From: Finn Thain <fthain@telegraphics.com.au>
To: "James E.J. Bottomley" <JBottomley@odin.com>,
Michael Schmitz <schmitzmic@gmail.com>,
linux-m68k@vger.kernel.org, linux-scsi@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: [PATCH 37/71] ncr5380: Standardize work queueing algorithm
Date: Wed, 18 Nov 2015 19:35:32 +1100 [thread overview]
Message-ID: <20151118083505.931718684@telegraphics.com.au> (raw)
In-Reply-To: 20151118083455.331768508@telegraphics.com.au
[-- Attachment #1: atari_NCR5380-main_running-and-queue_main --]
[-- Type: text/plain, Size: 6330 bytes --]
The complex main_running/queue_main mechanism is peculiar to
atari_NCR5380.c. It isn't SMP safe and offers little value given that
the work queue already offers concurrency management. Remove this
complexity to bring atari_NCR5380.c closer to NCR5380.c.
It is not a good idea to call the information transfer state machine from
queuecommand because, according to Documentation/scsi/scsi_mid_low_api.txt
that could happen in soft irq context. Fix this.
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
---
drivers/scsi/NCR5380.h | 1
drivers/scsi/atari_NCR5380.c | 80 +++----------------------------------------
2 files changed, 6 insertions(+), 75 deletions(-)
Index: linux/drivers/scsi/NCR5380.h
===================================================================
--- linux.orig/drivers/scsi/NCR5380.h 2015-11-18 19:33:49.000000000 +1100
+++ linux/drivers/scsi/NCR5380.h 2015-11-18 19:33:51.000000000 +1100
@@ -262,7 +262,6 @@ struct NCR5380_hostdata {
* transfer to handle chip overruns */
int retain_dma_intr;
struct work_struct main_task;
- volatile int main_running;
#ifdef SUPPORT_TAGS
struct tag_alloc TagAlloc[8][8]; /* 8 targets and 8 LUNs */
#endif
Index: linux/drivers/scsi/atari_NCR5380.c
===================================================================
--- linux.orig/drivers/scsi/atari_NCR5380.c 2015-11-18 19:33:48.000000000 +1100
+++ linux/drivers/scsi/atari_NCR5380.c 2015-11-18 19:33:51.000000000 +1100
@@ -611,36 +611,6 @@ static void NCR5380_print_phase(struct S
#endif
-/*
- * ++roman: New scheme of calling NCR5380_main()
- *
- * If we're not in an interrupt, we can call our main directly, it cannot be
- * already running. Else, we queue it on a task queue, if not 'main_running'
- * tells us that a lower level is already executing it. This way,
- * 'main_running' needs not be protected in a special way.
- *
- * queue_main() is a utility function for putting our main onto the task
- * queue, if main_running is false. It should be called only from a
- * interrupt or bottom half.
- */
-
-#include <linux/gfp.h>
-#include <linux/workqueue.h>
-#include <linux/interrupt.h>
-
-static inline void queue_main(struct NCR5380_hostdata *hostdata)
-{
- if (!hostdata->main_running) {
- /* If in interrupt and NCR5380_main() not already running,
- queue it on the 'immediate' task queue, to be processed
- immediately after the current interrupt processing has
- finished. */
- queue_work(hostdata->work_q, &hostdata->main_task);
- }
- /* else: nothing to do: the running NCR5380_main() will pick up
- any newly queued command. */
-}
-
/**
* NCR58380_info - report driver and host information
* @instance: relevant scsi host instance
@@ -723,8 +693,6 @@ static void __maybe_unused NCR5380_print
hostdata = (struct NCR5380_hostdata *)instance->hostdata;
local_irq_save(flags);
- printk("NCR5380: coroutine is%s running.\n",
- hostdata->main_running ? "" : "n't");
if (!hostdata->connected)
printk("scsi%d: no currently connected command\n", HOSTNO);
else
@@ -766,8 +734,6 @@ static int __maybe_unused NCR5380_show_i
hostdata = (struct NCR5380_hostdata *)instance->hostdata;
local_irq_save(flags);
- seq_printf(m, "NCR5380: coroutine is%s running.\n",
- hostdata->main_running ? "" : "n't");
if (!hostdata->connected)
seq_printf(m, "scsi%d: no currently connected command\n", HOSTNO);
else
@@ -991,17 +957,8 @@ static int NCR5380_queue_command(struct
dprintk(NDEBUG_QUEUES, "scsi%d: command added to %s of queue\n", H_NO(cmd),
(cmd->cmnd[0] == REQUEST_SENSE) ? "head" : "tail");
- /* If queue_command() is called from an interrupt (real one or bottom
- * half), we let queue_main() do the job of taking care about main. If it
- * is already running, this is a no-op, else main will be queued.
- *
- * If we're not in an interrupt, we can call NCR5380_main()
- * unconditionally, because it cannot be already running.
- */
- if (in_interrupt() || irqs_disabled())
- queue_main(hostdata);
- else
- NCR5380_main(&hostdata->main_task);
+ /* Kick off command processing */
+ queue_work(hostdata->work_q, &hostdata->main_task);
return 0;
}
@@ -1038,30 +995,11 @@ static void NCR5380_main(struct work_str
unsigned long flags;
/*
- * We run (with interrupts disabled) until we're sure that none of
- * the host adapters have anything that can be done, at which point
- * we set main_running to 0 and exit.
- *
- * Interrupts are enabled before doing various other internal
- * instructions, after we've decided that we need to run through
- * the loop again.
- *
- * this should prevent any race conditions.
- *
* ++roman: Just disabling the NCR interrupt isn't sufficient here,
* because also a timer int can trigger an abort or reset, which can
* alter queues and touch the Falcon lock.
*/
- /* Tell int handlers main() is now already executing. Note that
- no races are possible here. If an int comes in before
- 'main_running' is set here, and queues/executes main via the
- task queue, it doesn't do any harm, just this instance of main
- won't find any work left to do. */
- if (hostdata->main_running)
- return;
- hostdata->main_running = 1;
-
local_save_flags(flags);
do {
local_irq_disable(); /* Freeze request queues */
@@ -1176,11 +1114,6 @@ static void NCR5380_main(struct work_str
done = 0;
}
} while (!done);
-
- /* Better allow ints _after_ 'main_running' has been cleared, else
- an interrupt could believe we'll pick up the work it left for
- us, but we won't see it anymore here... */
- hostdata->main_running = 0;
local_irq_restore(flags);
}
@@ -1293,6 +1226,7 @@ static void NCR5380_dma_complete(struct
static irqreturn_t NCR5380_intr(int irq, void *dev_id)
{
struct Scsi_Host *instance = dev_id;
+ struct NCR5380_hostdata *hostdata = shost_priv(instance);
int done = 1, handled = 0;
unsigned char basr;
@@ -1361,11 +1295,9 @@ static irqreturn_t NCR5380_intr(int irq,
#endif
}
- if (!done) {
- dprintk(NDEBUG_INTR, "scsi%d: in int routine, calling main\n", HOSTNO);
- /* Put a call to NCR5380_main() on the queue... */
- queue_main(shost_priv(instance));
- }
+ if (!done)
+ queue_work(hostdata->work_q, &hostdata->main_task);
+
return IRQ_RETVAL(handled);
}
next prev parent reply other threads:[~2015-11-18 8:35 UTC|newest]
Thread overview: 182+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-11-18 8:34 [PATCH 00/71] More fixes, cleanup and modernization for NCR5380 drivers Finn Thain
2015-11-18 8:34 ` [PATCH 01/71] atari_scsi: Fix SCSI host ID setting Finn Thain
2015-11-19 7:37 ` Hannes Reinecke
2015-11-18 8:34 ` [PATCH 02/71] ncr5380: Remove redundant static variable initializers Finn Thain
2015-11-19 7:37 ` Hannes Reinecke
2015-11-18 8:34 ` [PATCH 03/71] ncr5380: Eliminate PDEBUG*, TDEBUG* and DTCDEBUG* macros Finn Thain
2015-11-19 7:38 ` Hannes Reinecke
2015-11-18 8:34 ` [PATCH 04/71] ncr5380: Remove more pointless macros Finn Thain
2015-11-19 7:38 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 05/71] ncr5380: Remove NCR5380_local_declare and NCR5380_setup macros Finn Thain
2015-11-19 7:41 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 06/71] ncr5380: Remove NCR5380_instance_name macro Finn Thain
2015-11-19 7:41 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 07/71] ncr5380: Split NCR5380_init() into two functions Finn Thain
2015-11-19 13:30 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 08/71] ncr5380: Move NCR53C400-specific code Finn Thain
2015-11-19 13:31 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 09/71] atari_NCR5380: Reset bus on driver initialization if required Finn Thain
2015-11-19 13:38 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 10/71] atari_NCR5380: Remove RESET_BOOT, CONFIG_ATARI_SCSI_TOSHIBA_DELAY and CONFIG_ATARI_SCSI_RESET_BOOT Finn Thain
2015-11-19 3:05 ` Michael Schmitz
2015-11-19 4:05 ` Finn Thain
2015-11-19 4:43 ` Michael Schmitz
2015-11-19 13:41 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 11/71] ncr5380: Simplify bus reset handlers Finn Thain
2015-11-19 13:56 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 12/71] ncr5380: Remove unused hostdata->aborted flag Finn Thain
2015-11-19 14:13 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 13/71] ncr5380: Remove redundant register writes Finn Thain
2015-11-19 14:14 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 14/71] ncr5380: Use return instead of goto in NCR5380_select() Finn Thain
2015-11-19 14:14 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 15/71] ncr5380: Always escalate bad target time-out " Finn Thain
2015-11-19 14:15 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 16/71] ncr5380: Proceed with next command after NCR5380_select() calls scsi_done Finn Thain
2015-11-19 14:16 ` Hannes Reinecke
2015-11-18 8:35 ` [PATCH 17/71] ncr5380: Keep BSY asserted when entering SELECTION phase Finn Thain
2015-11-18 8:35 ` [PATCH 18/71] ncr5380: Eliminate USLEEP_WAITLONG delay Finn Thain
2015-11-18 8:35 ` [PATCH 19/71] ncr5380: Cleanup bogus {request,release}_region() calls Finn Thain
2015-11-18 8:35 ` [PATCH 20/71] ncr5380: Introduce unbound workqueue Finn Thain
2015-11-18 8:35 ` [PATCH 21/71] ncr5380: Sleep when polling, if possible Finn Thain
2015-11-18 8:35 ` [PATCH 22/71] ncr5380: Eliminate selecting state Finn Thain
2015-11-24 19:08 ` Ondrej Zary
2015-11-25 3:17 ` Finn Thain
2015-11-25 21:59 ` Ondrej Zary
2015-11-26 10:23 ` Finn Thain
2015-11-26 19:35 ` Ondrej Zary
2015-11-26 22:32 ` Finn Thain
2015-11-28 23:13 ` Ondrej Zary
2015-11-29 0:46 ` Finn Thain
2015-11-29 9:09 ` Ondrej Zary
2015-11-29 9:27 ` Geert Uytterhoeven
2015-11-29 10:25 ` Finn Thain
2015-11-29 13:50 ` Geert Uytterhoeven
2015-11-30 2:12 ` Finn Thain
2015-11-18 8:35 ` [PATCH 23/71] ncr5380: Always retry arbitration and selection Finn Thain
2015-11-18 8:35 ` [PATCH 24/71] ncr5380: Implement NCR5380_dma_xfer_len and remove LIMIT_TRANSFERSIZE macro Finn Thain
2015-11-18 8:35 ` [PATCH 25/71] ncr5380: Rework disconnect versus poll logic Finn Thain
2015-11-18 8:35 ` [PATCH 26/71] ncr5380: Fix NCR5380_transfer_pio() result Finn Thain
2015-11-18 8:35 ` [PATCH 27/71] ncr5380: Add missing lock in eh_abort_handler Finn Thain
2015-11-18 8:35 ` [PATCH 28/71] ncr5380: Drop DEF_SCSI_QCMD macro Finn Thain
2015-11-18 8:35 ` [PATCH 29/71] ncr5380: Remove references to linked commands Finn Thain
2015-11-18 8:35 ` [PATCH 30/71] ncr5380: Add missing break after case MESSAGE_REJECT Finn Thain
2015-11-18 8:35 ` [PATCH 31/71] ncr5380: Fix !REQ timeout in do_abort() Finn Thain
2015-11-18 8:35 ` [PATCH 32/71] ncr5380: Fix bus phase " Finn Thain
2015-11-18 8:35 ` [PATCH 33/71] atari_NCR5380: Set do_abort() timeouts Finn Thain
2015-11-18 8:35 ` [PATCH 34/71] atari_NCR5380: Use arbitration timeout Finn Thain
2015-11-18 8:35 ` [PATCH 35/71] ncr5380: Dont wait for BUS FREE after disconnect Finn Thain
2015-11-18 8:35 ` [PATCH 36/71] ncr5380: Use work_struct instead of delayed_work Finn Thain
2015-11-18 8:35 ` Finn Thain [this message]
2015-11-18 8:35 ` [PATCH 38/71] ncr5380: Remove UNSAFE macro Finn Thain
2015-11-18 8:35 ` [PATCH 39/71] ncr5380: Standardize interrupt handling Finn Thain
2015-11-18 8:35 ` [PATCH 40/71] ncr5380: Introduce NCR5380_poll_politely2 Finn Thain
2015-11-18 8:35 ` [PATCH 41/71] ncr5380: Replace redundant flags with FLAG_NO_DMA_FIXUP Finn Thain
2015-11-18 8:35 ` [PATCH 42/71] ncr5380: Replace READ_OVERRUNS macro with FLAG_NO_DMA_FIXUPS Finn Thain
2015-11-18 8:35 ` [PATCH 43/71] ncr5380: Standardize reselection handling Finn Thain
2015-11-18 8:35 ` [PATCH 44/71] ncr5380: Fix off-by-one bug in extended_msg[] bounds check Finn Thain
2015-11-18 8:35 ` [PATCH 45/71] ncr5380: Cleanup #include directives Finn Thain
2015-11-18 8:35 ` [PATCH 46/71] ncr5380: Fix NDEBUG_NO_DATAOUT flag Finn Thain
2015-11-18 8:35 ` [PATCH 47/71] ncr5380: Fix and cleanup scsi_host_template initializers Finn Thain
2015-11-18 8:35 ` [PATCH 48/71] atari_NCR5380: Fix queue_size limit Finn Thain
2015-11-18 8:35 ` [PATCH 49/71] atari_NCR5380: Introduce FLAG_DTC3181E Finn Thain
2015-11-18 8:35 ` [PATCH 50/71] ncr5380: Change instance->host_lock to hostdata->lock Finn Thain
2015-11-18 8:35 ` [PATCH 51/71] ncr5380: Remove command list debug code Finn Thain
2015-11-18 8:35 ` [PATCH 52/71] ncr5380: Remove H_NO macro and introduce dsprintk Finn Thain
2015-11-18 8:35 ` [PATCH 53/71] ncr5380: Use shost_priv helper Finn Thain
2015-11-18 8:35 ` [PATCH 54/71] ncr5380: Use dsprintk() for queue debugging Finn Thain
2015-11-18 8:35 ` [PATCH 55/71] ncr5380: Remove LIST and REMOVE macros Finn Thain
2015-11-18 8:35 ` [PATCH 56/71] ncr5380: Remove redundant volatile qualifiers Finn Thain
2015-11-18 8:35 ` [PATCH 57/71] ncr5380: Use standard list data structure Finn Thain
2015-11-18 8:35 ` [PATCH 58/71] ncr5380: Refactor command completion Finn Thain
2015-11-18 8:35 ` [PATCH 59/71] ncr5380: Fix autosense bugs Finn Thain
2015-11-18 8:35 ` [PATCH 60/71] ncr5380: Implement new eh_abort_handler Finn Thain
2015-11-18 8:35 ` [PATCH 61/71] ncr5380: Fix EH during arbitration and selection Finn Thain
2015-11-18 8:35 ` [PATCH 62/71] ncr5380: Implement new eh_bus_reset_handler Finn Thain
2015-11-18 8:35 ` [PATCH 63/71] atari_NCR5380: Remove HOSTNO macro from printk() and seq_printf() calls Finn Thain
2015-11-18 8:35 ` [PATCH 64/71] atari_NCR5380: Eliminate HOSTNO macro Finn Thain
2015-11-18 8:36 ` [PATCH 65/71] atari_scsi, sun3_scsi: Remove global Scsi_Host pointer Finn Thain
2015-11-18 8:36 ` [PATCH 66/71] ncr5380: Fix soft lockups Finn Thain
2015-11-18 8:36 ` [PATCH 67/71] ncr5380: Cleanup comments Finn Thain
2015-11-18 8:36 ` [PATCH 68/71] ncr5380: Fix whitespace issues using regexp Finn Thain
2015-11-18 8:36 ` [PATCH 69/71] ncr5380: Merge changes from atari_NCR5380.c Finn Thain
2015-11-18 8:36 ` [PATCH 70/71] atari_NCR5380: Merge changes from NCR5380.c Finn Thain
2015-11-18 8:36 ` [PATCH 71/71] ncr5380: Cleanup whitespace and parentheses Finn Thain
2015-11-18 11:35 ` [PATCH 00/71] More fixes, cleanup and modernization for NCR5380 drivers Ondrej Zary
2015-11-19 2:24 ` Finn Thain
2015-11-19 2:52 ` Michael Schmitz
2015-11-19 7:48 ` Ondrej Zary
2015-11-19 22:50 ` Ondrej Zary
2015-11-20 1:41 ` Finn Thain
2015-11-20 7:21 ` Finn Thain
2015-11-20 7:33 ` Christoph Hellwig
2015-11-20 8:19 ` Finn Thain
2015-11-20 9:16 ` Ondrej Zary
2015-11-20 10:00 ` Christoph Hellwig
2015-11-20 10:53 ` Finn Thain
2015-11-20 11:40 ` Ondrej Zary
2015-11-20 11:45 ` Christoph Hellwig
2015-11-20 12:23 ` Geert Uytterhoeven
2015-11-20 12:46 ` Ondrej Zary
2015-11-20 7:35 ` Ondrej Zary
2015-11-20 18:29 ` Ondrej Zary
2015-11-21 1:58 ` Finn Thain
2015-11-21 13:01 ` Ondrej Zary
2015-11-21 23:07 ` Ondrej Zary
2015-11-21 23:32 ` Finn Thain
2015-11-23 22:55 ` Ondrej Zary
2015-11-24 1:21 ` Finn Thain
2015-11-24 8:04 ` Ondrej Zary
2015-11-24 9:13 ` Finn Thain
2015-11-24 12:03 ` Ondrej Zary
2015-11-24 18:04 ` Ondrej Zary
2015-11-24 21:40 ` Ondrej Zary
2015-11-25 2:10 ` Finn Thain
2015-11-25 9:04 ` Ondrej Zary
2015-11-25 11:50 ` Finn Thain
2015-11-25 23:01 ` Ondrej Zary
2015-11-25 21:34 ` [PATCH 72/71] ncr5380: Fix pseudo-DMA Ondrej Zary
2015-11-29 9:39 ` [RFC PATCH 73/71] ncr5380: Use runtime register mapping Ondrej Zary
2015-11-30 11:50 ` Finn Thain
2015-11-29 9:39 ` [RFC PATCH 74/71] ncr5380: Enable PDMA for NCR53C400A Ondrej Zary
2015-11-30 11:52 ` Finn Thain
2015-11-30 12:04 ` Finn Thain
2015-11-30 13:40 ` Ondrej Zary
2015-12-06 3:20 ` Finn Thain
2015-11-29 10:01 ` [RFC PATCH 75/71] ncr5380: Remove FLAG_DTC3181E Ondrej Zary
2015-11-30 4:56 ` Finn Thain
2015-12-03 23:03 ` [RFC PATCH 76/71] ncr5380: Enable PDMA for DTC chips Ondrej Zary
2015-12-04 0:12 ` Julian Calaby
2015-12-04 8:38 ` Finn Thain
2015-12-05 1:32 ` Julian Calaby
2015-12-05 2:12 ` Finn Thain
2015-12-05 2:38 ` Julian Calaby
2015-12-04 9:08 ` Finn Thain
2015-12-04 9:32 ` Ondrej Zary
2015-12-04 9:20 ` Finn Thain
2015-12-03 23:03 ` [RFC PATCH 77/71] ncr5380: Fix wait for 53C80 registers registers after PDMA Ondrej Zary
2015-12-04 21:00 ` [PATCH 73/71] ncr5380: Use runtime register mapping Ondrej Zary
2015-12-06 3:39 ` Finn Thain
2015-12-06 8:42 ` Geert Uytterhoeven
2015-12-04 21:02 ` [PATCH 74/71] ncr5380: Enable PDMA for NCR53C400A Ondrej Zary
2015-12-06 3:41 ` Finn Thain
2015-12-04 21:17 ` [PATCH 76/71] ncr5380: Enable PDMA for DTC chips Ondrej Zary
2015-12-06 3:40 ` Finn Thain
2015-12-06 22:47 ` Ondrej Zary
2015-12-05 21:18 ` [RFC PATCH 78/71] ncr5380: Add support for HP 53C400A-based cards (C2502) Ondrej Zary
2015-12-06 3:39 ` Finn Thain
2015-12-06 22:20 ` [PATCH v2 73/71] ncr5380: Use runtime register mapping Ondrej Zary
2015-12-06 22:21 ` [PATCH v2 74/71] ncr5380: Enable PDMA for NCR53C400A Ondrej Zary
2015-12-08 2:05 ` Finn Thain
2015-12-09 13:39 ` Ondrej Zary
2015-12-10 23:39 ` Finn Thain
2015-12-06 22:55 ` [PATCH v2 76/71] ncr5380: Enable PDMA for DTC chips Ondrej Zary
2015-12-06 23:17 ` [PATCH v2 77/71] ncr5380: Fix wait for 53C80 registers registers after PDMA Ondrej Zary
2015-12-07 3:16 ` Finn Thain
2015-12-07 8:08 ` Ondrej Zary
2015-12-06 23:20 ` [PATCH v3 76/71] ncr5380: Enable PDMA for DTC chips Ondrej Zary
2015-12-08 7:51 ` [PATCH v2 78/71] ncr5380: Add support for HP 53C400A-based cards (C2502) Ondrej Zary
2015-12-08 11:40 ` Finn Thain
2015-12-09 11:33 ` Ondrej Zary
2015-12-10 23:38 ` Finn Thain
2015-12-09 12:11 ` [PATCH v3 " Ondrej Zary
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20151118083505.931718684@telegraphics.com.au \
--to=fthain@telegraphics.com.au \
--cc=JBottomley@odin.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-m68k@vger.kernel.org \
--cc=linux-scsi@vger.kernel.org \
--cc=schmitzmic@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox