From: Ben Hutchings <bhutchings@solarflare.com>
To: David Miller <davem@davemloft.net>
Cc: <netdev@vger.kernel.org>, <linux-net-drivers@solarflare.com>
Subject: [PATCH net-next 31/32] sfc: Fix race in completion handling
Date: Thu, 22 Aug 2013 14:10:12 +0100 [thread overview]
Message-ID: <1377177012.1703.40.camel@bwh-desktop.uk.level5networks.com> (raw)
In-Reply-To: <1377175392.1703.5.camel@bwh-desktop.uk.level5networks.com>
When we poll for MCDI request completion, we don't hold the interface
lock while setting the response fields in struct efx_mcdi_iface.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/mcdi.c | 25 ++++++++++---------------
1 file changed, 10 insertions(+), 15 deletions(-)
diff --git a/drivers/net/ethernet/sfc/mcdi.c b/drivers/net/ethernet/sfc/mcdi.c
index 2c5ee89..1c8bf81 100644
--- a/drivers/net/ethernet/sfc/mcdi.c
+++ b/drivers/net/ethernet/sfc/mcdi.c
@@ -112,16 +112,6 @@ static void efx_mcdi_copyin(struct efx_nic *efx, unsigned cmd,
efx->type->mcdi_request(efx, hdr, hdr_len, inbuf, inlen);
}
-static void
-efx_mcdi_copyout(struct efx_nic *efx, efx_dword_t *outbuf, size_t outlen)
-{
- struct efx_mcdi_iface *mcdi = efx_mcdi(efx);
-
- BUG_ON(atomic_read(&mcdi->state) == MCDI_STATE_QUIESCENT);
-
- efx->type->mcdi_read_response(efx, outbuf, mcdi->resp_hdr_len, outlen);
-}
-
static int efx_mcdi_errno(unsigned int mcdi_err)
{
switch (mcdi_err) {
@@ -200,9 +190,11 @@ static int efx_mcdi_poll(struct efx_nic *efx)
/* Check for a reboot atomically with respect to efx_mcdi_copyout() */
rc = efx_mcdi_poll_reboot(efx);
if (rc) {
+ spin_lock_bh(&mcdi->iface_lock);
mcdi->resprc = rc;
mcdi->resp_hdr_len = 0;
mcdi->resp_data_len = 0;
+ spin_unlock_bh(&mcdi->iface_lock);
return 0;
}
@@ -231,7 +223,9 @@ static int efx_mcdi_poll(struct efx_nic *efx)
return -ETIMEDOUT;
}
+ spin_lock_bh(&mcdi->iface_lock);
efx_mcdi_read_response_header(efx);
+ spin_unlock_bh(&mcdi->iface_lock);
/* Return rc=0 like wait_event_timeout() */
return 0;
@@ -419,7 +413,7 @@ int efx_mcdi_rpc_finish(struct efx_nic *efx, unsigned cmd, size_t inlen,
"MC command 0x%x inlen %d mode %d timed out\n",
cmd, (int)inlen, mcdi->mode);
} else {
- size_t resplen;
+ size_t hdr_len, data_len;
/* At the very least we need a memory barrier here to ensure
* we pick up changes from efx_mcdi_ev_cpl(). Protect against
@@ -427,16 +421,17 @@ int efx_mcdi_rpc_finish(struct efx_nic *efx, unsigned cmd, size_t inlen,
* acquiring the iface_lock. */
spin_lock_bh(&mcdi->iface_lock);
rc = mcdi->resprc;
- resplen = mcdi->resp_data_len;
+ hdr_len = mcdi->resp_hdr_len;
+ data_len = mcdi->resp_data_len;
spin_unlock_bh(&mcdi->iface_lock);
BUG_ON(rc > 0);
if (rc == 0) {
- efx_mcdi_copyout(efx, outbuf,
- min(outlen, mcdi->resp_data_len));
+ efx->type->mcdi_read_response(efx, outbuf, hdr_len,
+ min(outlen, data_len));
if (outlen_actual != NULL)
- *outlen_actual = resplen;
+ *outlen_actual = data_len;
} else if (cmd == MC_CMD_REBOOT && rc == -EIO)
; /* Don't reset if MC_CMD_REBOOT returns EIO */
else if (rc == -EIO || rc == -EINTR) {
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
next prev parent reply other threads:[~2013-08-22 13:10 UTC|newest]
Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-08-22 12:43 Pull request: sfc-next 2013-08-22 Ben Hutchings
2013-08-22 12:44 ` [PATCH net-next 01/32] sfc: Fix lookup of default RX MAC filters when steered using ethtool Ben Hutchings
2013-08-22 12:44 ` [PATCH net-next 02/32] sfc: const-qualify source pointers for MMIO write functions Ben Hutchings
2013-08-22 12:45 ` [PATCH net-next 03/32] sfc: Use efx_mcdi_mon() to find efx_mcdi_mon structure from efx_nic Ben Hutchings
2013-08-22 12:45 ` [PATCH net-next 04/32] sfc: Move details of a Falcon bug workaround out of ethtool.c Ben Hutchings
2013-08-22 12:46 ` [PATCH net-next 05/32] sfc: Move more Falcon-specific code and definitions into falcon.c Ben Hutchings
2013-08-22 12:46 ` [PATCH net-next 06/32] sfc: Introduce and use MCDI_DECLARE_BUF macro Ben Hutchings
2013-08-22 12:47 ` [PATCH net-next 07/32] sfc: Rationalise MCDI buffer accessors Ben Hutchings
2013-08-22 12:48 ` [PATCH net-next 08/32] sfc: Fill out the set of MCDI accessors Ben Hutchings
2013-08-22 12:48 ` [PATCH net-next 09/32] sfc: Introduce and use MCDI_CTL_SDU_LEN_MAX_V1 macro for Siena-specific code Ben Hutchings
2013-08-22 12:49 ` [PATCH net-next 10/32] sfc: Use proper macros to declare and access MCDI arrays Ben Hutchings
2013-08-22 12:49 ` [PATCH net-next 11/32] sfc: Ensure MCDI buffers, but not lengths, are dword aligned Ben Hutchings
2013-08-22 12:50 ` [PATCH net-next 12/32] sfc: Add and use MCDI_SET_QWORD() and MCDI_SET_ARRAY_QWORD() Ben Hutchings
2013-08-22 12:50 ` [PATCH net-next 13/32] sfc: Move siena_reset_hw() and siena_map_reset_reason() into MCDI module Ben Hutchings
2013-08-22 12:50 ` [PATCH net-next 14/32] sfc: Move efx_mcdi_mac_reconfigure() to siena.c and rename Ben Hutchings
2013-08-22 12:52 ` [PATCH net-next 15/32] sfc: Collect all MCDI port functions into mcdi_port.c Ben Hutchings
2013-08-22 12:53 ` [PATCH net-next 16/32] sfc: Make efx_mcdi_init() call efx_mcdi_handle_assertion() Ben Hutchings
2013-08-22 12:54 ` [PATCH net-next 17/32] sfc: Make MCDI independent of Siena Ben Hutchings
2013-08-22 12:54 ` [PATCH net-next 18/32] sfc: Add GFP flags to efx_nic_alloc_buffer() and make most callers allow blocking Ben Hutchings
2013-08-22 12:55 ` [PATCH net-next 19/32] sfc: Make struct efx_special_buffer less special Ben Hutchings
2013-08-22 12:55 ` [PATCH net-next 20/32] sfc: Rename Falcon-architecture register definitions Ben Hutchings
2013-08-22 12:56 ` [PATCH net-next 21/32] sfc: Remove efx_process_channel_now() Ben Hutchings
2013-08-22 12:57 ` [PATCH net-next 22/32] sfc: Rework IRQ enable/disable Ben Hutchings
2013-08-22 12:59 ` [PATCH net-next 23/32] sfc: Limit scope of a Falcon A1 IRQ workaround Ben Hutchings
2013-08-22 13:00 ` [PATCH net-next 24/32] sfc: Stop RX refill before flushing RX queues Ben Hutchings
2013-08-22 13:00 ` [PATCH net-next 25/32] sfc: Remove bogus call to efx_release_tx_buffers() Ben Hutchings
2013-08-22 13:02 ` [PATCH net-next 26/32] sfc: Refactor queue teardown sequence to allow for EF10 flush behaviour Ben Hutchings
2013-08-22 13:05 ` [PATCH net-next 27/32] sfc: Move and rename Falcon/Siena common NIC operations Ben Hutchings
2013-08-22 13:05 ` [PATCH net-next 28/32] sfc: Translate MCDI error numbers received in events Ben Hutchings
2013-08-22 13:07 ` [PATCH net-next 29/32] sfc: Update MCDI protocol definitions for EF10 Ben Hutchings
2013-08-22 13:09 ` [PATCH net-next 30/32] sfc: Add support for MCDI v2 Ben Hutchings
2013-08-22 13:10 ` Ben Hutchings [this message]
2013-08-22 13:10 ` [PATCH net-next 32/32] sfc: Update and improve kernel-doc for efx_mcdi_state & efx_mcdi_iface Ben Hutchings
2013-08-22 19:09 ` Pull request: sfc-next 2013-08-22 David Miller
2013-08-22 19:16 ` Ben Hutchings
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=1377177012.1703.40.camel@bwh-desktop.uk.level5networks.com \
--to=bhutchings@solarflare.com \
--cc=davem@davemloft.net \
--cc=linux-net-drivers@solarflare.com \
--cc=netdev@vger.kernel.org \
/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