diff options
| author | Venugopal Iyer <Venu.Iyer@Sun.COM> | 2009-02-17 01:31:30 -0800 |
|---|---|---|
| committer | Venugopal Iyer <Venu.Iyer@Sun.COM> | 2009-02-17 01:31:30 -0800 |
| commit | ae6aa22afeb444ae208c287e7227a4a7c877f17a (patch) | |
| tree | 744dffd8856e6a2a16544575ca8773771051dada /usr/src/uts/common/io/softmac | |
| parent | d02310705313ee2fcefee164a4b26d1fa85e9d22 (diff) | |
| download | illumos-gate-ae6aa22afeb444ae208c287e7227a4a7c877f17a.tar.gz | |
PSARC/2009/099 dladm show-usage modifications
6726676 flow should not be seen by flowadm show-usage after the flow been removed by flowadm remove-flow
6766669 "dladm show-vnic -o" can't accept MACADDRESS
6773854 Per Tx ring flow control for UDP
6777547 mac_tx() should compute the hash if the passed hint is zero
6778557 nxge m_tx() should fanout to multiple rings for vnet scalability
6779356 sometimes packets are not classified to the correct flow
6783011 pre-existing subflows not initialized on a non-dls client when brought up
6786734 acctadm dladm_start_usagelog() calls need some work
6789760 mac perimeter deadlock due to dls_devnet_stat_update()
6789883 dladm show-link -s is adrift again.
6791099 mac_tx() frees the message but returns non-NULL cookie which causes panic
6791109 maxbw set on a link should not apply if this link is the underlying port of an aggregation
6791118 panic in mac_bcast_delete() unplumbing an IP interface
6791456 deleting last vnic interface causes bge interface to stop working
6791678 xvm guests don't communicate through vnics configured on vlan
6792164 race between mac_tx_is_flow_blocked() and mac_srs_group_teardown() could cause panic
6792546 paniced in bge_ring_tx()/freemsg() due to mp->b_next == NULL && mp->b_prev == NULL
6792555 paniced in mac_flow_walk_nolock() due to assertion failed: cnt == ft->ft_flow_count
6792871 multiple VLANs per MAC client cause hang in mac_flow_wait()
6792942 60% regression for Guest-to-Guest network throughput on snv106
6793278 the multicast addresses are not added to the aggregation port in certain scenarios
6793436 panic in mac_fini_macaddr() on mac_register() failure
6796850 SUNWcnetr postinstall script spews errors due to bad interface matching
6803378 need support for dls_bypass and rx fanout on non-ethernet media
Diffstat (limited to 'usr/src/uts/common/io/softmac')
| -rw-r--r-- | usr/src/uts/common/io/softmac/softmac_main.c | 16 |
1 files changed, 14 insertions, 2 deletions
diff --git a/usr/src/uts/common/io/softmac/softmac_main.c b/usr/src/uts/common/io/softmac/softmac_main.c index c940794d72..a44856c849 100644 --- a/usr/src/uts/common/io/softmac/softmac_main.c +++ b/usr/src/uts/common/io/softmac/softmac_main.c @@ -1042,17 +1042,26 @@ softmac_mac_recreate(mod_hash_key_t key, mod_hash_val_t *val, void *arg) return (MH_WALK_CONTINUE); } + /* + * Bumping up the smac_hold_cnt allows us to drop the lock. It also + * makes softmac_destroy() return failure on an attempted device detach. + * We don't want to hold the lock across calls to other subsystems + * like kstats, which will happen in the call to dls_devnet_recreate + */ + softmac->smac_hold_cnt++; + mutex_exit(&softmac->smac_mutex); + if (dls_mgmt_create(softmac->smac_devname, makedevice(softmac->smac_umajor, softmac->smac_uppa + 1), DATALINK_CLASS_PHYS, softmac->smac_media, B_TRUE, &linkid) != 0) { - mutex_exit(&softmac->smac_mutex); + softmac_rele_device((dls_dev_handle_t)softmac); return (MH_WALK_CONTINUE); } if ((err = softmac_update_info(softmac, &linkid)) != 0) { cmn_err(CE_WARN, "softmac: softmac_update_info() for %s " "failed (%d)", softmac->smac_devname, err); - mutex_exit(&softmac->smac_mutex); + softmac_rele_device((dls_dev_handle_t)softmac); return (MH_WALK_CONTINUE); } @@ -1069,7 +1078,10 @@ softmac_mac_recreate(mod_hash_key_t key, mod_hash_val_t *val, void *arg) } } + mutex_enter(&softmac->smac_mutex); softmac->smac_flags &= ~SOFTMAC_NEED_RECREATE; + ASSERT(softmac->smac_hold_cnt != 0); + softmac->smac_hold_cnt--; mutex_exit(&softmac->smac_mutex); return (MH_WALK_CONTINUE); |
