ixgbe: fix build when DEBUG is defined
Subsystems (2)
PATCH 1/9 ixgbe: fix build when DEBUG is defined
From: Alexander Duyck <[email protected]>
The ixgbe driver had issues when DEBUG was defined because the hw_dbg macro was incomplete. This patch completes the code based off of the code that already existed in the igb module.
---
drivers/net/ixgbe/ixgbe_common.h | 9 +++------ drivers/net/ixgbe/ixgbe_main.c | 12 ++++++++++++ 2 files changed, 15 insertions(+), 6 deletions(-)
drivers/net/ixgbe/ixgbe_common.h +3 -6
@@ -96,14 +96,11 @@ s32 ixgbe_write_analog_reg8_generic(struct ixgbe_hw *hw, u32 reg, u8 val);#define IXGBE_WRITE_FLUSH (a) IXGBE_READ_REG(a, IXGBE_STATUS)#ifdef DEBUG +extern char *ixgbe_get_hw_dev_name (struct ixgbe_hw *hw); #define hw_dbg(hw, format, arg...) \ -printk(KERN_DEBUG, "%s: " format, ixgbe_get_hw_dev_name(hw), ##arg); + printk(KERN_DEBUG "%s: " format, ixgbe_get_hw_dev_name(hw), ##arg)#else -static inline int __attribute__ ((format (printf,2 ,3 ))) -hw_dbg (struct ixgbe_hw *hw,const char *format, ...) -{ -return 0 ; -} +#define hw_dbg(hw, format, arg...) do {} while (0)#endif #endif /* IXGBE_COMMON */
drivers/net/ixgbe/ixgbe_main.c +12 -0
@@ -4987,8 +4987,20 @@ static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event, return ret_val ? NOTIFY_BAD : NOTIFY_DONE; } +#endif /* CONFIG_IXGBE_DCA */ +#ifdef DEBUG +/** +* ixgbe_get_hw_dev_name - return device name string +* used by hardware layer to print debugging information +**/ +char *ixgbe_get_hw_dev_name (struct ixgbe_hw *hw) +{ +struct ixgbe_adapter *adapter = hw->back; +return adapter->netdev->name; +} +#endif module_exit (ixgbe_exit_module);/* ixgbe_main.c */
PATCH 2/9 ixgbe: Fix ethtool output with advertised mode.
From: Don Skidmore <[email protected]>
Ethtool tries to get advertised speed from phy.autoneg_advertised. However for copper media this wasn't happening until later do to an other fix which moved mac.ops.setup_link_speed placement in ixgbe_link_config(). This patch will display the default advertised speeds if it can't yet get this information from phy.autoneg_advertised.
---
drivers/net/ixgbe/ixgbe_ethtool.c | 9 +++++++++ 1 files changed, 9 insertions(+), 0 deletions(-)
drivers/net/ixgbe/ixgbe_ethtool.c +9 -0
@@ -129,6 +129,15 @@ static int ixgbe_get_settings(struct net_device *netdev, ecmd->advertising |= ADVERTISED_10000baseT_Full;if (hw->phy.autoneg_advertised &IXGBE_LINK_SPEED_1GB_FULL ) ecmd->advertising |= ADVERTISED_1000baseT_Full; +/* +* It's possible that phy.autoneg_advertised may not be +* set yet. If so display what the default would be - +* both 1G and 10G supported. +*/ +if (!(ecmd->advertising & (ADVERTISED_1000baseT_Full | + ADVERTISED_10000baseT_Full))) + ecmd->advertising |= (ADVERTISED_10000baseT_Full | + ADVERTISED_1000baseT_Full); ecmd->port =PORT_TP ; }else if (hw->phy.media_type == ixgbe_media_type_backplane) {
Re: [PATCH 2/9] ixgbe: Fix ethtool output with advertised mode.
From: Jeff Kirsher <[email protected]>
Date: Wed, 01 Apr 2009 00:33:02 -0700
Quoted 10 lines
Ethtool tries to get advertised speed from phy.autoneg_advertised. However for copper media this wasn't happening until later do to an other fix which moved mac.ops.setup_link_speed placement in
ixgbe_link_config(). This patch will display the default advertised speeds if it can't yet get this information from phy.autoneg_advertised.Signed-off-by Don Skidmore <[email protected]>Signed-off-by Jeff Kirsher <[email protected]>
Applied.
PATCH 3/9 ixgbe: Fix DCB netlink layer for 82599 to enable Priority Flow Control
In reply to Jeff KirsherFrom: PJ Waskiewicz <[email protected]>
The priority flow control settings from the netlink layer aren't taking effect in the base driver. The boolean pfc_mode_enable in the dcb_config struct isn't being set, so the hardware configuration code is never reached.
---
drivers/net/ixgbe/ixgbe_dcb_nl.c | 6 +++++- 1 files changed, 5 insertions(+), 1 deletions(-)
drivers/net/ixgbe/ixgbe_dcb_nl.c +5 -1
@@ -90,6 +90,8 @@ int ixgbe_copy_dcb_cfg(struct ixgbe_dcb_config *src_dcb_cfg, src_dcb_cfg->tc_config[i -DCB_PFC_UP_ATTR_0 ].dcb_pfc; } + dst_dcb_cfg->pfc_mode_enable = src_dcb_cfg->pfc_mode_enable; +return 0 ; }
@@ -298,8 +300,10 @@ static void ixgbe_dcbnl_set_pfc_cfg(struct net_device *netdev, int priority, adapter->temp_dcb_cfg.tc_config[priority].dcb_pfc = setting;if (adapter->temp_dcb_cfg.tc_config[priority].dcb_pfc != - adapter->dcb_cfg.tc_config[priority].dcb_pfc) + adapter->dcb_cfg.tc_config[priority].dcb_pfc) { adapter->dcb_set_bitmap |=BIT_PFC ; + adapter->temp_dcb_cfg.pfc_mode_enable = true; + } } static void ixgbe_dcbnl_get_pfc_cfg(struct net_device *netdev, int priority,
Re: [PATCH 3/9] ixgbe: Fix DCB netlink layer for 82599 to enable Priority Flow Control
From: Jeff Kirsher <[email protected]>
Date: Wed, 01 Apr 2009 00:33:25 -0700
Quoted 7 lines
The priority flow control settings from the netlink layer aren't taking effect in the base driver. The boolean
pfc_mode_enablein thedcb_configstruct isn't being set, so the hardware configuration code is never reached.Signed-off-by Peter P Waskiewicz Jr <[email protected]>Signed-off-by Jeff Kirsher <[email protected]>
Applied.
PATCH 4/9 ixgbe: feature - driver to default with FC on.
In reply to Jeff KirsherFrom: Don Skidmore <[email protected]>
In the past flow control wasn't enabled by default under the incorrect assumption that this opened up us to a denial of service attack. However since any switch that forwarded flow control would be extremely msiconfigured and/or buggy, this concern no longer out weighs the preformance gains from having FC enabled.
---
drivers/net/ixgbe/ixgbe_main.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/net/ixgbe/ixgbe_main.c +1 -1
@@ -3169,7 +3169,7 @@ static int __devinit ixgbe_sw_init(struct ixgbe_adapter *adapter)#endif /* default flow control settings */ - hw->fc.requested_mode = ixgbe_fc_none; + hw->fc.requested_mode = ixgbe_fc_full; hw->fc.high_water =IXGBE_DEFAULT_FCRTH ; hw->fc.low_water =IXGBE_DEFAULT_FCRTL ; hw->fc.pause_time =IXGBE_DEFAULT_FCPAUSE ;
Re: [PATCH 4/9] ixgbe: feature - driver to default with FC on.
From: Jeff Kirsher <[email protected]>
Date: Wed, 01 Apr 2009 00:33:44 -0700
Quoted 11 lines
In the past flow control wasn't enabled by default under the incorrect assumption that this opened up us to a denial of service attack. However since any switch that forwarded flow control would be extremely msiconfigured and/or buggy, this concern no longer out weighs the preformance gains from having FC enabled.
Signed-off-by Don Skidmore <[email protected]>Signed-off-by Jeff Kirsher <[email protected]>
Applied.
PATCH 5/9 ixgbe: Fix 82598 MSI-X allocation on systems with more than 8 CPU cores
In reply to Jeff KirsherFrom: PJ Waskiewicz <[email protected]>
MSI-X allocation broke after the 82599 merge on systems with more than 8 CPU cores. 82598 drops back into MSI mode, which isn't sufficient to run full, efficient 10G line rate.
---
drivers/net/ixgbe/ixgbe_main.c | 11 ++++++----- 1 files changed, 6 insertions(+), 5 deletions(-)
drivers/net/ixgbe/ixgbe_main.c +6 -5
@@ -2934,6 +2934,7 @@ err_tx_ring_allocation:**/ static int ixgbe_set_interrupt_capability (struct ixgbe_adapter *adapter) { +struct ixgbe_hw *hw = &adapter->hw;int err =0 ;int vector, v_budget;
@@ -2948,12 +2949,12 @@ static int ixgbe_set_interrupt_capability(struct ixgbe_adapter *adapter)/* * At the same time, hardware can only support a maximum of -* MAX_MSIX_COUNT vectors. With features such as RSS and VMDq, -* we can easily reach upwards of 64 Rx descriptor queues and -* 32 Tx queues. Thus, we cap it off in those rare cases where -* the cpu count also exceeds our vector limit. +* hw.mac->max_msix_vectors vectors. With features +* such as RSS and VMDq, we can easily surpass the number of Rx and Tx +* descriptor queues supported by our device. Thus, we cap it off in +* those rare cases where the cpu count also exceeds our vector limit. */ - v_budget =min (v_budget,MAX_MSIX_COUNT ); + v_budget =min (v_budget, (int )hw->mac.max_msix_vectors);/* A failure in MSI-X entry allocation isn't fatal, but it does * mean we disable MSI-X capabilities of the adapter. */
Re: [PATCH 5/9] ixgbe: Fix 82598 MSI-X allocation on systems with more than 8 CPU cores
From: Jeff Kirsher <[email protected]>
Date: Wed, 01 Apr 2009 00:34:05 -0700
Quoted 6 lines
MSI-X allocation broke after the 82599 merge on systems with more than 8 CPU cores. 82598 drops back into MSI mode, which isn't sufficient to run full, efficient 10G line rate.
Signed-off-by Peter P Waskiewicz Jr <[email protected]>Signed-off-by Jeff Kirsher <[email protected]>
Applied.
PATCH 6/9 ixgbe: refactor tx buffer processing to use skb_dma_map/unmap
In reply to Jeff KirsherFrom: Alexander Duyck <[email protected]>
This patch resolves an issue with map single being used to map a buffer and then unmap page being used to unmap it. In addition it handles any error conditions that may be detected using skb_dma_map.
---
drivers/net/ixgbe/ixgbe_main.c | 79 +++++++++++++++++++++------------------- 1 files changed, 41 insertions(+), 38 deletions(-)
drivers/net/ixgbe/ixgbe_main.c +41 -38
@@ -187,15 +187,14 @@ static void ixgbe_unmap_and_free_tx_resource(struct ixgbe_adapter *adapter,struct ixgbe_tx_buffer *tx_buffer_info) { -if (tx_buffer_info->dma) { -pci_unmap_page (adapter->pdev, tx_buffer_info->dma, - tx_buffer_info->length,PCI_DMA_TODEVICE ); - tx_buffer_info->dma =0 ; - } + tx_buffer_info->dma =0 ;if (tx_buffer_info->skb) { +skb_dma_unmap (&adapter->pdev->dev, tx_buffer_info->skb, +DMA_TO_DEVICE );dev_kfree_skb_any (tx_buffer_info->skb); tx_buffer_info->skb =NULL ; } + tx_buffer_info->time_stamp =0 ;/* tx_buffer_info must be completely set up in the transmit path */ }
@@ -204,15 +203,11 @@ static inline bool ixgbe_check_tx_hang(struct ixgbe_adapter *adapter, unsigned int eop) {struct ixgbe_hw *hw = &adapter->hw; -u32 head, tail;/* Detect a transmit hang in hardware, this serializes the * check with the clearing of time_stamp and movement of eop */ - head =IXGBE_READ_REG (hw, tx_ring->head); - tail =IXGBE_READ_REG (hw, tx_ring->tail); adapter->detect_tx_hung = false; -if ((head != tail) && - tx_ring->tx_buffer_info[eop].time_stamp && +if (tx_ring->tx_buffer_info[eop].time_stamp &&time_after (jiffies, tx_ring->tx_buffer_info[eop].time_stamp +HZ ) && !(IXGBE_READ_REG (&adapter->hw,IXGBE_TFCS ) &IXGBE_TFCS_TXOFF )) {/* detected Tx unit hang */
@@ -227,7 +222,8 @@ static inline bool ixgbe_check_tx_hang(struct ixgbe_adapter *adapter," time_stamp <%lx>\n" " jiffies <%lx>\n" , tx_ring->queue_index, - head, tail, +IXGBE_READ_REG (hw, tx_ring->head), +IXGBE_READ_REG (hw, tx_ring->tail), tx_ring->next_to_use, eop, tx_ring->tx_buffer_info[eop].time_stamp, jiffies);return true;
@@ -4164,32 +4160,39 @@ static int ixgbe_tx_map(struct ixgbe_adapter *adapter, struct sk_buff *skb, unsigned int first) {struct ixgbe_tx_buffer *tx_buffer_info; -unsigned int len = skb->len; +unsigned int len =skb_headlen (skb);unsigned int offset =0 , size, count =0 , i;unsigned int nr_frags =skb_shinfo (skb)->nr_frags;unsigned int f; - - len -= skb->data_len; +dma_addr_t *map; i = tx_ring->next_to_use; +if (skb_dma_map (&adapter->pdev->dev, skb,DMA_TO_DEVICE )) { +dev_err (&adapter->pdev->dev,"TX DMA map failed\n" ); +return 0 ; + } + + map =skb_shinfo (skb)->dma_maps; +while (len) { tx_buffer_info = &tx_ring->tx_buffer_info[i]; size =min (len, (uint )IXGBE_MAX_DATA_PER_TXD ); tx_buffer_info->length = size; - tx_buffer_info->dma =pci_map_single (adapter->pdev, - skb->data + offset, - size,PCI_DMA_TODEVICE ); + tx_buffer_info->dma = map[0 ] + offset; tx_buffer_info->time_stamp = jiffies; tx_buffer_info->next_to_watch = i; len -= size; offset += size; count++; - i++; -if (i == tx_ring->count) - i =0 ; + +if (len) { + i++; +if (i == tx_ring->count) + i =0 ; + } }for (f =0 ; f < nr_frags; f++) {
@@ -4197,33 +4200,27 @@ static int ixgbe_tx_map(struct ixgbe_adapter *adapter, frag = &skb_shinfo (skb)->frags[f]; len = frag->size; - offset = frag->page_offset; + offset =0 ;while (len) { + i++; +if (i == tx_ring->count) + i =0 ; + tx_buffer_info = &tx_ring->tx_buffer_info[i]; size =min (len, (uint )IXGBE_MAX_DATA_PER_TXD ); tx_buffer_info->length = size; - tx_buffer_info->dma =pci_map_page (adapter->pdev, - frag->page, - offset, - size, -PCI_DMA_TODEVICE ); + tx_buffer_info->dma = map[f +1 ] + offset; tx_buffer_info->time_stamp = jiffies; tx_buffer_info->next_to_watch = i; len -= size; offset += size; count++; - i++; -if (i == tx_ring->count) - i =0 ; } } -if (i ==0 ) - i = tx_ring->count -1 ; -else - i = i -1 ; + tx_ring->tx_buffer_info[i].skb = skb; tx_ring->tx_buffer_info[first].next_to_watch = i;
@@ -4389,13 +4386,19 @@ static int ixgbe_xmit_frame(struct sk_buff *skb, struct net_device *netdev) (skb->ip_summed == CHECKSUM_PARTIAL)) tx_flags |=IXGBE_TX_FLAGS_CSUM ; -ixgbe_tx_queue (adapter, tx_ring, tx_flags, -ixgbe_tx_map (adapter, tx_ring, skb, first), - skb->len, hdr_len); + count =ixgbe_tx_map (adapter, tx_ring, skb, first); - netdev->trans_start = jiffies; +if (count) { +ixgbe_tx_queue (adapter, tx_ring, tx_flags, count, skb->len, + hdr_len); + netdev->trans_start = jiffies; +ixgbe_maybe_stop_tx (netdev, tx_ring,DESC_NEEDED ); -ixgbe_maybe_stop_tx (netdev, tx_ring,DESC_NEEDED ); + }else { +dev_kfree_skb_any (skb); + tx_ring->tx_buffer_info[first].time_stamp =0 ; + tx_ring->next_to_use = first; + }return NETDEV_TX_OK ; }
Re: [PATCH 6/9] ixgbe: refactor tx buffer processing to use skb_dma_map/unmap
From: Jeff Kirsher <[email protected]>
Date: Wed, 01 Apr 2009 00:34:23 -0700
Quoted 8 lines
This patch resolves an issue with map single being used to map a buffer and then unmap page being used to unmap it. In addition it handles any error conditions that may be detected using
skb_dma_map.Signed-off-by Alexander Duyck <[email protected]>Signed-off-by Jeff Kirsher <[email protected]>
Applied.
PATCH 7/9 ixgbe: Patch to fix driver panic while freeing up tx & rx resources
In reply to Jeff KirsherFrom: Mallikarjuna R Chilakala <[email protected]>
When network interface is made active we were not handling the error scenarios properly to clean up rx & tx resources which might result in a driver panic.
---
drivers/net/ixgbe/ixgbe_main.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-)
drivers/net/ixgbe/ixgbe_main.c +2 -2
@@ -3486,10 +3486,10 @@ err_up:ixgbe_release_hw_control (adapter);ixgbe_free_irq (adapter); err_req_irq: -ixgbe_free_all_rx_resources (adapter); err_setup_rx: -ixgbe_free_all_tx_resources (adapter); +ixgbe_free_all_rx_resources (adapter); err_setup_tx: +ixgbe_free_all_tx_resources (adapter);ixgbe_reset (adapter);return err;
Re: [PATCH 7/9] ixgbe: Patch to fix driver panic while freeing up tx & rx resources
From: Jeff Kirsher <[email protected]>
Date: Wed, 01 Apr 2009 00:34:44 -0700
Quoted 7 lines
When network interface is made active we were not handling the error scenarios properly to clean up rx & tx resources which might result in a driver panic.
Signed-off-by Mallikarjuna R Chilakala <[email protected]>Signed-off-by Jeff Kirsher <[email protected]>
Applied.
PATCH 8/9 ixgbe: fix ethtool -A|a behavior
In reply to Jeff KirsherFrom: Don Skidmore <[email protected]>
We were basicly ignoring ethtool users request for FC autoneg and replying to queries with a "best guess". This patch enables the driver to store if we want to enable/disable autoneg FC and do the correct behavior.
---
drivers/net/ixgbe/ixgbe_82598.c | 3 ++- drivers/net/ixgbe/ixgbe_common.c | 3 ++- drivers/net/ixgbe/ixgbe_ethtool.c | 19 ++++++++++++++++--- drivers/net/ixgbe/ixgbe_main.c | 2 ++ drivers/net/ixgbe/ixgbe_type.h | 1 + 5 files changed, 23 insertions(+), 5 deletions(-)
drivers/net/ixgbe/ixgbe_82598.c +2 -1
@@ -411,7 +411,8 @@ static s32 ixgbe_setup_fc_82598(struct ixgbe_hw *hw, s32 packetbuf_num)/* Decide whether to use autoneg or not. */ hw->mac.ops.check_link (hw, &speed, &link_up, false); -if (hw->phy.multispeed_fiber && (speed ==IXGBE_LINK_SPEED_1GB_FULL )) +if (!hw->fc.disable_fc_autoneg && hw->phy.multispeed_fiber && + (speed ==IXGBE_LINK_SPEED_1GB_FULL )) ret_val =ixgbe_fc_autoneg (hw);if (ret_val)
drivers/net/ixgbe/ixgbe_common.c +2 -1
@@ -1937,7 +1937,8 @@ s32 ixgbe_setup_fc_generic(struct ixgbe_hw *hw, s32 packetbuf_num)/* Decide whether to use autoneg or not. */ hw->mac.ops.check_link (hw, &speed, &link_up, false); -if (hw->phy.multispeed_fiber && (speed ==IXGBE_LINK_SPEED_1GB_FULL )) +if (!hw->fc.disable_fc_autoneg && hw->phy.multispeed_fiber && + (speed ==IXGBE_LINK_SPEED_1GB_FULL )) ret_val =ixgbe_fc_autoneg (hw);if (ret_val)
drivers/net/ixgbe/ixgbe_ethtool.c +16 -3
@@ -234,7 +234,16 @@ static void ixgbe_get_pauseparam(struct net_device *netdev,struct ixgbe_adapter *adapter =netdev_priv (netdev);struct ixgbe_hw *hw = &adapter->hw; - pause->autoneg = (hw->fc.current_mode == ixgbe_fc_full ?1 :0 ); +/* +* Flow Control Autoneg isn't on if +* - we didn't ask for it OR +* - it failed, we know this by tx & rx being off +*/ +if (hw->fc.disable_fc_autoneg || + (hw->fc.current_mode == ixgbe_fc_none)) + pause->autoneg =0 ; +else + pause->autoneg =1 ;if (hw->fc.current_mode == ixgbe_fc_rx_pause) { pause->rx_pause =1 ;
@@ -252,8 +261,12 @@ static int ixgbe_set_pauseparam(struct net_device *netdev,struct ixgbe_adapter *adapter =netdev_priv (netdev);struct ixgbe_hw *hw = &adapter->hw; -if ((pause->autoneg ==AUTONEG_ENABLE ) || - (pause->rx_pause && pause->tx_pause)) +if (pause->autoneg !=AUTONEG_ENABLE ) + hw->fc.disable_fc_autoneg = true; +else + hw->fc.disable_fc_autoneg = false; + +if (pause->rx_pause && pause->tx_pause) hw->fc.requested_mode = ixgbe_fc_full;else if (pause->rx_pause && !pause->tx_pause) hw->fc.requested_mode = ixgbe_fc_rx_pause;
drivers/net/ixgbe/ixgbe_main.c +2 -0
@@ -3167,10 +3167,12 @@ static int __devinit ixgbe_sw_init(struct ixgbe_adapter *adapter)/* default flow control settings */ hw->fc.requested_mode = ixgbe_fc_full; + hw->fc.current_mode = ixgbe_fc_full;/* init for ethtool output */ hw->fc.high_water =IXGBE_DEFAULT_FCRTH ; hw->fc.low_water =IXGBE_DEFAULT_FCRTL ; hw->fc.pause_time =IXGBE_DEFAULT_FCPAUSE ; hw->fc.send_xon = true; + hw->fc.disable_fc_autoneg = false;/* enable itr by default in dynamic mode */ adapter->itr_setting =1 ;
drivers/net/ixgbe/ixgbe_type.h +1 -0
@@ -2005,6 +2005,7 @@ struct ixgbe_fc_info {u16 pause_time;/* Flow Control Pause timer */ bool send_xon;/* Flow control send XON */ bool strict_ieee;/* Strict IEEE mode */ +bool disable_fc_autoneg;/* Turn off autoneg FC mode */ enum ixgbe_fc_mode current_mode;/* FC mode in effect */ enum ixgbe_fc_mode requested_mode;/* FC mode requested by caller */ };
Re: [PATCH 8/9] ixgbe: fix ethtool -A|a behavior
From: Jeff Kirsher <[email protected]>
Date: Wed, 01 Apr 2009 00:35:05 -0700
Quoted 7 lines
We were basicly ignoring ethtool users request for FC autoneg and replying to queries with a "best guess". This patch enables the driver to store if we want to enable/disable autoneg FC and do the correct behavior.
Signed-off-by Don Skidmore <[email protected]>Signed-off-by Jeff Kirsher <[email protected]>
Applied.
PATCH 9/9 ixgbe: Fix potential memory leak/driver panic issue while setting up Tx & Rx ring parameters
In reply to Jeff KirsherFrom: Mallikarjuna R Chilakala <[email protected]>
While setting up the ring parameters using ethtool the driver can panic or leak memory as ixgbe_open tries to setup tx & rx resources. The updated logic will use ixgbe_down/up after successful allocation of tx & rx resources
---
drivers/net/ixgbe/ixgbe_ethtool.c | 101 +++++++++++++++++++++---------------- 1 files changed, 58 insertions(+), 43 deletions(-)
drivers/net/ixgbe/ixgbe_ethtool.c +58 -43
@@ -734,9 +734,10 @@ static int ixgbe_set_ringparam(struct net_device *netdev, struct ethtool_ringparam *ring) {struct ixgbe_adapter *adapter =netdev_priv (netdev); -struct ixgbe_ring *temp_ring; +struct ixgbe_ring *temp_tx_ring, *temp_rx_ring;int i, err;u32 new_rx_count, new_tx_count; +bool need_update = false;if ((ring->rx_mini_pending) || (ring->rx_jumbo_pending))return -EINVAL ;
@@ -755,80 +756,94 @@ static int ixgbe_set_ringparam(struct net_device *netdev,return 0 ; } - temp_ring =kcalloc (adapter->num_tx_queues, -sizeof (struct ixgbe_ring ),GFP_KERNEL ); -if (!temp_ring) -return -ENOMEM ; -while (test_and_set_bit (__IXGBE_RESETTING, &adapter->state))msleep (1 ); -if (new_tx_count != adapter->tx_ring->count) { + temp_tx_ring =kcalloc (adapter->num_tx_queues, +sizeof (struct ixgbe_ring ),GFP_KERNEL ); +if (!temp_tx_ring) { + err = -ENOMEM ; + goto err_setup; + } + +if (new_tx_count != adapter->tx_ring_count) { +memcpy (temp_tx_ring, adapter->tx_ring, + adapter->num_tx_queues *sizeof (struct ixgbe_ring ));for (i =0 ; i < adapter->num_tx_queues; i++) { - temp_ring[i].count = new_tx_count; - err =ixgbe_setup_tx_resources (adapter, &temp_ring[i]); + temp_tx_ring[i].count = new_tx_count; + err =ixgbe_setup_tx_resources (adapter, + &temp_tx_ring[i]);if (err) {while (i) { i--;ixgbe_free_tx_resources (adapter, - &temp_ring[i]); + &temp_tx_ring[i]); } goto err_setup; } - temp_ring[i].v_idx = adapter->tx_ring[i].v_idx; + temp_tx_ring[i].v_idx = adapter->tx_ring[i].v_idx; } -if (netif_running (netdev)) - netdev->netdev_ops->ndo_stop (netdev); -ixgbe_reset_interrupt_capability (adapter); -ixgbe_napi_del_all (adapter); -INIT_LIST_HEAD (&netdev->napi_list); -kfree (adapter->tx_ring); - adapter->tx_ring = temp_ring; - temp_ring =NULL ; - adapter->tx_ring_count = new_tx_count; + need_update = true; } - temp_ring =kcalloc (adapter->num_rx_queues, -sizeof (struct ixgbe_ring ),GFP_KERNEL ); -if (!temp_ring) { -if (netif_running (netdev)) - netdev->netdev_ops->ndo_open (netdev); -return -ENOMEM ; + temp_rx_ring =kcalloc (adapter->num_rx_queues, +sizeof (struct ixgbe_ring ),GFP_KERNEL ); +if ((!temp_rx_ring) && (need_update)) { +for (i =0 ; i < adapter->num_tx_queues; i++) +ixgbe_free_tx_resources (adapter, &temp_tx_ring[i]); +kfree (temp_tx_ring); + err = -ENOMEM ; + goto err_setup; } -if (new_rx_count != adapter->rx_ring->count) { +if (new_rx_count != adapter->rx_ring_count) { +memcpy (temp_rx_ring, adapter->rx_ring, + adapter->num_rx_queues *sizeof (struct ixgbe_ring ));for (i =0 ; i < adapter->num_rx_queues; i++) { - temp_ring[i].count = new_rx_count; - err =ixgbe_setup_rx_resources (adapter, &temp_ring[i]); + temp_rx_ring[i].count = new_rx_count; + err =ixgbe_setup_rx_resources (adapter, + &temp_rx_ring[i]);if (err) {while (i) { i--;ixgbe_free_rx_resources (adapter, - &temp_ring[i]); + &temp_rx_ring[i]); } goto err_setup; } - temp_ring[i].v_idx = adapter->rx_ring[i].v_idx; + temp_rx_ring[i].v_idx = adapter->rx_ring[i].v_idx; } + need_update = true; + } + +/* if rings need to be updated, here's the place to do it in one shot */ +if (need_update) {if (netif_running (netdev)) - netdev->netdev_ops->ndo_stop (netdev); -ixgbe_reset_interrupt_capability (adapter); -ixgbe_napi_del_all (adapter); -INIT_LIST_HEAD (&netdev->napi_list); -kfree (adapter->rx_ring); - adapter->rx_ring = temp_ring; - temp_ring =NULL ; - - adapter->rx_ring_count = new_rx_count; +ixgbe_down (adapter); + +/* tx */ +if (new_tx_count != adapter->tx_ring_count) { +kfree (adapter->tx_ring); + adapter->tx_ring = temp_tx_ring; + temp_tx_ring =NULL ; + adapter->tx_ring_count = new_tx_count; + } + +/* rx */ +if (new_rx_count != adapter->rx_ring_count) { +kfree (adapter->rx_ring); + adapter->rx_ring = temp_rx_ring; + temp_rx_ring =NULL ; + adapter->rx_ring_count = new_rx_count; + } }/* success! */ err =0 ; -err_setup: -ixgbe_init_interrupt_scheme (adapter);if (netif_running (netdev)) - netdev->netdev_ops->ndo_open (netdev); +ixgbe_up (adapter); +err_setup:clear_bit (__IXGBE_RESETTING, &adapter->state);return err; }
Re: [PATCH 9/9] ixgbe: Fix potential memory leak/driver panic issue while setting up Tx & Rx ring parameters
From: Jeff Kirsher <[email protected]>
Date: Wed, 01 Apr 2009 00:35:24 -0700
Quoted 8 lines
While setting up the ring parameters using ethtool the driver can panic or leak memory as
ixgbe_opentries to setup tx & rx resources. The updated logic will use ixgbe_down/up after successful allocation of tx & rx resourcesSigned-off-by Mallikarjuna R Chilakala <[email protected]>Signed-off-by Peter P Waskiewicz Jr <[email protected]>Signed-off-by Jeff Kirsher <[email protected]>
Applied.
From: Jeff Kirsher <[email protected]>
Date: Wed, 01 Apr 2009 00:32:42 -0700
Quoted 7 lines
The ixgbe driver had issues when DEBUG was defined because the
hw_dbgmacro was incomplete. This patch completes the code based off of the code that already existed in the igb module.Signed-off-by Alexander Duyck <[email protected]>Signed-off-by Jeff Kirsher <[email protected]>
Applied.