From: Don Brace <don.brace@pmcs.com>
To: hch@infradead.org, webb.scales@hp.com,
james.bottomley@parallels.com, brace@pmcs.com
Cc: linux-scsi@vger.kernel.org
Subject: [PATCH 13/13] hpsa: remove spin lock around command allocation
Date: Fri, 14 Nov 2014 17:27:29 -0600 [thread overview]
Message-ID: <20141114232729.20808.72735.stgit@don-ProLiant-MicroServer-Gen8> (raw)
In-Reply-To: <20141114231145.20808.76898.stgit@don-ProLiant-MicroServer-Gen8>
From: Stephen M. Cameron <stephenmcameron@gmail.com>
It is already using atomic test_and_set_bit to do the
allocation.
There is some microscopic chance of starvation, but it is
so microscopic that it should never happen in reality.
Signed-off-by: Don Brace <don.brace@pmcs.com>
Reviewed-by: Webb Scales <webbnh@hp.com>
---
drivers/scsi/hpsa.c | 36 +++++++++++++++++++++++-------------
1 file changed, 23 insertions(+), 13 deletions(-)
diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c
index 72daca9..8a3d0de 100644
--- a/drivers/scsi/hpsa.c
+++ b/drivers/scsi/hpsa.c
@@ -4607,19 +4607,32 @@ static struct CommandList *cmd_alloc(struct ctlr_info *h)
int i;
union u64bit temp64;
dma_addr_t cmd_dma_handle, err_dma_handle;
- unsigned long flags;
+ int loopcount;
+
+ /* There is some *extremely* small but non-zero chance that that
+ * multiple threads could get in here, and one thread could
+ * be scanning through the list of bits looking for a free
+ * one, but the free ones are always behind him, and other
+ * threads sneak in behind him and eat them before he can
+ * get to them, so that while there is always a free one, a
+ * very unlucky thread might be starved anyway, never able to
+ * beat the other threads. In reality, this happens so
+ * infrequently as to be indistinguishable from never.
+ */
- spin_lock_irqsave(&h->lock, flags);
+ loopcount = 0;
do {
i = find_first_zero_bit(h->cmd_pool_bits, h->nr_cmds);
- if (i == h->nr_cmds) {
- spin_unlock_irqrestore(&h->lock, flags);
- return NULL;
- }
- } while (test_and_set_bit
- (i & (BITS_PER_LONG - 1),
- h->cmd_pool_bits + (i / BITS_PER_LONG)) != 0);
- spin_unlock_irqrestore(&h->lock, flags);
+ if (i == h->nr_cmds)
+ i = 0;
+ loopcount++;
+ } while (test_and_set_bit(i & (BITS_PER_LONG - 1),
+ h->cmd_pool_bits + (i / BITS_PER_LONG)) != 0 &&
+ loopcount < 10);
+
+ /* Thread got starved? We do not expect this to ever happen. */
+ if (loopcount >= 10)
+ return NULL;
c = h->cmd_pool + i;
memset(c, 0, sizeof(*c));
@@ -4679,13 +4692,10 @@ static struct CommandList *cmd_special_alloc(struct ctlr_info *h)
static void cmd_free(struct ctlr_info *h, struct CommandList *c)
{
int i;
- unsigned long flags;
i = c - h->cmd_pool;
- spin_lock_irqsave(&h->lock, flags);
clear_bit(i & (BITS_PER_LONG - 1),
h->cmd_pool_bits + (i / BITS_PER_LONG));
- spin_unlock_irqrestore(&h->lock, flags);
}
static void cmd_special_free(struct ctlr_info *h, struct CommandList *c)
next prev parent reply other threads:[~2014-11-14 23:25 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-11-14 23:26 [PATCH 00/13] hpsa update Don Brace
2014-11-14 23:26 ` [PATCH 01/13] hpsa: Clean up warnings from sparse Don Brace
2014-11-14 23:26 ` [PATCH 02/13] hpsa: remove dev_warn prints from RAID-1ADM Don Brace
2014-11-14 23:26 ` [PATCH 03/13] hpsa: fix a couple pci id table mistakes Don Brace
2014-11-14 23:26 ` [PATCH 04/13] hpsa: correct off-by-one sizing of chained SG block Don Brace
2014-11-14 23:26 ` [PATCH 05/13] hpsa: remove 'action required' phrasing Don Brace
2014-11-14 23:26 ` [PATCH 06/13] hpsa: fix allocation sizes for CISS_REPORT_LUNs commands Don Brace
2014-11-14 23:26 ` [PATCH 07/13] hpsa: fix endianness issue with scatter gather elements Don Brace
2014-11-14 23:27 ` [PATCH 08/13] hpsa: get rid of type/attribute/direction bit field where possible Don Brace
2014-11-14 23:27 ` [PATCH 09/13] hpsa: use atomics for commands_outstanding Don Brace
2014-11-14 23:27 ` [PATCH 10/13] hpsa: do not be so noisy about check conditions Don Brace
2014-11-14 23:27 ` [PATCH 11/13] hpsa: Convert SCSI LLD ->queuecommand() for host_lock less operation Don Brace
2014-11-14 23:27 ` [PATCH 12/13] hpsa: always call pci_set_master after pci_enable_device Don Brace
2014-11-14 23:27 ` Don Brace [this message]
2014-11-20 15:55 ` [PATCH 00/13] hpsa update Christoph Hellwig
2014-11-24 19:22 ` brace
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=20141114232729.20808.72735.stgit@don-ProLiant-MicroServer-Gen8 \
--to=don.brace@pmcs.com \
--cc=brace@pmcs.com \
--cc=hch@infradead.org \
--cc=james.bottomley@parallels.com \
--cc=linux-scsi@vger.kernel.org \
--cc=webb.scales@hp.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