From: Finn Thain <fthain@telegraphics.com.au>
To: Hannes Reinecke <hare@suse.de>
Cc: "James E.J. Bottomley" <JBottomley@parallels.com>,
Michael Schmitz <schmitzmic@gmail.com>,
Sam Creasey <sammy@sammy.net>,
linux-scsi@vger.kernel.org, linux-m68k@vger.kernel.org
Subject: Re: [PATCH v2 29/36] atari_NCR5380: Refactor Falcon locking
Date: Fri, 31 Oct 2014 18:20:09 +1100 (AEDT) [thread overview]
Message-ID: <alpine.LNX.2.00.1410311523590.7380@nippy.intranet> (raw)
In-Reply-To: <5451F67A.2000708@suse.de>
On Thu, 30 Oct 2014, Hannes Reinecke wrote:
> On 10/27/2014 06:26 AM, Finn Thain wrote:
> > Simplify falcon_release_lock_if_possible() by making callers
> > responsible for disabling local IRQ's, which they must do anyway to
> > correctly synchronize the ST DMA "lock" with core driver data
> > structures. Move this synchronization logic to the core driver with
> > which it is tightly coupled.
> >
> > Other LLD's like sun3_scsi and mac_scsi that can make use of this core
> > driver can just stub out the NCR5380_acquire_dma_irq() and
> > NCR5380_release_dma_irq() calls so the compiler will eliminate the ST
> > DMA code.
> >
> > Remove a redundant local_irq_save/restore pair (irq's are disabled for
> > interrupt handlers these days). Revise the locking for
> > atari_scsi_bus_reset(): use local_irq_save/restore() instead of
> > atari_turnoff/turnon_irq(). There is no guarantee that atari_scsi
> > still holds the ST DMA lock during EH, so atari_turnoff/turnon_irq()
> > could end up dropping an IDE or floppy interrupt.
> >
> > Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
> >
> > ---
> > drivers/scsi/atari_NCR5380.c | 72 ++++++++++++++++++++++++++-----------------
> > drivers/scsi/atari_scsi.c | 47 ++++++++++------------------
> > 2 files changed, 62 insertions(+), 57 deletions(-)
> >
> > Index: linux/drivers/scsi/atari_NCR5380.c
> > ===================================================================
> > --- linux.orig/drivers/scsi/atari_NCR5380.c 2014-10-27 16:25:54.000000000 +1100
> > +++ linux/drivers/scsi/atari_NCR5380.c 2014-10-27 16:25:55.000000000 +1100
> > @@ -926,7 +926,7 @@ static int NCR5380_queue_command(struct
> > * alter queues and touch the lock.
> > */
> > /* perhaps stop command timer here */
> > - if (!falcon_get_lock())
> > + if (!NCR5380_acquire_dma_irq(instance))
> > return SCSI_MLQUEUE_HOST_BUSY;
> > /* perhaps restart command timer here */
> >
> > @@ -962,6 +962,18 @@ static int NCR5380_queue_command(struct
> > return 0;
> > }
> >
> > +static inline void maybe_release_dma_irq(struct Scsi_Host *instance)
> > +{
> > + struct NCR5380_hostdata *hostdata = shost_priv(instance);
> > +
> > + /* Caller does the locking needed to set & test these data atomically */
> > + if (!hostdata->disconnected_queue &&
> > + !hostdata->issue_queue &&
> > + !hostdata->connected &&
> > + !hostdata->retain_dma_intr)
> > + NCR5380_release_dma_irq(instance);
> > +}
> > +
> > /*
> > * Function : NCR5380_main (void)
> > *
> > @@ -1084,9 +1096,11 @@ static void NCR5380_main(struct work_str
> > cmd_get_tag(tmp, tmp->cmnd[0] != REQUEST_SENSE);
> > #endif
> > if (!NCR5380_select(instance, tmp)) {
> > + local_irq_disable();
> > hostdata->retain_dma_intr--;
> > /* release if target did not response! */
> > - falcon_release_lock_if_possible(hostdata);
> > + maybe_release_dma_irq(instance);
> > + local_irq_restore(flags);
> > break;
> > } else {
> > local_irq_disable();
> > @@ -2085,11 +2099,12 @@ static void NCR5380_information_transfer
> > case COMMAND_COMPLETE:
> > /* Accept message by clearing ACK */
> > NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
> > - /* ++guenther: possible race with Falcon locking */
> > - hostdata->retain_dma_intr++;
> > - hostdata->connected = NULL;
> > dprintk(NDEBUG_QUEUES, "scsi%d: command for target %d, lun %llu "
> > "completed\n", HOSTNO, cmd->device->id, cmd->device->lun);
> > +
> > + local_irq_save(flags);
> > + hostdata->retain_dma_intr++;
> > + hostdata->connected = NULL;
> > #ifdef SUPPORT_TAGS
> > cmd_free_tag(cmd);
> > if (status_byte(cmd->SCp.Status) == QUEUE_FULL) {
> > @@ -2148,17 +2163,17 @@ static void NCR5380_information_transfer
> >
> > dprintk(NDEBUG_AUTOSENSE, "scsi%d: performing request sense\n", HOSTNO);
> >
> > - local_irq_save(flags);
> > LIST(cmd,hostdata->issue_queue);
> > SET_NEXT(cmd, hostdata->issue_queue);
> > hostdata->issue_queue = (struct scsi_cmnd *) cmd;
> > - local_irq_restore(flags);
> > dprintk(NDEBUG_QUEUES, "scsi%d: REQUEST SENSE added to head of "
> > "issue queue\n", H_NO(cmd));
> > } else {
> > cmd->scsi_done(cmd);
> > }
> >
> > + local_irq_restore(flags);
> > +
> > NCR5380_write(SELECT_ENABLE_REG, hostdata->id_mask);
> > /*
> > * Restore phase bits to 0 so an interrupted selection,
> Probably a stupid question, but you're disabling local interrupts.
> scsi_done() OTOH is triggering a soft irq.
Isn't that how all bottom halves work?
> Does this work?
Yes, I've tested that particular code. I'm pretty sure Michael has too.
A quick search turned up another example of this in
zfcp_fsf_fcp_cmnd_handler().
> Shouldn't we enable interrupts before calling scsi_done() ?
I don't see anything in the scsi_done() implementation in scsi_lib.c that
would be problematic. The description says "This function is interrupt
context safe". The same would have to apply to scsi_eh_done().
--
next prev parent reply other threads:[~2014-10-31 7:20 UTC|newest]
Thread overview: 100+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-10-27 5:26 [PATCH v2 00/36] Fixes, cleanups and modernization for NCR5380 drivers Finn Thain
2014-10-27 5:26 ` [PATCH v2 01/36] ncr5380: Use printk() not pr_debug() Finn Thain
2014-10-27 5:26 ` [PATCH v2 02/36] ncr5380: Remove unused hostdata fields Finn Thain
2014-10-27 5:26 ` [PATCH v2 03/36] ncr5380: Fix compiler warnings and __setup options Finn Thain
2014-10-27 5:26 ` [PATCH v2 04/36] ncr5380: Remove unused macros Finn Thain
2014-10-27 5:26 ` [PATCH v2 05/36] ncr5380: Remove useless prototypes Finn Thain
2014-10-29 14:41 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 06/36] ncr5380: Remove more " Finn Thain
2014-10-29 14:44 ` Hannes Reinecke
2014-11-09 12:19 ` Finn Thain
2014-11-11 12:47 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 07/36] ncr5380: Cleanup TAG_NEXT and TAG_NONE macros Finn Thain
2014-10-29 14:45 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 08/36] ncr5380: Remove redundant AUTOSENSE macro Finn Thain
2014-10-29 15:57 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 09/36] ncr5380: Remove duplicate comments Finn Thain
2014-10-29 15:59 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 10/36] ncr5380: Fix SCSI_IRQ_NONE bugs Finn Thain
2014-10-29 16:07 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 11/36] ncr5380: Remove NCR5380_STATS Finn Thain
2014-10-29 16:11 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 12/36] ncr5380: Cleanup host info() methods Finn Thain
2014-10-29 16:17 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 13/36] ncr5380: Move static PDMA spin counters to host data Finn Thain
2014-10-30 7:31 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 14/36] ncr5380: Remove pointless compiler command line override macros Finn Thain
2014-10-30 7:34 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 15/36] ncr5380: Remove *_RELEASE macros Finn Thain
2014-10-30 7:36 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 16/36] ncr5380: Drop legacy scsi.h include Finn Thain
2014-10-30 7:37 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 17/36] dmx3191d: Use NO_IRQ Finn Thain
2014-10-30 7:38 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 18/36] mac_scsi: Remove header Finn Thain
2014-10-30 7:39 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 19/36] mac_scsi: Add module option to Kconfig Finn Thain
2014-10-30 7:44 ` Hannes Reinecke
2014-10-31 7:17 ` Finn Thain
2014-10-31 8:33 ` Hannes Reinecke
2014-11-09 12:17 ` Finn Thain
2014-11-10 7:02 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 20/36] mac_scsi: Cleanup PDMA code Finn Thain
2014-10-30 7:46 ` Hannes Reinecke
2014-10-30 8:45 ` Hannes Reinecke
2014-10-31 7:18 ` Finn Thain
2014-10-31 8:34 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 21/36] mac_scsi: Convert to platform device Finn Thain
2014-10-30 7:55 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 22/36] atari_scsi: Fix atari_scsi deadlocks on Falcon Finn Thain
2014-10-30 8:07 ` Hannes Reinecke
2014-11-07 18:08 ` Michael Schmitz
2014-11-08 0:37 ` Finn Thain
2014-11-08 7:22 ` Michael Schmitz
2014-10-27 5:26 ` [PATCH v2 23/36] atari_scsi: Convert to platform device Finn Thain
2014-10-30 8:17 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 24/36] atari_scsi: Remove header Finn Thain
2014-10-30 8:18 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 25/36] sun3_scsi: Convert to platform device Finn Thain
2014-10-30 8:20 ` Hannes Reinecke
2014-11-09 10:25 ` Geert Uytterhoeven
2014-11-09 12:12 ` Finn Thain
2014-11-09 12:18 ` Geert Uytterhoeven
2014-10-27 5:26 ` [PATCH v2 26/36] sun3_scsi: Move macro definitions Finn Thain
2014-10-30 8:20 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 27/36] ncr5380: Remove ENABLE_IRQ/DISABLE_IRQ macros Finn Thain
2014-10-30 8:21 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 28/36] atari_NCR5380: Refactor Falcon special cases Finn Thain
2014-10-30 8:23 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 29/36] atari_NCR5380: Refactor Falcon locking Finn Thain
2014-10-30 8:27 ` Hannes Reinecke
2014-10-31 7:20 ` Finn Thain [this message]
2014-11-09 12:18 ` Finn Thain
2014-11-10 7:03 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 30/36] atari_NCR5380: Merge from sun3_NCR5380.c Finn Thain
2014-10-30 8:33 ` Hannes Reinecke
2014-10-31 7:21 ` Finn Thain
2014-10-31 8:40 ` Hannes Reinecke
2014-10-31 12:48 ` Finn Thain
2014-10-27 5:26 ` [PATCH v2 31/36] sun3_scsi: Adopt atari_NCR5380 core driver and remove sun3_NCR5380.c Finn Thain
2014-10-30 8:35 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 32/36] atari_NCR5380: Merge from NCR5380.c Finn Thain
2014-10-30 8:37 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 33/36] atari_NCR5380: Introduce FLAG_TAGGED_QUEUING Finn Thain
2014-10-30 8:39 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 34/36] atari_NCR5380: Move static TagAlloc array to host data Finn Thain
2014-10-30 8:41 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 35/36] atari_NCR5380: Move static co-routine variables " Finn Thain
2014-10-30 8:42 ` Hannes Reinecke
2014-10-27 5:26 ` [PATCH v2 36/36] atari_NCR5380: Remove RESET_RUN_DONE macro Finn Thain
2014-10-30 8:44 ` Hannes Reinecke
2014-11-02 5:28 ` [PATCH v2 00/36] Fixes, cleanups and modernization for NCR5380 drivers Michael Schmitz
2014-11-02 6:12 ` Finn Thain
[not found] ` <5455E340.7080305@gmail.com>
2014-11-04 23:36 ` Michael Schmitz
2014-11-05 7:56 ` David Gálvez
2014-11-05 8:10 ` Geert Uytterhoeven
2014-11-06 2:09 ` Michael Schmitz
2014-11-06 4:50 ` Finn Thain
2014-11-06 18:54 ` Michael Schmitz
2014-11-07 2:34 ` Finn Thain
2014-11-09 10:33 ` Geert Uytterhoeven
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=alpine.LNX.2.00.1410311523590.7380@nippy.intranet \
--to=fthain@telegraphics.com.au \
--cc=JBottomley@parallels.com \
--cc=hare@suse.de \
--cc=linux-m68k@vger.kernel.org \
--cc=linux-scsi@vger.kernel.org \
--cc=sammy@sammy.net \
--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