp;
struct isp_fc *fc = ISP_FC_PC(isp, chan);
if (xpt_create_path(&tp, NULL, cam_sim_path(fc->sim), tgt, CAM_LUN_WILDCARD) == CAM_REQ_CMP) {
xpt_async(AC_LOST_DEVICE, tp, NULL);
xpt_free_path(tp);
}
}
/*
* Gone Device Timer Function- when we have decided that a device has gone
* away, we wait a specific period of time prior to telling the OS it has
* gone away.
*
* This timer function fires once a second and then scans the port database
* for devices that are marked dead but still have a virtual target assigned.
* We decrement a counter for that port database entry, and when it hits zero,
* we tell the OS the device has gone away.
*/
static void
isp_gdt(void *arg)
{
struct isp_fc *fc = arg;
taskqueue_enqueue(taskqueue_thread, &fc->gtask);
}
static void
isp_gdt_task(void *arg, int pending)
{
struct isp_fc *fc = arg;
ispsoftc_t *isp = fc->isp;
int chan = fc - isp->isp_osinfo.pc.fc;
fcportdb_t *lp;
struct ac_contract ac;
struct ac_device_changed *adc;
int dbidx, more_to_do = 0;
ISP_LOCK(isp);
isp_prt(isp, ISP_LOGDEBUG0, "Chan %d GDT timer expired", chan);
for (dbidx = 0; dbidx < MAX_FC_TARG; dbidx++) {
lp = &FCPARAM(isp, chan)->portdb[dbidx];
if (lp->state != FC_PORTDB_STATE_ZOMBIE) {
continue;
}
if (lp->gone_timer != 0) {
lp->gone_timer -= 1;
more_to_do++;
continue;
}
isp_prt(isp, ISP_LOGCONFIG, prom3, chan, dbidx, lp->portid, "Gone Device Timeout");
if (lp->is_target) {
lp->is_target = 0;
isp_make_gone(isp, lp, chan, dbidx);
}
if (lp->is_initiator) {
lp->is_initiator = 0;
ac.contract_number = AC_CONTRACT_DEV_CHG;
adc = (struct ac_device_changed *) ac.contract_data;
adc->wwpn = lp->port_wwn;
adc->port = lp->portid;
adc->target = dbidx;
adc->arrived = 0;
xpt_async(AC_CONTRACT, fc->path, &ac);
}
lp->state = FC_PORTDB_STATE_NIL;
}
if (fc->ready) {
if (more_to_do) {
callout_reset(&fc->gdt, hz, isp_gdt, fc);
} else {
callout_deactivate(&fc->gdt);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_prt(isp, ISP_LOG_SANCFG, "Chan %d Stopping Gone Device Timer @ %lu", chan, (unsigned long) time_uptime);
}
}
ISP_UNLOCK(isp);
}
/*
* When loop goes down we remember the time and freeze CAM command queue.
* During some time period we are trying to reprobe the loop. But if we
* fail, we tell the OS that devices have gone away and drop the freeze.
*
* We don't clear the devices out of our port database because, when loop
* come back up, we have to do some actual cleanup with the chip at that
* point (implicit PLOGO, e.g., to get the chip's port database state right).
*/
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
static void
isp_loop_changed(ispsoftc_t *isp, int chan)
{
fcparam *fcp = FCPARAM(isp, chan);
struct isp_fc *fc = ISP_FC_PC(isp, chan);
if (fc->loop_down_time)
return;
isp_prt(isp, ISP_LOG_SANCFG|ISP_LOGDEBUG0, "Chan %d Loop changed", chan);
if (fcp->role & ISP_ROLE_INITIATOR)
isp_freeze_loopdown(isp, chan);
fc->loop_dead = 0;
fc->loop_down_time = time_uptime;
wakeup(fc);
}
static void
isp_loop_up(ispsoftc_t *isp, int chan)
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
{
struct isp_fc *fc = ISP_FC_PC(isp, chan);
isp_prt(isp, ISP_LOG_SANCFG|ISP_LOGDEBUG0, "Chan %d Loop is up", chan);
fc->loop_seen_once = 1;
fc->loop_dead = 0;
fc->loop_down_time = 0;
isp_unfreeze_loopdown(isp, chan);
}
static void
isp_loop_dead(ispsoftc_t *isp, int chan)
{
fcparam *fcp = FCPARAM(isp, chan);
struct isp_fc *fc = ISP_FC_PC(isp, chan);
fcportdb_t *lp;
struct ac_contract ac;
struct ac_device_changed *adc;
int dbidx, i;
isp_prt(isp, ISP_LOG_SANCFG|ISP_LOGDEBUG0, "Chan %d Loop is dead", chan);
/*
* Notify to the OS all targets who we now consider have departed.
*/
for (dbidx = 0; dbidx < MAX_FC_TARG; dbidx++) {
lp = &fcp->portdb[dbidx];
if (lp->state == FC_PORTDB_STATE_NIL)
continue;
/*
* XXX: CLEAN UP AND COMPLETE ANY PENDING COMMANDS FIRST!
*/
for (i = 0; i < isp->isp_maxcmds; i++) {
struct ccb_scsiio *xs;
if (ISP_H2HT(isp->isp_xflist[i].handle) != ISP_HANDLE_INITIATOR) {
continue;
}
if ((xs = isp->isp_xflist[i].cmd) == NULL) {
continue;
}
if (dbidx != XS_TGT(xs)) {
continue;
}
isp_prt(isp, ISP_LOGWARN, "command handle 0x%x for %d.%d.%jx orphaned by loop down timeout",
isp->isp_xflist[i].handle, chan, XS_TGT(xs),
(uintmax_t)XS_LUN(xs));
}
isp_prt(isp, ISP_LOGCONFIG, prom3, chan, dbidx, lp->portid, "Loop Down Timeout");
if (lp->is_target) {
lp->is_target = 0;
isp_make_gone(isp, lp, chan, dbidx);
}
if (lp->is_initiator) {
lp->is_initiator = 0;
ac.contract_number = AC_CONTRACT_DEV_CHG;
adc = (struct ac_device_changed *) ac.contract_data;
adc->wwpn = lp->port_wwn;
adc->port = lp->portid;
adc->target = dbidx;
adc->arrived = 0;
xpt_async(AC_CONTRACT, fc->path, &ac);
}
}
isp_unfreeze_loopdown(isp, chan);
fc->loop_dead = 1;
fc->loop_down_time = 0;
}
static void
isp_kthread(void *arg)
{
struct isp_fc *fc = arg;
ispsoftc_t *isp = fc->isp;
int chan = fc - isp->isp_osinfo.pc.fc;
int slp = 0, d;
int lb, lim;
mtx_lock(&isp->isp_osinfo.lock);
while (isp->isp_osinfo.is_exiting == 0) {
isp_prt(isp, ISP_LOG_SANCFG|ISP_LOGDEBUG0,
"Chan %d Checking FC state", chan);
lb = isp_fc_runstate(isp, chan, 250000);
isp_prt(isp, ISP_LOG_SANCFG|ISP_LOGDEBUG0,
"Chan %d FC got to %s state", chan,
isp_fc_loop_statename(lb));
/*
* Our action is different based upon whether we're supporting
* Initiator mode or not. If we are, we might freeze the simq
* when loop is down and set all sorts of different delays to
* check again.
*
* If not, we simply just wait for loop to come up.
*/
if (lb == LOOP_READY || lb < 0) {
slp = 0;
} else {
/*
* If we've never seen loop up and we've waited longer
* than quickboot time, or we've seen loop up but we've
* waited longer than loop_down_limit, give up and go
* to sleep until loop comes up.
*/
if (fc->loop_seen_once == 0)
lim = isp_quickboot_time;
else
lim = fc->loop_down_limit;
d = time_uptime - fc->loop_down_time;
if (d >= lim)
slp = 0;
else if (d < 10)
slp = 1;
else if (d < 30)
slp = 5;
else if (d < 60)
slp = 10;
else if (d < 120)
slp = 20;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
else
slp = 30;
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
}
if (slp == 0) {
if (lb == LOOP_READY)
isp_loop_up(isp, chan);
else
isp_loop_dead(isp, chan);
}
isp_prt(isp, ISP_LOG_SANCFG|ISP_LOGDEBUG0,
"Chan %d sleep for %d seconds", chan, slp);
msleep(fc, &isp->isp_osinfo.lock, PRIBIO, "ispf", slp * hz);
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
}
fc->num_threads -= 1;
mtx_unlock(&isp->isp_osinfo.lock);
kthread_exit();
}
static void
isp_action(struct cam_sim *sim, union ccb *ccb)
{
int bus, tgt, ts, error;
ispsoftc_t *isp;
struct ccb_trans_settings *cts;
CAM_DEBUG(ccb->ccb_h.path, CAM_DEBUG_TRACE, ("isp_action\n"));
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp = (ispsoftc_t *)cam_sim_softc(sim);
mtx_assert(&isp->isp_lock, MA_OWNED);
isp_prt(isp, ISP_LOGDEBUG2, "isp_action code %x", ccb->ccb_h.func_code);
ISP_PCMD(ccb) = NULL;
switch (ccb->ccb_h.func_code) {
case XPT_SCSI_IO: /* Execute the requested I/O operation */
bus = XS_CHANNEL(ccb);
/*
* Do a couple of preliminary checks...
*/
if ((ccb->ccb_h.flags & CAM_CDB_POINTER) != 0) {
if ((ccb->ccb_h.flags & CAM_CDB_PHYS) != 0) {
ccb->ccb_h.status = CAM_REQ_INVALID;
isp_done((struct ccb_scsiio *) ccb);
break;
}
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ccb->csio.req_map = NULL;
#ifdef DIAGNOSTIC
if (ccb->ccb_h.target_id >= ISP_MAX_TARGETS(isp)) {
xpt_print(ccb->ccb_h.path, "invalid target\n");
ccb->ccb_h.status = CAM_PATH_INVALID;
} else if (ISP_MAX_LUNS(isp) > 0 &&
ccb->ccb_h.target_lun >= ISP_MAX_LUNS(isp)) {
xpt_print(ccb->ccb_h.path, "invalid lun\n");
ccb->ccb_h.status = CAM_PATH_INVALID;
}
if (ccb->ccb_h.status == CAM_PATH_INVALID) {
xpt_done(ccb);
break;
}
#endif
ccb->csio.scsi_status = SCSI_STATUS_OK;
if (isp_get_pcmd(isp, ccb)) {
isp_prt(isp, ISP_LOGWARN, "out of PCMDs");
cam_freeze_devq(ccb->ccb_h.path);
cam_release_devq(ccb->ccb_h.path, RELSIM_RELEASE_AFTER_TIMEOUT, 0, 250, 0);
ccb->ccb_h.status = CAM_REQUEUE_REQ;
xpt_done(ccb);
break;
}
error = isp_start((XS_T *) ccb);
switch (error) {
case CMD_QUEUED:
ccb->ccb_h.status |= CAM_SIM_QUEUED;
if (ccb->ccb_h.timeout == CAM_TIME_INFINITY) {
break;
}
ts = ccb->ccb_h.timeout;
if (ts == CAM_TIME_DEFAULT) {
ts = 60*1000;
}
ts = isp_mstohz(ts);
callout_reset(&PISP_PCMD(ccb)->wdog, ts, isp_watchdog, ccb);
break;
case CMD_RQLATER:
/*
* We get this result if the loop isn't ready
* or if the device in question has gone zombie.
*/
if (ISP_FC_PC(isp, bus)->loop_dead) {
isp_prt(isp, ISP_LOGDEBUG0,
"%d.%jx loop is dead",
XS_TGT(ccb), (uintmax_t)XS_LUN(ccb));
ccb->ccb_h.status = CAM_SEL_TIMEOUT;
isp_done((struct ccb_scsiio *) ccb);
break;
}
isp_prt(isp, ISP_LOGDEBUG0, "%d.%jx retry later",
XS_TGT(ccb), (uintmax_t)XS_LUN(ccb));
cam_freeze_devq(ccb->ccb_h.path);
cam_release_devq(ccb->ccb_h.path, RELSIM_RELEASE_AFTER_TIMEOUT, 0, 1000, 0);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ccb->ccb_h.status = CAM_REQUEUE_REQ;
isp_free_pcmd(isp, ccb);
xpt_done(ccb);
break;
case CMD_EAGAIN:
isp_free_pcmd(isp, ccb);
cam_freeze_devq(ccb->ccb_h.path);
cam_release_devq(ccb->ccb_h.path, RELSIM_RELEASE_AFTER_TIMEOUT, 0, 100, 0);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ccb->ccb_h.status = CAM_REQUEUE_REQ;
xpt_done(ccb);
break;
case CMD_COMPLETE:
isp_done((struct ccb_scsiio *) ccb);
break;
default:
isp_prt(isp, ISP_LOGERR, "What's this? 0x%x at %d in file %s", error, __LINE__, __FILE__);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ccb->ccb_h.status = CAM_REQUEUE_REQ;
isp_free_pcmd(isp, ccb);
xpt_done(ccb);
}
break;
#ifdef ISP_TARGET_MODE
case XPT_EN_LUN: /* Enable/Disable LUN as a target */
if (ccb->cel.enable) {
isp_enable_lun(isp, ccb);
} else {
isp_disable_lun(isp, ccb);
}
break;
case XPT_IMMED_NOTIFY:
case XPT_IMMEDIATE_NOTIFY: /* Add Immediate Notify Resource */
case XPT_ACCEPT_TARGET_IO: /* Add Accept Target IO Resource */
{
tstate_t *tptr = get_lun_statep(isp, XS_CHANNEL(ccb), ccb->ccb_h.target_lun);
if (tptr == NULL) {
tptr = get_lun_statep(isp, XS_CHANNEL(ccb), CAM_LUN_WILDCARD);
}
if (tptr == NULL) {
const char *str;
uint32_t tag;
if (ccb->ccb_h.func_code == XPT_IMMEDIATE_NOTIFY) {
str = "XPT_IMMEDIATE_NOTIFY";
tag = ccb->cin1.seq_id;
} else {
tag = ccb->atio.tag_id;
str = "XPT_ACCEPT_TARGET_IO";
}
ISP_PATH_PRT(isp, ISP_LOGWARN, ccb->ccb_h.path, "%s: [0x%x] no state pointer found for %s\n", __func__, tag, str);
dump_tstates(isp, XS_CHANNEL(ccb));
ccb->ccb_h.status = CAM_DEV_NOT_THERE;
break;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ccb->ccb_h.spriv_field0 = 0;
ccb->ccb_h.spriv_ptr1 = isp;
if (ccb->ccb_h.func_code == XPT_ACCEPT_TARGET_IO) {
if (ccb->atio.tag_id) {
atio_private_data_t *atp = isp_find_atpd(isp, tptr, ccb->atio.tag_id);
if (atp) {
isp_put_atpd(isp, tptr, atp);
}
}
tptr->atio_count++;
SLIST_INSERT_HEAD(&tptr->atios, &ccb->ccb_h, sim_links.sle);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_PATH_PRT(isp, ISP_LOGTDEBUG2, ccb->ccb_h.path, "Put FREE ATIO (tag id 0x%x), count now %d\n",
ccb->atio.tag_id, tptr->atio_count);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
ccb->atio.tag_id = 0;
} else if (ccb->ccb_h.func_code == XPT_IMMEDIATE_NOTIFY) {
if (ccb->cin1.tag_id) {
inot_private_data_t *ntp = isp_find_ntpd(isp, tptr, ccb->cin1.tag_id, ccb->cin1.seq_id);
if (ntp) {
isp_put_ntpd(isp, tptr, ntp);
}
}
tptr->inot_count++;
SLIST_INSERT_HEAD(&tptr->inots, &ccb->ccb_h, sim_links.sle);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_PATH_PRT(isp, ISP_LOGTDEBUG2, ccb->ccb_h.path, "Put FREE INOT, (seq id 0x%x) count now %d\n",
ccb->cin1.seq_id, tptr->inot_count);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
ccb->cin1.seq_id = 0;
} else if (ccb->ccb_h.func_code == XPT_IMMED_NOTIFY) {
tptr->inot_count++;
SLIST_INSERT_HEAD(&tptr->inots, &ccb->ccb_h, sim_links.sle);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_PATH_PRT(isp, ISP_LOGTDEBUG2, ccb->ccb_h.path, "Put FREE INOT, (seq id 0x%x) count now %d\n",
ccb->cin1.seq_id, tptr->inot_count);
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
ccb->cin1.seq_id = 0;
}
rls_lun_statep(isp, tptr);
ccb->ccb_h.status = CAM_REQ_INPROG;
break;
}
case XPT_NOTIFY_ACK:
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
break;
case XPT_NOTIFY_ACKNOWLEDGE: /* notify ack */
{
tstate_t *tptr;
inot_private_data_t *ntp;
/*
* XXX: Because we cannot guarantee that the path information in the notify acknowledge ccb
* XXX: matches that for the immediate notify, we have to *search* for the notify structure
*/
/*
* All the relevant path information is in the associated immediate notify
*/
ISP_PATH_PRT(isp, ISP_LOGTDEBUG0, ccb->ccb_h.path, "%s: [0x%x] NOTIFY ACKNOWLEDGE for 0x%x seen\n", __func__, ccb->cna2.tag_id, ccb->cna2.seq_id);
ntp = get_ntp_from_tagdata(isp, ccb->cna2.tag_id, ccb->cna2.seq_id, &tptr);
if (ntp == NULL) {
ISP_PATH_PRT(isp, ISP_LOGWARN, ccb->ccb_h.path, "%s: [0x%x] XPT_NOTIFY_ACKNOWLEDGE of 0x%x cannot find ntp private data\n", __func__,
ccb->cna2.tag_id, ccb->cna2.seq_id);
ccb->ccb_h.status = CAM_DEV_NOT_THERE;
xpt_done(ccb);
break;
}
if (isp_handle_platform_target_notify_ack(isp, &ntp->rd.nt,
(ccb->ccb_h.flags & CAM_SEND_STATUS) ? ccb->cna2.arg : 0)) {
rls_lun_statep(isp, tptr);
cam_freeze_devq(ccb->ccb_h.path);
cam_release_devq(ccb->ccb_h.path, RELSIM_RELEASE_AFTER_TIMEOUT, 0, 1000, 0);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status |= CAM_REQUEUE_REQ;
break;
}
isp_put_ntpd(isp, tptr, ntp);
rls_lun_statep(isp, tptr);
ccb->ccb_h.status = CAM_REQ_CMP;
ISP_PATH_PRT(isp, ISP_LOGTDEBUG0, ccb->ccb_h.path, "%s: [0x%x] calling xpt_done for tag 0x%x\n", __func__, ccb->cna2.tag_id, ccb->cna2.seq_id);
xpt_done(ccb);
break;
}
case XPT_CONT_TARGET_IO:
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_target_start_ctio(isp, ccb, FROM_CAM);
break;
#endif
case XPT_RESET_DEV: /* BDR the specified SCSI device */
bus = cam_sim_bus(xpt_path_sim(ccb->ccb_h.path));
tgt = ccb->ccb_h.target_id;
tgt |= (bus << 16);
error = isp_control(isp, ISPCTL_RESET_DEV, bus, tgt);
if (error) {
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
} else {
/*
* If we have a FC device, reset the Command
* Reference Number, because the target will expect
* that we re-start the CRN at 1 after a reset.
*/
2015-11-14 19:47:17 +00:00
if (IS_FC(isp))
isp_fcp_reset_crn(isp, bus, tgt, /*tgt_set*/ 1);
ccb->ccb_h.status = CAM_REQ_CMP;
}
xpt_done(ccb);
break;
case XPT_ABORT: /* Abort the specified CCB */
{
union ccb *accb = ccb->cab.abort_ccb;
switch (accb->ccb_h.func_code) {
#ifdef ISP_TARGET_MODE
case XPT_ACCEPT_TARGET_IO:
isp_target_mark_aborted(isp, ccb);
break;
#endif
case XPT_SCSI_IO:
error = isp_control(isp, ISPCTL_ABORT_CMD, accb);
if (error) {
ccb->ccb_h.status = CAM_UA_ABORT;
} else {
ccb->ccb_h.status = CAM_REQ_CMP;
}
break;
default:
ccb->ccb_h.status = CAM_REQ_INVALID;
break;
}
/*
* This is not a queued CCB, so the caller expects it to be
* complete when control is returned.
*/
break;
}
#define IS_CURRENT_SETTINGS(c) (c->type == CTS_TYPE_CURRENT_SETTINGS)
case XPT_SET_TRAN_SETTINGS: /* Nexus Settings */
cts = &ccb->cts;
if (!IS_CURRENT_SETTINGS(cts)) {
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
break;
}
tgt = cts->ccb_h.target_id;
bus = cam_sim_bus(xpt_path_sim(cts->ccb_h.path));
if (IS_SCSI(isp)) {
struct ccb_trans_settings_scsi *scsi = &cts->proto_specific.scsi;
struct ccb_trans_settings_spi *spi = &cts->xport_specific.spi;
sdparam *sdp = SDPARAM(isp, bus);
uint16_t *dptr;
if (spi->valid == 0 && scsi->valid == 0) {
ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
break;
}
/*
* We always update (internally) from goal_flags
* so any request to change settings just gets
* vectored to that location.
*/
dptr = &sdp->isp_devparam[tgt].goal_flags;
if ((spi->valid & CTS_SPI_VALID_DISC) != 0) {
if ((spi->flags & CTS_SPI_FLAGS_DISC_ENB) != 0)
*dptr |= DPARM_DISC;
else
*dptr &= ~DPARM_DISC;
}
if ((scsi->valid & CTS_SCSI_VALID_TQ) != 0) {
if ((scsi->flags & CTS_SCSI_FLAGS_TAG_ENB) != 0)
*dptr |= DPARM_TQING;
else
*dptr &= ~DPARM_TQING;
}
if ((spi->valid & CTS_SPI_VALID_BUS_WIDTH) != 0) {
if (spi->bus_width == MSG_EXT_WDTR_BUS_16_BIT)
*dptr |= DPARM_WIDE;
else
*dptr &= ~DPARM_WIDE;
}
/*
* XXX: FIX ME
*/
if ((spi->valid & CTS_SPI_VALID_SYNC_OFFSET) && (spi->valid & CTS_SPI_VALID_SYNC_RATE) && (spi->sync_period && spi->sync_offset)) {
*dptr |= DPARM_SYNC;
/*
* XXX: CHECK FOR LEGALITY
*/
sdp->isp_devparam[tgt].goal_period = spi->sync_period;
sdp->isp_devparam[tgt].goal_offset = spi->sync_offset;
} else {
*dptr &= ~DPARM_SYNC;
}
isp_prt(isp, ISP_LOGDEBUG0, "SET (%d.%d.%jx) to flags %x off %x per %x", bus, tgt, (uintmax_t)cts->ccb_h.target_lun, sdp->isp_devparam[tgt].goal_flags,
sdp->isp_devparam[tgt].goal_offset, sdp->isp_devparam[tgt].goal_period);
sdp->isp_devparam[tgt].dev_update = 1;
sdp->update = 1;
}
ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
break;
case XPT_GET_TRAN_SETTINGS:
cts = &ccb->cts;
tgt = cts->ccb_h.target_id;
bus = cam_sim_bus(xpt_path_sim(cts->ccb_h.path));
if (IS_FC(isp)) {
fcparam *fcp = FCPARAM(isp, bus);
struct ccb_trans_settings_scsi *scsi = &cts->proto_specific.scsi;
struct ccb_trans_settings_fc *fc = &cts->xport_specific.fc;
cts->protocol = PROTO_SCSI;
cts->protocol_version = SCSI_REV_2;
cts->transport = XPORT_FC;
cts->transport_version = 0;
scsi->valid = CTS_SCSI_VALID_TQ;
scsi->flags = CTS_SCSI_FLAGS_TAG_ENB;
fc->valid = CTS_FC_VALID_SPEED;
fc->bitrate = 100000;
fc->bitrate *= fcp->isp_gbspeed;
if (tgt < MAX_FC_TARG) {
fcportdb_t *lp = &fcp->portdb[tgt];
fc->wwnn = lp->node_wwn;
fc->wwpn = lp->port_wwn;
fc->port = lp->portid;
fc->valid |= CTS_FC_VALID_WWNN | CTS_FC_VALID_WWPN | CTS_FC_VALID_PORT;
}
} else {
struct ccb_trans_settings_scsi *scsi = &cts->proto_specific.scsi;
struct ccb_trans_settings_spi *spi = &cts->xport_specific.spi;
sdparam *sdp = SDPARAM(isp, bus);
uint16_t dval, pval, oval;
if (IS_CURRENT_SETTINGS(cts)) {
sdp->isp_devparam[tgt].dev_refresh = 1;
sdp->update = 1;
(void) isp_control(isp, ISPCTL_UPDATE_PARAMS, bus);
dval = sdp->isp_devparam[tgt].actv_flags;
oval = sdp->isp_devparam[tgt].actv_offset;
pval = sdp->isp_devparam[tgt].actv_period;
} else {
dval = sdp->isp_devparam[tgt].nvrm_flags;
oval = sdp->isp_devparam[tgt].nvrm_offset;
pval = sdp->isp_devparam[tgt].nvrm_period;
}
cts->protocol = PROTO_SCSI;
cts->protocol_version = SCSI_REV_2;
cts->transport = XPORT_SPI;
cts->transport_version = 2;
spi->valid = 0;
scsi->valid = 0;
spi->flags = 0;
scsi->flags = 0;
if (dval & DPARM_DISC) {
spi->flags |= CTS_SPI_FLAGS_DISC_ENB;
}
if ((dval & DPARM_SYNC) && oval && pval) {
spi->sync_offset = oval;
spi->sync_period = pval;
} else {
spi->sync_offset = 0;
spi->sync_period = 0;
}
spi->valid |= CTS_SPI_VALID_SYNC_OFFSET;
spi->valid |= CTS_SPI_VALID_SYNC_RATE;
spi->valid |= CTS_SPI_VALID_BUS_WIDTH;
if (dval & DPARM_WIDE) {
spi->bus_width = MSG_EXT_WDTR_BUS_16_BIT;
} else {
spi->bus_width = MSG_EXT_WDTR_BUS_8_BIT;
}
if (cts->ccb_h.target_lun != CAM_LUN_WILDCARD) {
scsi->valid = CTS_SCSI_VALID_TQ;
if (dval & DPARM_TQING) {
scsi->flags |= CTS_SCSI_FLAGS_TAG_ENB;
}
spi->valid |= CTS_SPI_VALID_DISC;
}
isp_prt(isp, ISP_LOGDEBUG0, "GET %s (%d.%d.%jx) to flags %x off %x per %x", IS_CURRENT_SETTINGS(cts)? "ACTIVE" : "NVRAM",
bus, tgt, (uintmax_t)cts->ccb_h.target_lun, dval, oval, pval);
}
ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
break;
case XPT_CALC_GEOMETRY:
cam_calc_geometry(&ccb->ccg, 1);
xpt_done(ccb);
break;
case XPT_RESET_BUS: /* Reset the specified bus */
bus = cam_sim_bus(sim);
error = isp_control(isp, ISPCTL_RESET_BUS, bus);
if (error) {
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
xpt_done(ccb);
break;
}
if (bootverbose) {
xpt_print(ccb->ccb_h.path, "reset bus on channel %d\n", bus);
}
if (IS_FC(isp)) {
xpt_async(AC_BUS_RESET, ISP_FC_PC(isp, bus)->path, 0);
} else {
xpt_async(AC_BUS_RESET, ISP_SPI_PC(isp, bus)->path, 0);
}
ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
break;
case XPT_TERM_IO: /* Terminate the I/O process */
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
break;
case XPT_SET_SIM_KNOB: /* Set SIM knobs */
{
struct ccb_sim_knob *kp = &ccb->knob;
fcparam *fcp;
if (!IS_FC(isp)) {
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
break;
}
bus = cam_sim_bus(xpt_path_sim(kp->ccb_h.path));
fcp = FCPARAM(isp, bus);
if (kp->xport_specific.fc.valid & KNOB_VALID_ADDRESS) {
fcp->isp_wwnn = ISP_FC_PC(isp, bus)->def_wwnn = kp->xport_specific.fc.wwnn;
fcp->isp_wwpn = ISP_FC_PC(isp, bus)->def_wwpn = kp->xport_specific.fc.wwpn;
isp_prt(isp, ISP_LOGALL, "Setting Channel %d wwns to 0x%jx 0x%jx", bus, fcp->isp_wwnn, fcp->isp_wwpn);
}
ccb->ccb_h.status = CAM_REQ_CMP;
if (kp->xport_specific.fc.valid & KNOB_VALID_ROLE) {
int rchange = 0;
int newrole = 0;
switch (kp->xport_specific.fc.role) {
case KNOB_ROLE_NONE:
if (fcp->role != ISP_ROLE_NONE) {
rchange = 1;
newrole = ISP_ROLE_NONE;
}
break;
case KNOB_ROLE_TARGET:
if (fcp->role != ISP_ROLE_TARGET) {
rchange = 1;
newrole = ISP_ROLE_TARGET;
}
break;
case KNOB_ROLE_INITIATOR:
if (fcp->role != ISP_ROLE_INITIATOR) {
rchange = 1;
newrole = ISP_ROLE_INITIATOR;
}
break;
case KNOB_ROLE_BOTH:
if (fcp->role != ISP_ROLE_BOTH) {
rchange = 1;
newrole = ISP_ROLE_BOTH;
}
break;
}
if (rchange) {
ISP_PATH_PRT(isp, ISP_LOGCONFIG, ccb->ccb_h.path, "changing role on from %d to %d\n", fcp->role, newrole);
if (isp_control(isp, ISPCTL_CHANGE_ROLE,
bus, newrole) != 0) {
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
xpt_done(ccb);
break;
}
}
}
xpt_done(ccb);
break;
}
case XPT_GET_SIM_KNOB_OLD: /* Get SIM knobs -- compat value */
case XPT_GET_SIM_KNOB: /* Get SIM knobs */
{
struct ccb_sim_knob *kp = &ccb->knob;
if (IS_FC(isp)) {
fcparam *fcp;
bus = cam_sim_bus(xpt_path_sim(kp->ccb_h.path));
fcp = FCPARAM(isp, bus);
kp->xport_specific.fc.wwnn = fcp->isp_wwnn;
kp->xport_specific.fc.wwpn = fcp->isp_wwpn;
switch (fcp->role) {
case ISP_ROLE_NONE:
kp->xport_specific.fc.role = KNOB_ROLE_NONE;
break;
case ISP_ROLE_TARGET:
kp->xport_specific.fc.role = KNOB_ROLE_TARGET;
break;
case ISP_ROLE_INITIATOR:
kp->xport_specific.fc.role = KNOB_ROLE_INITIATOR;
break;
case ISP_ROLE_BOTH:
kp->xport_specific.fc.role = KNOB_ROLE_BOTH;
break;
}
kp->xport_specific.fc.valid = KNOB_VALID_ADDRESS | KNOB_VALID_ROLE;
ccb->ccb_h.status = CAM_REQ_CMP;
} else {
ccb->ccb_h.status = CAM_REQ_INVALID;
}
xpt_done(ccb);
break;
}
case XPT_PATH_INQ: /* Path routing inquiry */
{
struct ccb_pathinq *cpi = &ccb->cpi;
cpi->version_num = 1;
#ifdef ISP_TARGET_MODE
if (IS_FC(isp) && ISP_CAP_TMODE(isp) && ISP_CAP_SCCFW(isp))
cpi->target_sprt = PIT_PROCESSOR | PIT_DISCONNECT | PIT_TERM_IO;
else
#endif
cpi->target_sprt = 0;
cpi->hba_eng_cnt = 0;
cpi->max_target = ISP_MAX_TARGETS(isp) - 1;
cpi->max_lun = ISP_MAX_LUNS(isp) == 0 ?
255 : ISP_MAX_LUNS(isp) - 1;
cpi->bus_id = cam_sim_bus(sim);
if (isp->isp_osinfo.sixtyfourbit)
cpi->maxio = (ISP_NSEG64_MAX - 1) * PAGE_SIZE;
else
cpi->maxio = (ISP_NSEG_MAX - 1) * PAGE_SIZE;
bus = cam_sim_bus(xpt_path_sim(cpi->ccb_h.path));
if (IS_FC(isp)) {
fcparam *fcp = FCPARAM(isp, bus);
cpi->hba_misc = PIM_NOBUSRESET | PIM_UNMAPPED;
#if __FreeBSD_version >= 1000700
cpi->hba_misc |= PIM_EXTLUNS;
#endif
#if __FreeBSD_version >= 1000039
cpi->hba_misc |= PIM_NOSCAN;
#endif
/*
* Because our loop ID can shift from time to time,
* make our initiator ID out of range of our bus.
*/
cpi->initiator_id = cpi->max_target + 1;
Add a number of interrelated CAM feature enhancements and bug fixes. NOTE: These changes will require recompilation of any userland applications, like cdrecord, xmcd, etc., that use the CAM passthrough interface. A make world is recommended. camcontrol.[c8]: - We now support two new commands, "tags" and "negotiate". - The tags commands allows users to view the number of tagged openings for a device as well as a number of other related parameters, and it allows users to set tagged openings for a device. - The negotiate command allows users to enable and disable disconnection and tagged queueing, set sync rates, offsets and bus width. Note that not all of those features are available for all controllers. Only the adv, ahc, and ncr drivers fully support all of the features at this point. Some cards do not allow the setting of sync rates, offsets and the like, and some of the drivers don't have any facilities to do so. Some drivers, like the adw driver, only support enabling or disabling sync negotiation, but do not support setting sync rates. - new description in the camcontrol man page of how to format a disk - cleanup of the camcontrol inquiry command - add support in the 'devlist' command for skipping unconfigured devices if -v was not specified on the command line. - make use of the new base_transfer_speed in the path inquiry CCB. - fix CCB bzero cases cam_xpt.c, cam_sim.[ch], cam_ccb.h: - new flags on many CCB function codes to designate whether they're non-immediate, use a user-supplied CCB, and can only be passed from userland programs via the xpt device. Use these flags in the transport layer and pass driver to categorize CCBs. - new flag in the transport layer device matching code for device nodes that indicates whether a device is unconfigured - bump the CAM version from 0x10 to 0x11 - Change the CAM ioctls to use the version as their group code, so we can force users to recompile code even when the CCB size doesn't change. - add + fill in a new value in the path inquiry CCB, base_transfer_speed. Remove a corresponding field from the cam_sim structure, and add code to every SIM to set this field to the proper value. - Fix the set transfer settings code in the transport layer. scsi_cd.c: - make some variables volatile instead of just casting them in various places - fix a race condition in the changer code - attach unless we get a "logical unit not supported" error. This should fix all of the cases where people have devices that return weird errors when they don't have media in the drive. scsi_da.c: - attach unless we get a "logical unit not supported" error scsi_pass.c: - for immediate CCBs, just malloc a CCB to send the user request in. This gets rid of the 'held' count problem in camcontrol tags. scsi_pass.h: - change the CAM ioctls to use the CAM version as their group code. adv driver: - Allow changing the sync rate and offset separately. adw driver - Allow changing the sync rate and offset separately. aha driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. ahc driver: - Allow setting offset and sync rate separately bt driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. NCR driver: - Fix the ultra/ultra 2 negotiation bug - allow setting both the sync rate and offset separately Other HBA drivers: - Put code in to set the base_transfer_speed field for XPT_GET_TRAN_SETTINGS CCBs. Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
/*
* Set base transfer capabilities for Fibre Channel, for this HBA.
Add a number of interrelated CAM feature enhancements and bug fixes. NOTE: These changes will require recompilation of any userland applications, like cdrecord, xmcd, etc., that use the CAM passthrough interface. A make world is recommended. camcontrol.[c8]: - We now support two new commands, "tags" and "negotiate". - The tags commands allows users to view the number of tagged openings for a device as well as a number of other related parameters, and it allows users to set tagged openings for a device. - The negotiate command allows users to enable and disable disconnection and tagged queueing, set sync rates, offsets and bus width. Note that not all of those features are available for all controllers. Only the adv, ahc, and ncr drivers fully support all of the features at this point. Some cards do not allow the setting of sync rates, offsets and the like, and some of the drivers don't have any facilities to do so. Some drivers, like the adw driver, only support enabling or disabling sync negotiation, but do not support setting sync rates. - new description in the camcontrol man page of how to format a disk - cleanup of the camcontrol inquiry command - add support in the 'devlist' command for skipping unconfigured devices if -v was not specified on the command line. - make use of the new base_transfer_speed in the path inquiry CCB. - fix CCB bzero cases cam_xpt.c, cam_sim.[ch], cam_ccb.h: - new flags on many CCB function codes to designate whether they're non-immediate, use a user-supplied CCB, and can only be passed from userland programs via the xpt device. Use these flags in the transport layer and pass driver to categorize CCBs. - new flag in the transport layer device matching code for device nodes that indicates whether a device is unconfigured - bump the CAM version from 0x10 to 0x11 - Change the CAM ioctls to use the version as their group code, so we can force users to recompile code even when the CCB size doesn't change. - add + fill in a new value in the path inquiry CCB, base_transfer_speed. Remove a corresponding field from the cam_sim structure, and add code to every SIM to set this field to the proper value. - Fix the set transfer settings code in the transport layer. scsi_cd.c: - make some variables volatile instead of just casting them in various places - fix a race condition in the changer code - attach unless we get a "logical unit not supported" error. This should fix all of the cases where people have devices that return weird errors when they don't have media in the drive. scsi_da.c: - attach unless we get a "logical unit not supported" error scsi_pass.c: - for immediate CCBs, just malloc a CCB to send the user request in. This gets rid of the 'held' count problem in camcontrol tags. scsi_pass.h: - change the CAM ioctls to use the CAM version as their group code. adv driver: - Allow changing the sync rate and offset separately. adw driver - Allow changing the sync rate and offset separately. aha driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. ahc driver: - Allow setting offset and sync rate separately bt driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. NCR driver: - Fix the ultra/ultra 2 negotiation bug - allow setting both the sync rate and offset separately Other HBA drivers: - Put code in to set the base_transfer_speed field for XPT_GET_TRAN_SETTINGS CCBs. Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
*/
if (IS_25XX(isp)) {
cpi->base_transfer_speed = 8000000;
} else if (IS_24XX(isp)) {
cpi->base_transfer_speed = 4000000;
} else if (IS_23XX(isp)) {
cpi->base_transfer_speed = 2000000;
} else {
cpi->base_transfer_speed = 1000000;
}
cpi->hba_inquiry = PI_TAG_ABLE;
cpi->transport = XPORT_FC;
cpi->transport_version = 0;
cpi->xport_specific.fc.wwnn = fcp->isp_wwnn;
cpi->xport_specific.fc.wwpn = fcp->isp_wwpn;
cpi->xport_specific.fc.port = fcp->isp_portid;
cpi->xport_specific.fc.bitrate = fcp->isp_gbspeed * 1000;
} else {
sdparam *sdp = SDPARAM(isp, bus);
cpi->hba_inquiry = PI_SDTR_ABLE|PI_TAG_ABLE|PI_WIDE_16;
cpi->hba_misc = PIM_UNMAPPED;
cpi->initiator_id = sdp->isp_initiator_id;
Add a number of interrelated CAM feature enhancements and bug fixes. NOTE: These changes will require recompilation of any userland applications, like cdrecord, xmcd, etc., that use the CAM passthrough interface. A make world is recommended. camcontrol.[c8]: - We now support two new commands, "tags" and "negotiate". - The tags commands allows users to view the number of tagged openings for a device as well as a number of other related parameters, and it allows users to set tagged openings for a device. - The negotiate command allows users to enable and disable disconnection and tagged queueing, set sync rates, offsets and bus width. Note that not all of those features are available for all controllers. Only the adv, ahc, and ncr drivers fully support all of the features at this point. Some cards do not allow the setting of sync rates, offsets and the like, and some of the drivers don't have any facilities to do so. Some drivers, like the adw driver, only support enabling or disabling sync negotiation, but do not support setting sync rates. - new description in the camcontrol man page of how to format a disk - cleanup of the camcontrol inquiry command - add support in the 'devlist' command for skipping unconfigured devices if -v was not specified on the command line. - make use of the new base_transfer_speed in the path inquiry CCB. - fix CCB bzero cases cam_xpt.c, cam_sim.[ch], cam_ccb.h: - new flags on many CCB function codes to designate whether they're non-immediate, use a user-supplied CCB, and can only be passed from userland programs via the xpt device. Use these flags in the transport layer and pass driver to categorize CCBs. - new flag in the transport layer device matching code for device nodes that indicates whether a device is unconfigured - bump the CAM version from 0x10 to 0x11 - Change the CAM ioctls to use the version as their group code, so we can force users to recompile code even when the CCB size doesn't change. - add + fill in a new value in the path inquiry CCB, base_transfer_speed. Remove a corresponding field from the cam_sim structure, and add code to every SIM to set this field to the proper value. - Fix the set transfer settings code in the transport layer. scsi_cd.c: - make some variables volatile instead of just casting them in various places - fix a race condition in the changer code - attach unless we get a "logical unit not supported" error. This should fix all of the cases where people have devices that return weird errors when they don't have media in the drive. scsi_da.c: - attach unless we get a "logical unit not supported" error scsi_pass.c: - for immediate CCBs, just malloc a CCB to send the user request in. This gets rid of the 'held' count problem in camcontrol tags. scsi_pass.h: - change the CAM ioctls to use the CAM version as their group code. adv driver: - Allow changing the sync rate and offset separately. adw driver - Allow changing the sync rate and offset separately. aha driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. ahc driver: - Allow setting offset and sync rate separately bt driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. NCR driver: - Fix the ultra/ultra 2 negotiation bug - allow setting both the sync rate and offset separately Other HBA drivers: - Put code in to set the base_transfer_speed field for XPT_GET_TRAN_SETTINGS CCBs. Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
cpi->base_transfer_speed = 3300;
cpi->transport = XPORT_SPI;
cpi->transport_version = 2;
}
cpi->protocol = PROTO_SCSI;
cpi->protocol_version = SCSI_REV_2;
strncpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
strncpy(cpi->hba_vid, "Qlogic", HBA_IDLEN);
strncpy(cpi->dev_name, cam_sim_name(sim), DEV_IDLEN);
cpi->unit_number = cam_sim_unit(sim);
cpi->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
break;
}
default:
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
break;
}
}
#define ISPDDB (CAM_DEBUG_INFO|CAM_DEBUG_TRACE|CAM_DEBUG_CDB)
void
isp_done(XS_T *sccb)
{
ispsoftc_t *isp = XS_ISP(sccb);
uint32_t status;
if (XS_NOERR(sccb))
XS_SETERR(sccb, CAM_REQ_CMP);
if ((sccb->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP && (sccb->scsi_status != SCSI_STATUS_OK)) {
sccb->ccb_h.status &= ~CAM_STATUS_MASK;
if ((sccb->scsi_status == SCSI_STATUS_CHECK_COND) && (sccb->ccb_h.status & CAM_AUTOSNS_VALID) == 0) {
sccb->ccb_h.status |= CAM_AUTOSENSE_FAIL;
} else {
sccb->ccb_h.status |= CAM_SCSI_STATUS_ERROR;
}
}
sccb->ccb_h.status &= ~CAM_SIM_QUEUED;
status = sccb->ccb_h.status & CAM_STATUS_MASK;
if (status != CAM_REQ_CMP) {
if (status != CAM_SEL_TIMEOUT)
isp_prt(isp, ISP_LOGDEBUG0,
"target %d lun %jx CAM status 0x%x SCSI status 0x%x",
XS_TGT(sccb), (uintmax_t)XS_LUN(sccb),
sccb->ccb_h.status, sccb->scsi_status);
Close a race in the isp(4) driver that caused devices to disappear and not automatically come back if they were gone for a short period of time. The isp(4) driver has a 30 second gone device timer that gets activated whenever a device goes away. If the device comes back before the timer expires, we don't send a notification to CAM that it has gone away. If, however, there is a command sent to the device while it is gone and before it comes back, the isp(4) driver sends the command back with CAM_SEL_TIMEOUT status. CAM responds to the CAM_SEL_TIMEOUT status by removing the device. In the case where a device comes back within the 30 second gone device timer window, though, we weren't telling CAM the device came back. So, fix this by tracking whether we have told CAM the device is gone, and if we have, send a rescan if it comes back within the 30 second window. ispvar.h: In the fcportdb_t structure, add a new bitfield, reported_gone. This gets set whenever we return a command with CAM_SEL_TIMEOUT status on a Fibre Channel device. isp_freebsd.c: In isp_done(), if we're sending CAM_SEL_TIMEOUT for for a command sent to a FC device, set the reported_gone bit. In isp_async(), in the ISPASYNC_DEV_STAYED case, rescan the device in question if it is mapped to a target ID and has been reported gone. In isp_make_here(), take a port database entry argument, and clear the reported_gone bit when we send a rescan to CAM. In isp_make_gone(), take a port database entry as an argument, and set the reported_gone bit when we send an async event telling CAM consumers that the device is gone. Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:51:12 +00:00
else if ((IS_FC(isp))
&& (XS_TGT(sccb) < MAX_FC_TARG)) {
fcparam *fcp;
fcp = FCPARAM(isp, XS_CHANNEL(sccb));
fcp->portdb[XS_TGT(sccb)].is_target = 0;
Close a race in the isp(4) driver that caused devices to disappear and not automatically come back if they were gone for a short period of time. The isp(4) driver has a 30 second gone device timer that gets activated whenever a device goes away. If the device comes back before the timer expires, we don't send a notification to CAM that it has gone away. If, however, there is a command sent to the device while it is gone and before it comes back, the isp(4) driver sends the command back with CAM_SEL_TIMEOUT status. CAM responds to the CAM_SEL_TIMEOUT status by removing the device. In the case where a device comes back within the 30 second gone device timer window, though, we weren't telling CAM the device came back. So, fix this by tracking whether we have told CAM the device is gone, and if we have, send a rescan if it comes back within the 30 second window. ispvar.h: In the fcportdb_t structure, add a new bitfield, reported_gone. This gets set whenever we return a command with CAM_SEL_TIMEOUT status on a Fibre Channel device. isp_freebsd.c: In isp_done(), if we're sending CAM_SEL_TIMEOUT for for a command sent to a FC device, set the reported_gone bit. In isp_async(), in the ISPASYNC_DEV_STAYED case, rescan the device in question if it is mapped to a target ID and has been reported gone. In isp_make_here(), take a port database entry argument, and clear the reported_gone bit when we send a rescan to CAM. In isp_make_gone(), take a port database entry as an argument, and set the reported_gone bit when we send an async event telling CAM consumers that the device is gone. Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:51:12 +00:00
}
if ((sccb->ccb_h.status & CAM_DEV_QFRZN) == 0) {
sccb->ccb_h.status |= CAM_DEV_QFRZN;
xpt_freeze_devq(sccb->ccb_h.path, 1);
}
}
if ((CAM_DEBUGGED(sccb->ccb_h.path, ISPDDB)) && (sccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
xpt_print(sccb->ccb_h.path, "cam completion status 0x%x\n", sccb->ccb_h.status);
}
if (ISP_PCMD(sccb)) {
if (callout_active(&PISP_PCMD(sccb)->wdog))
callout_stop(&PISP_PCMD(sccb)->wdog);
isp_free_pcmd(isp, (union ccb *) sccb);
}
xpt_done((union ccb *) sccb);
}
void
isp_async(ispsoftc_t *isp, ispasync_t cmd, ...)
{
int bus;
static const char prom[] = "Chan %d [%d] WWPN 0x%16jx PortID 0x%06x handle 0x%x %s %s";
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
char buf[64];
char *msg = NULL;
target_id_t tgt;
fcportdb_t *lp;
struct isp_fc *fc;
struct cam_path *tmppath;
struct ac_contract ac;
struct ac_device_changed *adc;
va_list ap;
switch (cmd) {
case ISPASYNC_NEW_TGT_PARAMS:
{
struct ccb_trans_settings_scsi *scsi;
struct ccb_trans_settings_spi *spi;
int flags, tgt;
sdparam *sdp;
struct ccb_trans_settings cts;
memset(&cts, 0, sizeof (struct ccb_trans_settings));
va_start(ap, cmd);
bus = va_arg(ap, int);
tgt = va_arg(ap, int);
va_end(ap);
sdp = SDPARAM(isp, bus);
if (xpt_create_path(&tmppath, NULL, cam_sim_path(ISP_SPI_PC(isp, bus)->sim), tgt, CAM_LUN_WILDCARD) != CAM_REQ_CMP) {
isp_prt(isp, ISP_LOGWARN, "isp_async cannot make temp path for %d.%d", tgt, bus);
break;
}
flags = sdp->isp_devparam[tgt].actv_flags;
cts.type = CTS_TYPE_CURRENT_SETTINGS;
cts.protocol = PROTO_SCSI;
cts.transport = XPORT_SPI;
scsi = &cts.proto_specific.scsi;
spi = &cts.xport_specific.spi;
if (flags & DPARM_TQING) {
scsi->valid |= CTS_SCSI_VALID_TQ;
scsi->flags |= CTS_SCSI_FLAGS_TAG_ENB;
}
if (flags & DPARM_DISC) {
spi->valid |= CTS_SPI_VALID_DISC;
spi->flags |= CTS_SPI_FLAGS_DISC_ENB;
}
spi->flags |= CTS_SPI_VALID_BUS_WIDTH;
if (flags & DPARM_WIDE) {
spi->bus_width = MSG_EXT_WDTR_BUS_16_BIT;
} else {
spi->bus_width = MSG_EXT_WDTR_BUS_8_BIT;
}
if (flags & DPARM_SYNC) {
spi->valid |= CTS_SPI_VALID_SYNC_RATE;
spi->valid |= CTS_SPI_VALID_SYNC_OFFSET;
spi->sync_period = sdp->isp_devparam[tgt].actv_period;
spi->sync_offset = sdp->isp_devparam[tgt].actv_offset;
}
isp_prt(isp, ISP_LOGDEBUG2, "NEW_TGT_PARAMS bus %d tgt %d period %x offset %x flags %x", bus, tgt, sdp->isp_devparam[tgt].actv_period, sdp->isp_devparam[tgt].actv_offset, flags);
xpt_setup_ccb(&cts.ccb_h, tmppath, 1);
xpt_async(AC_TRANSFER_NEG, tmppath, &cts);
xpt_free_path(tmppath);
break;
}
case ISPASYNC_BUS_RESET:
{
va_start(ap, cmd);
bus = va_arg(ap, int);
va_end(ap);
isp_prt(isp, ISP_LOGINFO, "SCSI bus reset on bus %d detected", bus);
if (IS_FC(isp)) {
xpt_async(AC_BUS_RESET, ISP_FC_PC(isp, bus)->path, NULL);
} else {
xpt_async(AC_BUS_RESET, ISP_SPI_PC(isp, bus)->path, NULL);
}
break;
}
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
case ISPASYNC_LIP:
if (msg == NULL)
msg = "LIP Received";
/* FALLTHROUGH */
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
case ISPASYNC_LOOP_RESET:
if (msg == NULL)
msg = "LOOP Reset";
/* FALLTHROUGH */
case ISPASYNC_LOOP_DOWN:
if (msg == NULL)
msg = "LOOP Down";
va_start(ap, cmd);
bus = va_arg(ap, int);
va_end(ap);
2015-11-14 19:47:17 +00:00
isp_fcp_reset_crn(isp, bus, /*tgt*/0, /*tgt_set*/ 0);
isp_loop_changed(isp, bus);
isp_prt(isp, ISP_LOGINFO, "Chan %d %s", bus, msg);
break;
case ISPASYNC_LOOP_UP:
va_start(ap, cmd);
bus = va_arg(ap, int);
va_end(ap);
isp_loop_changed(isp, bus);
isp_prt(isp, ISP_LOGINFO, "Chan %d Loop UP", bus);
break;
case ISPASYNC_DEV_ARRIVED:
va_start(ap, cmd);
bus = va_arg(ap, int);
lp = va_arg(ap, fcportdb_t *);
va_end(ap);
fc = ISP_FC_PC(isp, bus);
tgt = FC_PORTDB_TGT(isp, bus, lp);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_gen_role_str(buf, sizeof (buf), lp->prli_word3);
isp_prt(isp, ISP_LOGCONFIG, prom, bus, tgt, lp->port_wwn, lp->portid, lp->handle, buf, "arrived");
if ((FCPARAM(isp, bus)->role & ISP_ROLE_INITIATOR) &&
(lp->prli_word3 & PRLI_WD3_TARGET_FUNCTION)) {
lp->is_target = 1;
2015-11-14 19:47:17 +00:00
isp_fcp_reset_crn(isp, bus, tgt, /*tgt_set*/ 1);
Close a race in the isp(4) driver that caused devices to disappear and not automatically come back if they were gone for a short period of time. The isp(4) driver has a 30 second gone device timer that gets activated whenever a device goes away. If the device comes back before the timer expires, we don't send a notification to CAM that it has gone away. If, however, there is a command sent to the device while it is gone and before it comes back, the isp(4) driver sends the command back with CAM_SEL_TIMEOUT status. CAM responds to the CAM_SEL_TIMEOUT status by removing the device. In the case where a device comes back within the 30 second gone device timer window, though, we weren't telling CAM the device came back. So, fix this by tracking whether we have told CAM the device is gone, and if we have, send a rescan if it comes back within the 30 second window. ispvar.h: In the fcportdb_t structure, add a new bitfield, reported_gone. This gets set whenever we return a command with CAM_SEL_TIMEOUT status on a Fibre Channel device. isp_freebsd.c: In isp_done(), if we're sending CAM_SEL_TIMEOUT for for a command sent to a FC device, set the reported_gone bit. In isp_async(), in the ISPASYNC_DEV_STAYED case, rescan the device in question if it is mapped to a target ID and has been reported gone. In isp_make_here(), take a port database entry argument, and clear the reported_gone bit when we send a rescan to CAM. In isp_make_gone(), take a port database entry as an argument, and set the reported_gone bit when we send an async event telling CAM consumers that the device is gone. Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:51:12 +00:00
isp_make_here(isp, lp, bus, tgt);
}
if ((FCPARAM(isp, bus)->role & ISP_ROLE_TARGET) &&
(lp->prli_word3 & PRLI_WD3_INITIATOR_FUNCTION)) {
lp->is_initiator = 1;
ac.contract_number = AC_CONTRACT_DEV_CHG;
adc = (struct ac_device_changed *) ac.contract_data;
adc->wwpn = lp->port_wwn;
adc->port = lp->portid;
adc->target = tgt;
adc->arrived = 1;
xpt_async(AC_CONTRACT, fc->path, &ac);
}
break;
case ISPASYNC_DEV_CHANGED:
va_start(ap, cmd);
bus = va_arg(ap, int);
lp = va_arg(ap, fcportdb_t *);
va_end(ap);
fc = ISP_FC_PC(isp, bus);
tgt = FC_PORTDB_TGT(isp, bus, lp);
isp_gen_role_str(buf, sizeof (buf), lp->new_prli_word3);
isp_prt(isp, ISP_LOGCONFIG, prom, bus, tgt, lp->port_wwn, lp->new_portid, lp->handle, buf, "changed");
changed:
if (lp->is_target !=
((FCPARAM(isp, bus)->role & ISP_ROLE_INITIATOR) &&
(lp->new_prli_word3 & PRLI_WD3_TARGET_FUNCTION))) {
lp->is_target = !lp->is_target;
if (lp->is_target) {
2015-11-14 19:47:17 +00:00
isp_fcp_reset_crn(isp, bus, tgt, /*tgt_set*/ 1);
isp_make_here(isp, lp, bus, tgt);
} else {
isp_make_gone(isp, lp, bus, tgt);
2015-11-14 19:47:17 +00:00
isp_fcp_reset_crn(isp, bus, tgt, /*tgt_set*/ 1);
}
}
if (lp->is_initiator !=
((FCPARAM(isp, bus)->role & ISP_ROLE_TARGET) &&
(lp->new_prli_word3 & PRLI_WD3_INITIATOR_FUNCTION))) {
lp->is_initiator = !lp->is_initiator;
ac.contract_number = AC_CONTRACT_DEV_CHG;
adc = (struct ac_device_changed *) ac.contract_data;
adc->wwpn = lp->port_wwn;
adc->port = lp->portid;
adc->target = tgt;
adc->arrived = lp->is_initiator;
xpt_async(AC_CONTRACT, fc->path, &ac);
}
break;
case ISPASYNC_DEV_STAYED:
va_start(ap, cmd);
bus = va_arg(ap, int);
lp = va_arg(ap, fcportdb_t *);
va_end(ap);
fc = ISP_FC_PC(isp, bus);
tgt = FC_PORTDB_TGT(isp, bus, lp);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_gen_role_str(buf, sizeof (buf), lp->prli_word3);
isp_prt(isp, ISP_LOGCONFIG, prom, bus, tgt, lp->port_wwn, lp->portid, lp->handle, buf, "stayed");
goto changed;
case ISPASYNC_DEV_GONE:
va_start(ap, cmd);
bus = va_arg(ap, int);
lp = va_arg(ap, fcportdb_t *);
va_end(ap);
fc = ISP_FC_PC(isp, bus);
tgt = FC_PORTDB_TGT(isp, bus, lp);
/*
* If this has a virtual target or initiator set the isp_gdt
* timer running on it to delay its departure.
*/
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_gen_role_str(buf, sizeof (buf), lp->prli_word3);
if (lp->is_target || lp->is_initiator) {
lp->state = FC_PORTDB_STATE_ZOMBIE;
lp->gone_timer = fc->gone_device_time;
isp_prt(isp, ISP_LOGCONFIG, prom, bus, tgt, lp->port_wwn, lp->portid, lp->handle, buf, "gone zombie");
if (fc->ready && !callout_active(&fc->gdt)) {
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_prt(isp, ISP_LOG_SANCFG|ISP_LOGDEBUG0, "Chan %d Starting Gone Device Timer with %u seconds time now %lu", bus, lp->gone_timer, (unsigned long)time_uptime);
callout_reset(&fc->gdt, hz, isp_gdt, fc);
}
break;
}
isp_prt(isp, ISP_LOGCONFIG, prom, bus, tgt, lp->port_wwn, lp->portid, lp->handle, buf, "gone");
break;
case ISPASYNC_CHANGE_NOTIFY:
{
char *msg;
int evt, nphdl, nlstate, portid, reason;
va_start(ap, cmd);
bus = va_arg(ap, int);
evt = va_arg(ap, int);
if (evt == ISPASYNC_CHANGE_PDB) {
nphdl = va_arg(ap, int);
nlstate = va_arg(ap, int);
reason = va_arg(ap, int);
} else if (evt == ISPASYNC_CHANGE_SNS) {
portid = va_arg(ap, int);
} else {
nphdl = NIL_HANDLE;
nlstate = reason = 0;
}
va_end(ap);
fc = ISP_FC_PC(isp, bus);
if (evt == ISPASYNC_CHANGE_PDB) {
msg = "Port Database Changed";
isp_prt(isp, ISP_LOGINFO,
"Chan %d %s (nphdl 0x%x state 0x%x reason 0x%x)",
bus, msg, nphdl, nlstate, reason);
} else if (evt == ISPASYNC_CHANGE_SNS) {
msg = "Name Server Database Changed";
isp_prt(isp, ISP_LOGINFO, "Chan %d %s (PortID 0x%06x)",
bus, msg, portid);
} else {
msg = "Other Change Notify";
isp_prt(isp, ISP_LOGINFO, "Chan %d %s", bus, msg);
}
isp_loop_changed(isp, bus);
break;
}
#ifdef ISP_TARGET_MODE
case ISPASYNC_TARGET_NOTIFY:
{
isp_notify_t *notify;
va_start(ap, cmd);
notify = va_arg(ap, isp_notify_t *);
va_end(ap);
switch (notify->nt_ncode) {
case NT_ABORT_TASK:
case NT_ABORT_TASK_SET:
case NT_CLEAR_ACA:
case NT_CLEAR_TASK_SET:
case NT_LUN_RESET:
case NT_TARGET_RESET:
case NT_QUERY_TASK_SET:
case NT_QUERY_ASYNC_EVENT:
/*
* These are task management functions.
*/
isp_handle_platform_target_tmf(isp, notify);
break;
case NT_BUS_RESET:
case NT_LIP_RESET:
case NT_LINK_UP:
case NT_LINK_DOWN:
case NT_HBA_RESET:
/*
* No action need be taken here.
*/
break;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
case NT_GLOBAL_LOGOUT:
case NT_LOGOUT:
/*
* This is device arrival/departure notification
*/
isp_handle_platform_target_notify_ack(isp, notify, 0);
break;
default:
isp_prt(isp, ISP_LOGALL, "target notify code 0x%x", notify->nt_ncode);
isp_handle_platform_target_notify_ack(isp, notify, 0);
break;
}
break;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
case ISPASYNC_TARGET_NOTIFY_ACK:
{
void *inot;
va_start(ap, cmd);
inot = va_arg(ap, void *);
va_end(ap);
if (isp_notify_ack(isp, inot)) {
isp_tna_t *tp = malloc(sizeof (*tp), M_DEVBUF, M_NOWAIT);
if (tp) {
tp->isp = isp;
if (inot) {
memcpy(tp->data, inot, sizeof (tp->data));
tp->not = tp->data;
} else {
tp->not = NULL;
}
callout_init_mtx(&tp->timer, &isp->isp_lock, 0);
callout_reset(&tp->timer, 5,
isp_refire_notify_ack, tp);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
} else {
isp_prt(isp, ISP_LOGERR, "you lose- cannot allocate a notify refire");
}
}
break;
}
case ISPASYNC_TARGET_ACTION:
{
isphdr_t *hp;
va_start(ap, cmd);
hp = va_arg(ap, isphdr_t *);
va_end(ap);
switch (hp->rqs_entry_type) {
default:
isp_prt(isp, ISP_LOGWARN, "%s: unhandled target action 0x%x", __func__, hp->rqs_entry_type);
break;
case RQSTYPE_NOTIFY:
if (IS_24XX(isp)) {
isp_handle_platform_notify_24xx(isp, (in_fcentry_24xx_t *) hp);
} else {
isp_handle_platform_notify_fc(isp, (in_fcentry_t *) hp);
}
break;
case RQSTYPE_ATIO:
isp_handle_platform_atio7(isp, (at7_entry_t *) hp);
break;
case RQSTYPE_ATIO2:
isp_handle_platform_atio2(isp, (at2_entry_t *) hp);
break;
case RQSTYPE_CTIO7:
case RQSTYPE_CTIO3:
case RQSTYPE_CTIO2:
case RQSTYPE_CTIO:
isp_handle_platform_ctio(isp, hp);
break;
case RQSTYPE_ABTS_RCVD:
{
abts_t *abts = (abts_t *)hp;
isp_notify_t notify, *nt = &notify;
tstate_t *tptr;
fcportdb_t *lp;
uint16_t chan;
uint32_t sid, did;
did = (abts->abts_did_hi << 16) | abts->abts_did_lo;
sid = (abts->abts_sid_hi << 16) | abts->abts_sid_lo;
ISP_MEMZERO(nt, sizeof (isp_notify_t));
nt->nt_hba = isp;
nt->nt_did = did;
nt->nt_nphdl = abts->abts_nphdl;
nt->nt_sid = sid;
isp_find_chan_by_did(isp, did, &chan);
if (chan == ISP_NOCHAN) {
nt->nt_tgt = TGT_ANY;
} else {
nt->nt_tgt = FCPARAM(isp, chan)->isp_wwpn;
if (isp_find_pdb_by_handle(isp, chan, abts->abts_nphdl, &lp)) {
nt->nt_wwn = lp->port_wwn;
} else {
nt->nt_wwn = INI_ANY;
}
}
/*
* Try hard to find the lun for this command.
*/
tptr = get_lun_statep_from_tag(isp, chan, abts->abts_rxid_task);
if (tptr) {
nt->nt_lun = tptr->ts_lun;
rls_lun_statep(isp, tptr);
} else {
nt->nt_lun = LUN_ANY;
}
nt->nt_need_ack = 1;
nt->nt_tagval = abts->abts_rxid_task;
nt->nt_tagval |= (((uint64_t) abts->abts_rxid_abts) << 32);
if (abts->abts_rxid_task == ISP24XX_NO_TASK) {
isp_prt(isp, ISP_LOGTINFO, "[0x%x] ABTS from N-Port handle 0x%x Port 0x%06x has no task id (rx_id 0x%04x ox_id 0x%04x)",
abts->abts_rxid_abts, abts->abts_nphdl, sid, abts->abts_rx_id, abts->abts_ox_id);
} else {
isp_prt(isp, ISP_LOGTINFO, "[0x%x] ABTS from N-Port handle 0x%x Port 0x%06x for task 0x%x (rx_id 0x%04x ox_id 0x%04x)",
abts->abts_rxid_abts, abts->abts_nphdl, sid, abts->abts_rxid_task, abts->abts_rx_id, abts->abts_ox_id);
}
nt->nt_channel = chan;
nt->nt_ncode = NT_ABORT_TASK;
nt->nt_lreserved = hp;
isp_handle_platform_target_tmf(isp, nt);
break;
}
}
break;
}
#endif
case ISPASYNC_FW_CRASH:
{
uint16_t mbox1, mbox6;
mbox1 = ISP_READ(isp, OUTMAILBOX1);
if (IS_DUALBUS(isp)) {
mbox6 = ISP_READ(isp, OUTMAILBOX6);
} else {
mbox6 = 0;
}
isp_prt(isp, ISP_LOGERR, "Internal Firmware Error on bus %d @ RISC Address 0x%x", mbox6, mbox1);
mbox1 = isp->isp_osinfo.mbox_sleep_ok;
isp->isp_osinfo.mbox_sleep_ok = 0;
isp_reinit(isp, 1);
isp->isp_osinfo.mbox_sleep_ok = mbox1;
isp_async(isp, ISPASYNC_FW_RESTARTED, NULL);
break;
}
default:
isp_prt(isp, ISP_LOGERR, "unknown isp_async event %d", cmd);
break;
}
}
/*
* Locks are held before coming here.
*/
void
isp_uninit(ispsoftc_t *isp)
{
if (IS_24XX(isp)) {
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_RESET);
} else {
ISP_WRITE(isp, HCCR, HCCR_CMD_RESET);
}
ISP_DISABLE_INTS(isp);
}
uint64_t
isp_default_wwn(ispsoftc_t * isp, int chan, int isactive, int iswwnn)
{
uint64_t seed;
struct isp_fc *fc = ISP_FC_PC(isp, chan);
/* First try to use explicitly configured WWNs. */
seed = iswwnn ? fc->def_wwnn : fc->def_wwpn;
if (seed)
return (seed);
/* Otherwise try to use WWNs from NVRAM. */
if (isactive) {
seed = iswwnn ? FCPARAM(isp, chan)->isp_wwnn_nvram :
FCPARAM(isp, chan)->isp_wwpn_nvram;
if (seed)
return (seed);
}
/* If still no WWNs, try to steal them from the first channel. */
if (chan > 0) {
seed = iswwnn ? ISP_FC_PC(isp, 0)->def_wwnn :
ISP_FC_PC(isp, 0)->def_wwpn;
if (seed == 0) {
seed = iswwnn ? FCPARAM(isp, 0)->isp_wwnn_nvram :
FCPARAM(isp, 0)->isp_wwpn_nvram;
}
}
/* If still nothing -- improvise. */
if (seed == 0) {
seed = 0x400000007F000000ull + device_get_unit(isp->isp_dev);
if (!iswwnn)
seed ^= 0x0100000000000000ULL;
}
/* For additional channels we have to improvise even more. */
if (!iswwnn && chan > 0) {
/*
* We'll stick our channel number plus one first into bits
* 57..59 and thence into bits 52..55 which allows for 8 bits
* of channel which is enough for our maximum of 255 channels.
*/
seed ^= 0x0100000000000000ULL;
seed ^= ((uint64_t) (chan + 1) & 0xf) << 56;
seed ^= ((uint64_t) ((chan + 1) >> 4) & 0xf) << 52;
}
return (seed);
}
void
isp_prt(ispsoftc_t *isp, int level, const char *fmt, ...)
{
int loc;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
char lbuf[200];
va_list ap;
if (level != ISP_LOGALL && (level & isp->isp_dblev) == 0) {
return;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
snprintf(lbuf, sizeof (lbuf), "%s: ", device_get_nameunit(isp->isp_dev));
loc = strlen(lbuf);
va_start(ap, fmt);
vsnprintf(&lbuf[loc], sizeof (lbuf) - loc - 1, fmt, ap);
va_end(ap);
printf("%s\n", lbuf);
}
void
isp_xs_prt(ispsoftc_t *isp, XS_T *xs, int level, const char *fmt, ...)
{
va_list ap;
if (level != ISP_LOGALL && (level & isp->isp_dblev) == 0) {
return;
}
xpt_print_path(xs->ccb_h.path);
va_start(ap, fmt);
vprintf(fmt, ap);
va_end(ap);
printf("\n");
}
uint64_t
isp_nanotime_sub(struct timespec *b, struct timespec *a)
{
uint64_t elapsed;
struct timespec x = *b;
timespecsub(&x, a);
elapsed = GET_NANOSEC(&x);
if (elapsed == 0)
elapsed++;
return (elapsed);
}
int
isp_mbox_acquire(ispsoftc_t *isp)
{
if (isp->isp_osinfo.mboxbsy) {
return (1);
} else {
isp->isp_osinfo.mboxcmd_done = 0;
isp->isp_osinfo.mboxbsy = 1;
return (0);
}
}
void
isp_mbox_wait_complete(ispsoftc_t *isp, mbreg_t *mbp)
{
unsigned int usecs = mbp->timeout;
unsigned int max, olim, ilim;
if (usecs == 0) {
usecs = MBCMD_DEFAULT_TIMEOUT;
}
max = isp->isp_mbxwrk0 + 1;
if (isp->isp_osinfo.mbox_sleep_ok) {
unsigned int ms = (usecs + 999) / 1000;
isp->isp_osinfo.mbox_sleep_ok = 0;
isp->isp_osinfo.mbox_sleeping = 1;
for (olim = 0; olim < max; olim++) {
msleep(&isp->isp_mbxworkp, &isp->isp_osinfo.lock, PRIBIO, "ispmbx_sleep", isp_mstohz(ms));
if (isp->isp_osinfo.mboxcmd_done) {
break;
}
}
isp->isp_osinfo.mbox_sleep_ok = 1;
isp->isp_osinfo.mbox_sleeping = 0;
} else {
for (olim = 0; olim < max; olim++) {
for (ilim = 0; ilim < usecs; ilim += 100) {
uint16_t isr, sema, info;
if (isp->isp_osinfo.mboxcmd_done) {
break;
}
if (ISP_READ_ISR(isp, &isr, &sema, &info)) {
isp_intr(isp, isr, sema, info);
if (isp->isp_osinfo.mboxcmd_done) {
break;
}
}
ISP_DELAY(100);
}
if (isp->isp_osinfo.mboxcmd_done) {
break;
}
}
}
if (isp->isp_osinfo.mboxcmd_done == 0) {
isp_prt(isp, ISP_LOGWARN, "%s Mailbox Command (0x%x) Timeout (%uus) (started @ %s:%d)",
isp->isp_osinfo.mbox_sleep_ok? "Interrupting" : "Polled", isp->isp_lastmbxcmd, usecs, mbp->func, mbp->lineno);
mbp->param[0] = MBOX_TIMEOUT;
isp->isp_osinfo.mboxcmd_done = 1;
}
}
void
isp_mbox_notify_done(ispsoftc_t *isp)
{
if (isp->isp_osinfo.mbox_sleeping) {
wakeup(&isp->isp_mbxworkp);
}
isp->isp_osinfo.mboxcmd_done = 1;
}
void
isp_mbox_release(ispsoftc_t *isp)
{
isp->isp_osinfo.mboxbsy = 0;
}
int
isp_fc_scratch_acquire(ispsoftc_t *isp, int chan)
{
int ret = 0;
if (isp->isp_osinfo.pc.fc[chan].fcbsy) {
ret = -1;
} else {
isp->isp_osinfo.pc.fc[chan].fcbsy = 1;
}
return (ret);
}
int
isp_mstohz(int ms)
{
int hz;
struct timeval t;
t.tv_sec = ms / 1000;
t.tv_usec = (ms % 1000) * 1000;
hz = tvtohz(&t);
if (hz < 0) {
hz = 0x7fffffff;
}
if (hz == 0) {
hz = 1;
}
return (hz);
}
void
isp_platform_intr(void *arg)
{
ispsoftc_t *isp = arg;
uint16_t isr, sema, info;
ISP_LOCK(isp);
isp->isp_intcnt++;
if (ISP_READ_ISR(isp, &isr, &sema, &info))
isp_intr(isp, isr, sema, info);
else
isp->isp_intbogus++;
ISP_UNLOCK(isp);
}
void
isp_common_dmateardown(ispsoftc_t *isp, struct ccb_scsiio *csio, uint32_t hdl)
{
if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
bus_dmamap_sync(isp->isp_osinfo.dmat, PISP_PCMD(csio)->dmap, BUS_DMASYNC_POSTREAD);
} else {
bus_dmamap_sync(isp->isp_osinfo.dmat, PISP_PCMD(csio)->dmap, BUS_DMASYNC_POSTWRITE);
}
bus_dmamap_unload(isp->isp_osinfo.dmat, PISP_PCMD(csio)->dmap);
}
Fix Fibre Channel Command Reference Number handling in the isp(4) driver. The Command Reference Number is used for precise delivery of commands, and is part of the FC-Tape functionality set. (This is only enabled for devices that support precise delivery of commands.) It is an 8-bit unsigned number that increments from 1 to 255. The commands sent by the initiator must be processed by the target in CRN order if the CRN is non-zero. There are certain scenarios where the Command Reference Number sequence needs to be reset. When the target is power cycled, for instance, the initiator needs to reset the CRN to 1. The initiator will know this because it will see a LIP (when directly connected) or get a logout/login event (when connected to a switch). The isp(4) driver was not resetting the CRN when a target went away and came back. When it saw the target again after a power cycle, it would continue the CRN sequence where it left off. The target would ignore the command because the CRN sequence is supposed to be reset to 1 after a power cycle or other similar event. The symptom that the user would see is that there would be lots of aborted INQUIRY commands after a tape library was power cycled, and the library would fail to probe. The INQUIRY commands were being ignored by the tape drive due to the CRN issue mentioned above. isp_freebsd.c: Add a new function, isp_fcp_reset_crn(). This will reset all of the CRNs for a given port, or the CRNs for all LUNs on a target. Reset the CRNs for all targets on a port when we get a LIP, loop reset, or loop down event. Reset the CRN for a particular target when it arrives, is changed or departs. This is less precise behavior than the clearing behavior specified in the FCP-4 spec (which says that it should be reset for PRLI, PRLO, PLOGI and LOGO), but this is the level of information we have here. If this is insufficient, then we will need to add more precise notification from the lower level isp(4) code. isp_freebsd.h: Add a prototype for isp_fcp_reset_crn(). Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:41:28 +00:00
/*
* Reset the command reference number for all LUNs on a specific target
* (needed when a target arrives again) or for all targets on a port
* (needed for events like a LIP).
*/
void
2015-11-14 19:47:17 +00:00
isp_fcp_reset_crn(ispsoftc_t *isp, int chan, uint32_t tgt, int tgt_set)
Fix Fibre Channel Command Reference Number handling in the isp(4) driver. The Command Reference Number is used for precise delivery of commands, and is part of the FC-Tape functionality set. (This is only enabled for devices that support precise delivery of commands.) It is an 8-bit unsigned number that increments from 1 to 255. The commands sent by the initiator must be processed by the target in CRN order if the CRN is non-zero. There are certain scenarios where the Command Reference Number sequence needs to be reset. When the target is power cycled, for instance, the initiator needs to reset the CRN to 1. The initiator will know this because it will see a LIP (when directly connected) or get a logout/login event (when connected to a switch). The isp(4) driver was not resetting the CRN when a target went away and came back. When it saw the target again after a power cycle, it would continue the CRN sequence where it left off. The target would ignore the command because the CRN sequence is supposed to be reset to 1 after a power cycle or other similar event. The symptom that the user would see is that there would be lots of aborted INQUIRY commands after a tape library was power cycled, and the library would fail to probe. The INQUIRY commands were being ignored by the tape drive due to the CRN issue mentioned above. isp_freebsd.c: Add a new function, isp_fcp_reset_crn(). This will reset all of the CRNs for a given port, or the CRNs for all LUNs on a target. Reset the CRNs for all targets on a port when we get a LIP, loop reset, or loop down event. Reset the CRN for a particular target when it arrives, is changed or departs. This is less precise behavior than the clearing behavior specified in the FCP-4 spec (which says that it should be reset for PRLI, PRLO, PLOGI and LOGO), but this is the level of information we have here. If this is insufficient, then we will need to add more precise notification from the lower level isp(4) code. isp_freebsd.h: Add a prototype for isp_fcp_reset_crn(). Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:41:28 +00:00
{
2015-11-14 19:47:17 +00:00
struct isp_fc *fc = ISP_FC_PC(isp, chan);
Fix Fibre Channel Command Reference Number handling in the isp(4) driver. The Command Reference Number is used for precise delivery of commands, and is part of the FC-Tape functionality set. (This is only enabled for devices that support precise delivery of commands.) It is an 8-bit unsigned number that increments from 1 to 255. The commands sent by the initiator must be processed by the target in CRN order if the CRN is non-zero. There are certain scenarios where the Command Reference Number sequence needs to be reset. When the target is power cycled, for instance, the initiator needs to reset the CRN to 1. The initiator will know this because it will see a LIP (when directly connected) or get a logout/login event (when connected to a switch). The isp(4) driver was not resetting the CRN when a target went away and came back. When it saw the target again after a power cycle, it would continue the CRN sequence where it left off. The target would ignore the command because the CRN sequence is supposed to be reset to 1 after a power cycle or other similar event. The symptom that the user would see is that there would be lots of aborted INQUIRY commands after a tape library was power cycled, and the library would fail to probe. The INQUIRY commands were being ignored by the tape drive due to the CRN issue mentioned above. isp_freebsd.c: Add a new function, isp_fcp_reset_crn(). This will reset all of the CRNs for a given port, or the CRNs for all LUNs on a target. Reset the CRNs for all targets on a port when we get a LIP, loop reset, or loop down event. Reset the CRN for a particular target when it arrives, is changed or departs. This is less precise behavior than the clearing behavior specified in the FCP-4 spec (which says that it should be reset for PRLI, PRLO, PLOGI and LOGO), but this is the level of information we have here. If this is insufficient, then we will need to add more precise notification from the lower level isp(4) code. isp_freebsd.h: Add a prototype for isp_fcp_reset_crn(). Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:41:28 +00:00
struct isp_nexus *nxp;
2015-11-14 19:47:17 +00:00
int i;
Fix Fibre Channel Command Reference Number handling in the isp(4) driver. The Command Reference Number is used for precise delivery of commands, and is part of the FC-Tape functionality set. (This is only enabled for devices that support precise delivery of commands.) It is an 8-bit unsigned number that increments from 1 to 255. The commands sent by the initiator must be processed by the target in CRN order if the CRN is non-zero. There are certain scenarios where the Command Reference Number sequence needs to be reset. When the target is power cycled, for instance, the initiator needs to reset the CRN to 1. The initiator will know this because it will see a LIP (when directly connected) or get a logout/login event (when connected to a switch). The isp(4) driver was not resetting the CRN when a target went away and came back. When it saw the target again after a power cycle, it would continue the CRN sequence where it left off. The target would ignore the command because the CRN sequence is supposed to be reset to 1 after a power cycle or other similar event. The symptom that the user would see is that there would be lots of aborted INQUIRY commands after a tape library was power cycled, and the library would fail to probe. The INQUIRY commands were being ignored by the tape drive due to the CRN issue mentioned above. isp_freebsd.c: Add a new function, isp_fcp_reset_crn(). This will reset all of the CRNs for a given port, or the CRNs for all LUNs on a target. Reset the CRNs for all targets on a port when we get a LIP, loop reset, or loop down event. Reset the CRN for a particular target when it arrives, is changed or departs. This is less precise behavior than the clearing behavior specified in the FCP-4 spec (which says that it should be reset for PRLI, PRLO, PLOGI and LOGO), but this is the level of information we have here. If this is insufficient, then we will need to add more precise notification from the lower level isp(4) code. isp_freebsd.h: Add a prototype for isp_fcp_reset_crn(). Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:41:28 +00:00
if (tgt_set == 0)
2015-11-14 19:47:17 +00:00
isp_prt(isp, ISP_LOGDEBUG0,
"Chan %d resetting CRN on all targets", chan);
Fix Fibre Channel Command Reference Number handling in the isp(4) driver. The Command Reference Number is used for precise delivery of commands, and is part of the FC-Tape functionality set. (This is only enabled for devices that support precise delivery of commands.) It is an 8-bit unsigned number that increments from 1 to 255. The commands sent by the initiator must be processed by the target in CRN order if the CRN is non-zero. There are certain scenarios where the Command Reference Number sequence needs to be reset. When the target is power cycled, for instance, the initiator needs to reset the CRN to 1. The initiator will know this because it will see a LIP (when directly connected) or get a logout/login event (when connected to a switch). The isp(4) driver was not resetting the CRN when a target went away and came back. When it saw the target again after a power cycle, it would continue the CRN sequence where it left off. The target would ignore the command because the CRN sequence is supposed to be reset to 1 after a power cycle or other similar event. The symptom that the user would see is that there would be lots of aborted INQUIRY commands after a tape library was power cycled, and the library would fail to probe. The INQUIRY commands were being ignored by the tape drive due to the CRN issue mentioned above. isp_freebsd.c: Add a new function, isp_fcp_reset_crn(). This will reset all of the CRNs for a given port, or the CRNs for all LUNs on a target. Reset the CRNs for all targets on a port when we get a LIP, loop reset, or loop down event. Reset the CRN for a particular target when it arrives, is changed or departs. This is less precise behavior than the clearing behavior specified in the FCP-4 spec (which says that it should be reset for PRLI, PRLO, PLOGI and LOGO), but this is the level of information we have here. If this is insufficient, then we will need to add more precise notification from the lower level isp(4) code. isp_freebsd.h: Add a prototype for isp_fcp_reset_crn(). Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:41:28 +00:00
else
2015-11-14 19:47:17 +00:00
isp_prt(isp, ISP_LOGDEBUG0,
"Chan %d resetting CRN on target %u", chan, tgt);
Fix Fibre Channel Command Reference Number handling in the isp(4) driver. The Command Reference Number is used for precise delivery of commands, and is part of the FC-Tape functionality set. (This is only enabled for devices that support precise delivery of commands.) It is an 8-bit unsigned number that increments from 1 to 255. The commands sent by the initiator must be processed by the target in CRN order if the CRN is non-zero. There are certain scenarios where the Command Reference Number sequence needs to be reset. When the target is power cycled, for instance, the initiator needs to reset the CRN to 1. The initiator will know this because it will see a LIP (when directly connected) or get a logout/login event (when connected to a switch). The isp(4) driver was not resetting the CRN when a target went away and came back. When it saw the target again after a power cycle, it would continue the CRN sequence where it left off. The target would ignore the command because the CRN sequence is supposed to be reset to 1 after a power cycle or other similar event. The symptom that the user would see is that there would be lots of aborted INQUIRY commands after a tape library was power cycled, and the library would fail to probe. The INQUIRY commands were being ignored by the tape drive due to the CRN issue mentioned above. isp_freebsd.c: Add a new function, isp_fcp_reset_crn(). This will reset all of the CRNs for a given port, or the CRNs for all LUNs on a target. Reset the CRNs for all targets on a port when we get a LIP, loop reset, or loop down event. Reset the CRN for a particular target when it arrives, is changed or departs. This is less precise behavior than the clearing behavior specified in the FCP-4 spec (which says that it should be reset for PRLI, PRLO, PLOGI and LOGO), but this is the level of information we have here. If this is insufficient, then we will need to add more precise notification from the lower level isp(4) code. isp_freebsd.h: Add a prototype for isp_fcp_reset_crn(). Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:41:28 +00:00
for (i = 0; i < NEXUS_HASH_WIDTH; i++) {
2015-11-14 19:47:17 +00:00
for (nxp = fc->nexus_hash[i]; nxp != NULL; nxp = nxp->next) {
if (tgt_set == 0 || tgt == nxp->tgt)
Fix Fibre Channel Command Reference Number handling in the isp(4) driver. The Command Reference Number is used for precise delivery of commands, and is part of the FC-Tape functionality set. (This is only enabled for devices that support precise delivery of commands.) It is an 8-bit unsigned number that increments from 1 to 255. The commands sent by the initiator must be processed by the target in CRN order if the CRN is non-zero. There are certain scenarios where the Command Reference Number sequence needs to be reset. When the target is power cycled, for instance, the initiator needs to reset the CRN to 1. The initiator will know this because it will see a LIP (when directly connected) or get a logout/login event (when connected to a switch). The isp(4) driver was not resetting the CRN when a target went away and came back. When it saw the target again after a power cycle, it would continue the CRN sequence where it left off. The target would ignore the command because the CRN sequence is supposed to be reset to 1 after a power cycle or other similar event. The symptom that the user would see is that there would be lots of aborted INQUIRY commands after a tape library was power cycled, and the library would fail to probe. The INQUIRY commands were being ignored by the tape drive due to the CRN issue mentioned above. isp_freebsd.c: Add a new function, isp_fcp_reset_crn(). This will reset all of the CRNs for a given port, or the CRNs for all LUNs on a target. Reset the CRNs for all targets on a port when we get a LIP, loop reset, or loop down event. Reset the CRN for a particular target when it arrives, is changed or departs. This is less precise behavior than the clearing behavior specified in the FCP-4 spec (which says that it should be reset for PRLI, PRLO, PLOGI and LOGO), but this is the level of information we have here. If this is insufficient, then we will need to add more precise notification from the lower level isp(4) code. isp_freebsd.h: Add a prototype for isp_fcp_reset_crn(). Sponsored by: Spectra Logic MFC after: 1 week
2015-01-08 17:41:28 +00:00
nxp->crnseed = 0;
}
}
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
int
isp_fcp_next_crn(ispsoftc_t *isp, uint8_t *crnp, XS_T *cmd)
{
lun_id_t lun;
uint32_t chan, tgt;
struct isp_fc *fc;
struct isp_nexus *nxp;
int idx;
if (IS_2100(isp))
return (0);
chan = XS_CHANNEL(cmd);
tgt = XS_TGT(cmd);
lun = XS_LUN(cmd);
fc = &isp->isp_osinfo.pc.fc[chan];
idx = NEXUS_HASH(tgt, lun);
nxp = fc->nexus_hash[idx];
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
while (nxp) {
if (nxp->tgt == tgt && nxp->lun == lun)
break;
nxp = nxp->next;
}
if (nxp == NULL) {
nxp = fc->nexus_free_list;
if (nxp == NULL) {
nxp = malloc(sizeof (struct isp_nexus), M_DEVBUF, M_ZERO|M_NOWAIT);
if (nxp == NULL) {
return (-1);
}
} else {
fc->nexus_free_list = nxp->next;
}
nxp->tgt = tgt;
nxp->lun = lun;
nxp->next = fc->nexus_hash[idx];
fc->nexus_hash[idx] = nxp;
}
2015-11-14 19:47:17 +00:00
if (nxp->crnseed == 0)
nxp->crnseed = 1;
PISP_PCMD(cmd)->crn = nxp->crnseed;
*crnp = nxp->crnseed++;
return (0);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
/*
* We enter with the lock held
*/
void
isp_timer(void *arg)
{
ispsoftc_t *isp = arg;
#ifdef ISP_TARGET_MODE
isp_tmcmd_restart(isp);
#endif
More rototilling with target mode in an attemp to get multiple CCB at a time outstanding reliable. It's not there yet, but this is the direction to go in so might as well commit. So far, multiple at a time CCBs work (see ISP_INTERNAL_TARGET test mode), but it fails if there are more downstream than the SIM wants to handle and SRR is sort of confused when this happens, plus it is not entirely quite clear what one does if a CCB/CTIO fails and you have more in flight (that don't fail, say) and more queued up at the SIM level that haven't been started yet. Some of this is driven because there apparently is no flow control to requeue XPT_CONTINUE_IO requests like there are for XPT_SCSI_IO requests. It is also more driven in that the few target mode periph drivers there are are not really set up for handling pushback- heck most of them don't even check for errors (and what would they really do with them anyway? It's the initiator's problem, really....). The data transfer arithmetic has been worked over again to handle multiple outstanding commands, so you have a notion of what's been moved already as well as what's currently in flight. It turns that this led to uncovering a REPORT_LUNS bug in the ISP_INTERNAL_TARGET code which was sending back 24 bytes of rpl data instead of the specified 16. What happened furthermore here is that sending back 16 bytes and reporting an overrun of 8 bytes made the initiator (running FC-Tape aware f/w) mad enough to request, and keep requesting, another FCP response (I guess it didn't like the answer so kept asking for it again). Sponsored by: Spectralogic MFC after: 1 month
2012-08-08 18:24:33 +00:00
callout_reset(&isp->isp_osinfo.tmo, isp_timer_count, isp_timer, isp);
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_ecmd_t *
isp_get_ecmd(ispsoftc_t *isp)
{
isp_ecmd_t *ecmd = isp->isp_osinfo.ecmd_free;
if (ecmd) {
isp->isp_osinfo.ecmd_free = ecmd->next;
}
return (ecmd);
}
void
isp_put_ecmd(ispsoftc_t *isp, isp_ecmd_t *ecmd)
{
ecmd->next = isp->isp_osinfo.ecmd_free;
isp->isp_osinfo.ecmd_free = ecmd;
}