net driver fixes
Subsystems (7)
PATCH 1/1 net driver fixes
Includes several fixes from my slice of the akpm patch pie.
Please pull from 'davem-fixes' branch of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6.git davem-fixes
to receive the following updates:
MAINTAINERS | 4 +++- drivers/net/8139too.c | 3 ++- drivers/net/atl1e/atl1e_main.c | 2 -- drivers/net/au1000_eth.c | 3 ++- drivers/net/bonding/bond_alb.c | 24 ++++++++++++++++++++++++ drivers/net/cs89x0.c | 2 -- drivers/net/ehea/ehea_phyp.c | 2 +- drivers/net/ehea/ehea_qmr.c | 3 ++- drivers/net/forcedeth.c | 16 +++++++++++++--- drivers/net/mlx4/alloc.c | 1 + drivers/net/ne.c | 9 ++++++++- drivers/net/r8169.c | 25 ++++++++++++++----------- drivers/net/s2io.c | 18 ++++++++++-------- drivers/net/s2io.h | 1 + drivers/net/skfp/pmf.c | 29 +++++++++++------------------ drivers/net/smc91x.c | 3 ++- drivers/net/smc91x.h | 2 ++ drivers/net/tulip/de2104x.c | 1 - drivers/net/usb/hso.c | 9 +++++---- drivers/net/wan/hdlc_x25.c | 8 +++++--- 20 files changed, 106 insertions(+), 59 deletions(-)
Adrian Bunk (1):
[netdrvr/usb] hso_create_bulk_serial_device(): fix a double free
Andrew Morton (1):
drivers/net/mlx4/alloc.c needs mm.h
Breno Leitao (1):
s2io: Fix enabling VLAN tag stripping at driver initialization
Chris Snook (1):
MAINTAINERS: add Atheros maintainer for atlx
David Fries (1):
[netdrvr] ne: Fix suspend and resume for ISA PnP cards.
Francois Romieu (1):
r8169: fix RxMissed register access
Hannes Hering (1):
ehea: Fix DLPAR memory handling
Krzysztof Halasa (1):
wan/hdlc_x25.c: fix a NULL dereference
Magnus Damm (2):
smc91x: fix nowait printout
smc91x: SMC_IO_SHIFT platform data support for default case
Martin Gebert (1):
[netdrvr] au1000_eth: Spinlock initialisation fix
Mats Erik Andersson (1):
8139too: [cosmetic] fix incorrect register for flash-rom
Matthew Wilcox (1):
[netdrvr] atl1e: Don't take the mdio_lock in atl1e_probe
Sebastien Dugue (1):
ehea: fix phyp debugging typo
Takashi Iwai (1):
drivers/net/skfp/pmf.c: Use offsetof() macro
Thomas Bogendoerfer (1):
tulip: Fix dead 21041 ethernet after ifconfig down
Vlad Yasevich (1):
bonding: Do not tx-balance some IPv6 packets on ALB/TLB bonds
Yinghai Lu (1):
forcedeth: call restore mac addr in nv_shutdown path
Yoichi Yuasa (1):
cs89x0: fix warning release_irq label
MAINTAINERS +3 -1
@@ -750,11 +750,13 @@ P: Ville Syrjala M: [email protected] S: Maintained -ATL1 ETHERNET DRIVER +ATLX ETHERNET DRIVERS P: Jay Cliburn M: [email protected] P: Chris Snook M: [email protected] +P: Jie Yang +M: [email protected] L: [email protected] W: http://sourceforge.net/projects/atl1 W: http://atl1.sourceforge.net
drivers/net/8139too.c +2 -1
@@ -309,7 +309,7 @@ enum RTL8139_registers { Cfg9346 =0x50 , Config0 =0x51 , Config1 =0x52 , - FlashReg =0x54 , + TimerInt =0x54 , MediaStatus =0x58 , Config3 =0x59 , Config4 = 0x5A, /* absent on RTL-8139A */
@@ -325,6 +325,7 @@ enum RTL8139_registers {FIFOTMS =0x70 ,/* FIFO Control and test. */ CSCR =0x74 ,/* Chip Status and Configuration Register. */ PARA78 =0x78 , + FlashReg =0xD4 ,/* Communication with Flash ROM, four bytes. */ PARA7c =0x7c ,/* Magic transceiver parameter register. */ Config5 = 0xD8, /* absent on RTL-8139A */ };
drivers/net/atl1e/atl1e_main.c +0 -2
@@ -2390,9 +2390,7 @@ static int __devinit atl1e_probe(struct pci_dev *pdev, }/* Init GPHY as early as possible due to power saving issue */ -spin_lock (&adapter->mdio_lock);atl1e_phy_init (&adapter->hw); -spin_unlock (&adapter->mdio_lock);/* reset the controller to * put the device in a known good starting state */ err =atl1e_reset_hw (&adapter->hw);
drivers/net/au1000_eth.c +2 -1
@@ -653,6 +653,8 @@ static struct net_device * au1000_probe(int port_num) aup = dev->priv; +spin_lock_init (&aup->lock); +/* Allocate the data buffers */ /* Snooping works fine with eth on all au1xxx */ aup->vaddr = (u32)dma_alloc_noncoherent(NULL, MAX_BUF_SIZE *
@@ -753,7 +755,6 @@ static struct net_device * au1000_probe(int port_num) aup->tx_db_inuse[i] = pDB; } -spin_lock_init (&aup->lock); dev->base_addr = base; dev->irq = irq; dev->open = au1000_open;
drivers/net/bonding/bond_alb.c +24 -0
@@ -38,6 +38,7 @@#include <linux/in.h> #include <net/ipx.h> #include <net/arp.h> +#include <net/ipv6.h> #include <asm/byteorder.h> #include "bonding.h" #include "bond_alb.h"
@@ -81,6 +82,7 @@#define RLB_PROMISC_TIMEOUT 10*ALB_TIMER_TICKS_PER_SECstatic const u8 mac_bcast[ETH_ALEN ] = {0xff ,0xff ,0xff ,0xff ,0xff ,0xff }; +static const u8 mac_v6_allmcast[ETH_ALEN ] = {0x33 ,0x33 ,0x00 ,0x00 ,0x00 ,0x01 };static const int alb_delta_in_ticks =HZ /ALB_TIMER_TICKS_PER_SEC ;#pragma pack(1)
@@ -1290,6 +1292,7 @@ int bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev)u32 hash_index =0 ;const u8 *hash_start =NULL ;int res =1 ; +struct ipv6hdr *ip6hdr;skb_reset_mac_header (skb); eth_data =eth_hdr (skb);
@@ -1319,11 +1322,32 @@ int bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev) }break ;case ETH_P_IPV6 : +/* IPv6 doesn't really use broadcast mac address, but leave +* that here just in case. +*/ if (memcmp (eth_data->h_dest, mac_bcast,ETH_ALEN ) ==0 ) { do_tx_balance =0 ;break ; } +/* IPv6 uses all-nodes multicast as an equivalent to +* broadcasts in IPv4. +*/ +if (memcmp (eth_data->h_dest, mac_v6_allmcast,ETH_ALEN ) ==0 ) { + do_tx_balance =0 ; +break ; + } + +/* Additianally, DAD probes should not be tx-balanced as that +* will lead to false positives for duplicate addresses and +* prevent address configuration from working. +*/ + ip6hdr =ipv6_hdr (skb); +if (ipv6_addr_any (&ip6hdr->saddr)) { + do_tx_balance =0 ; +break ; + } + hash_start = (char *)&(ipv6_hdr (skb)->daddr); hash_size =sizeof (ipv6_hdr (skb)->daddr);break ;
drivers/net/cs89x0.c +0 -2
@@ -1397,9 +1397,7 @@ net_open(struct net_device *dev) release_dma:#if ALLOW_DMA free_dma (dev->dma); -#endif release_irq: -#if ALLOW_DMArelease_dma_buff (lp);#endif writereg (dev, PP_LineCTL,readreg (dev, PP_LineCTL) & ~(SERIAL_TX_ON |SERIAL_RX_ON ));
drivers/net/ehea/ehea_phyp.c +1 -1
@@ -535,7 +535,7 @@ u64 ehea_h_query_ehea(const u64 adapter_handle, void *cb_addr) cb_logaddr,/* R5 */ 0, 0, 0, 0, 0); /* R6-R10 */#ifdef DEBUG -ehea_dmp (cb_addr,sizeof (struct hcp_query_ehea ),"hcp_query_ehea" ); +ehea_dump (cb_addr,sizeof (struct hcp_query_ehea ),"hcp_query_ehea" );#endif return hret; }
drivers/net/ehea/ehea_qmr.c +2 -1
@@ -595,7 +595,8 @@ static int ehea_create_busmap_callback(unsigned long pfn, end_section = start_section + ((nr_pages *PAGE_SIZE ) /EHEA_SECTSIZE ); mr_len = *(unsigned long *)arg; - ehea_bmap =kzalloc (sizeof (struct ehea_bmap ),GFP_KERNEL ); +if (!ehea_bmap) + ehea_bmap =kzalloc (sizeof (struct ehea_bmap ),GFP_KERNEL );if (!ehea_bmap)return -ENOMEM ;
drivers/net/forcedeth.c +13 -3
@@ -5643,6 +5643,7 @@ static int __devinit nv_probe(struct pci_dev *pci_dev, const struct pci_device_i dev->dev_addr[4 ] = (np->orig_mac[0 ] >>8 ) &0xff ; dev->dev_addr[5 ] = (np->orig_mac[0 ] >>0 ) &0xff ;writel (txreg|NVREG_TRANSMITPOLL_MAC_ADDR_REV , base + NvRegTransmitPoll); +printk (KERN_DEBUG "nv_probe: set workaround bit for reversed mac addr\n" ); }memcpy (dev->perm_addr, dev->dev_addr, dev->addr_len);
@@ -5890,14 +5891,12 @@ static void nv_restore_phy(struct net_device *dev) } } -static void __devexit nv_remove(struct pci_dev *pci_dev) +static void nv_restore_mac_addr (struct pci_dev *pci_dev) {struct net_device *dev =pci_get_drvdata (pci_dev);struct fe_priv *np =netdev_priv (dev);u8 __iomem *base =get_hwbase (dev); -unregister_netdev (dev); -/* special op: write back the misordered MAC address - otherwise * the next nv_probe would see a wrong address. */
@@ -5905,6 +5904,15 @@ static void __devexit nv_remove(struct pci_dev *pci_dev)writel (np->orig_mac[1 ], base + NvRegMacAddrB);writel (readl (base + NvRegTransmitPoll) & ~NVREG_TRANSMITPOLL_MAC_ADDR_REV , base + NvRegTransmitPoll); +} + +static void __devexit nv_remove(struct pci_dev *pci_dev) +{ +struct net_device *dev =pci_get_drvdata (pci_dev); + +unregister_netdev (dev); + +nv_restore_mac_addr (pci_dev);/* restore any phy related changes */ nv_restore_phy (dev);
@@ -5975,6 +5983,8 @@ static void nv_shutdown(struct pci_dev *pdev)if (netif_running (dev))nv_close (dev); +nv_restore_mac_addr (pdev); +pci_disable_device (pdev);if (system_state ==SYSTEM_POWER_OFF ) {if (pci_enable_wake (pdev, PCI_D3cold, np->wolenabled))
drivers/net/mlx4/alloc.c +1 -0
@@ -33,6 +33,7 @@#include <linux/errno.h> #include <linux/slab.h> +#include <linux/mm.h> #include <linux/bitmap.h> #include <linux/dma-mapping.h> #include <linux/vmalloc.h>
drivers/net/ne.c +8 -1
@@ -844,8 +844,12 @@ static int ne_drv_suspend(struct platform_device *pdev, pm_message_t state) {struct net_device *dev =platform_get_drvdata (pdev); -if (netif_running (dev)) +if (netif_running (dev)) { +struct pnp_dev *idev = (struct pnp_dev *)ei_status.priv;netif_device_detach (dev); +if (idev) +pnp_stop_dev (idev); + }return 0 ; }
@@ -854,6 +858,9 @@ static int ne_drv_resume(struct platform_device *pdev)struct net_device *dev =platform_get_drvdata (pdev);if (netif_running (dev)) { +struct pnp_dev *idev = (struct pnp_dev *)ei_status.priv; +if (idev) +pnp_start_dev (idev);ne_reset_8390 (dev);NS8390p_init (dev,1 );netif_device_attach (dev);
drivers/net/r8169.c +14 -11
@@ -2099,8 +2099,6 @@ static void rtl_hw_start_8168(struct net_device *dev)RTL_R8 (IntrMask); -RTL_W32 (RxMissed,0 ); -rtl_set_rx_mode (dev);RTL_W8 (ChipCmd, CmdTxEnb | CmdRxEnb);
@@ -2143,8 +2141,6 @@ static void rtl_hw_start_8101(struct net_device *dev)RTL_R8 (IntrMask); -RTL_W32 (RxMissed,0 ); -rtl_set_rx_mode (dev);RTL_W8 (ChipCmd, CmdTxEnb | CmdRxEnb);
@@ -2922,6 +2918,17 @@ static int rtl8169_poll(struct napi_struct *napi, int budget)return work_done; } +static void rtl8169_rx_missed (struct net_device *dev,void __iomem *ioaddr) +{ +struct rtl8169_private *tp =netdev_priv (dev); + +if (tp->mac_version >RTL_GIGA_MAC_VER_06 ) +return ; + + dev->stats.rx_missed_errors += (RTL_R32 (RxMissed) &0xffffff ); +RTL_W32 (RxMissed,0 ); +} +static void rtl8169_down (struct net_device *dev) {struct rtl8169_private *tp =netdev_priv (dev);
@@ -2939,9 +2946,7 @@ core_down:rtl8169_asic_down (ioaddr); -/* Update the error counts. */ - dev->stats.rx_missed_errors +=RTL_R32 (RxMissed); -RTL_W32 (RxMissed,0 ); +rtl8169_rx_missed (dev, ioaddr);spin_unlock_irq (&tp->lock);
@@ -3063,8 +3068,7 @@ static struct net_device_stats *rtl8169_get_stats(struct net_device *dev)if (netif_running (dev)) {spin_lock_irqsave (&tp->lock, flags); - dev->stats.rx_missed_errors +=RTL_R32 (RxMissed); -RTL_W32 (RxMissed,0 ); +rtl8169_rx_missed (dev, ioaddr);spin_unlock_irqrestore (&tp->lock, flags); }
@@ -3089,8 +3093,7 @@ static int rtl8169_suspend(struct pci_dev *pdev, pm_message_t state)rtl8169_asic_down (ioaddr); - dev->stats.rx_missed_errors +=RTL_R32 (RxMissed); -RTL_W32 (RxMissed,0 ); +rtl8169_rx_missed (dev, ioaddr);spin_unlock_irq (&tp->lock);
drivers/net/s2io.c +10 -8
@@ -371,9 +371,6 @@ static void s2io_vlan_rx_register(struct net_device *dev, flags[i]); } -/* A flag indicating whether 'RX_PA_CFG_STRIP_VLAN_TAG' bit is set or not */ -static int vlan_strip_flag; -/* Unregister the vlan */ static void s2io_vlan_rx_kill_vid (struct net_device *dev,unsigned long vid) {
@@ -2303,7 +2300,7 @@ static int start_nic(struct s2io_nic *nic) val64 =readq (&bar0->rx_pa_cfg); val64 &= ~RX_PA_CFG_STRIP_VLAN_TAG ;writeq (val64, &bar0->rx_pa_cfg); - vlan_strip_flag =0 ; + nic->vlan_strip_flag =0 ; }/*
@@ -5010,7 +5007,7 @@ static void s2io_set_multicast(struct net_device *dev) val64 =readq (&bar0->rx_pa_cfg); val64 &= ~RX_PA_CFG_STRIP_VLAN_TAG ;writeq (val64, &bar0->rx_pa_cfg); - vlan_strip_flag =0 ; + sp->vlan_strip_flag =0 ; } val64 =readq (&bar0->mac_cfg);
@@ -5032,7 +5029,7 @@ static void s2io_set_multicast(struct net_device *dev) val64 =readq (&bar0->rx_pa_cfg); val64 |=RX_PA_CFG_STRIP_VLAN_TAG ;writeq (val64, &bar0->rx_pa_cfg); - vlan_strip_flag =1 ; + sp->vlan_strip_flag =1 ; } val64 =readq (&bar0->mac_cfg);
@@ -8206,6 +8203,11 @@ s2io_init_nic(struct pci_dev *pdev, const struct pci_device_id *pre)/* Initialize device name */ sprintf (sp->name,"%s Neterion %s" , dev->name, sp->product_name); +if (vlan_tag_strip) + sp->vlan_strip_flag =1 ; +else + sp->vlan_strip_flag =0 ; +/* * Make Link state as off at this point, when the Link change * interrupt comes the state will be automatically changed to
@@ -8311,7 +8313,7 @@ static int check_L2_lro_capable(u8 *buffer, struct iphdr **ip,* If vlan stripping is disabled and the frame is VLAN tagged, * shift the offset by the VLAN header size bytes. */ -if ((!vlan_strip_flag) && +if ((!sp->vlan_strip_flag) && (rxdp->Control_1 &RXD_FRAME_VLAN_TAG )) ip_off +=HEADER_VLAN_SIZE ; }else {
@@ -8592,7 +8594,7 @@ static void queue_rx_frame(struct sk_buff *skb, u16 vlan_tag) skb->protocol =eth_type_trans (skb, dev);if (sp->vlgrp && vlan_tag - && (vlan_strip_flag)) { + && (sp->vlan_strip_flag)) {/* Queueing the vlan frame to the upper layer */ if (sp->config.napi)vlan_hwaccel_receive_skb (skb, sp->vlgrp, vlan_tag);
drivers/net/s2io.h +1 -0
@@ -962,6 +962,7 @@ struct s2io_nic {int task_flag;unsigned long long start_time;struct vlan_group *vlgrp; +int vlan_strip_flag;#define MSIX_FLG 0xA5int num_entries;struct msix_entry *entries;
drivers/net/skfp/pmf.c +11 -18
@@ -44,17 +44,10 @@ static SMbuf *smt_build_pmf_response(struct s_smc *smc, struct smt_header *req, int set, int local);static int port_to_mib (struct s_smc *smc,int p); -#define MOFFSS (e) ((int)&(((struct fddi_mib *)0)->e)) -#define MOFFSA (e) ((int) (((struct fddi_mib *)0)->e)) - -#define MOFFMS (e) ((int)&(((struct fddi_mib_m *)0)->e)) -#define MOFFMA (e) ((int) (((struct fddi_mib_m *)0)->e)) - -#define MOFFAS (e) ((int)&(((struct fddi_mib_a *)0)->e)) -#define MOFFAA (e) ((int) (((struct fddi_mib_a *)0)->e)) - -#define MOFFPS (e) ((int)&(((struct fddi_mib_p *)0)->e)) -#define MOFFPA (e) ((int) (((struct fddi_mib_p *)0)->e)) +#define MOFFSS (e) offsetof(struct fddi_mib, e) +#define MOFFMS (e) offsetof(struct fddi_mib_m, e) +#define MOFFAS (e) offsetof(struct fddi_mib_a, e) +#define MOFFPS (e) offsetof(struct fddi_mib_p, e)#define AC_G 0x01/* Get */
@@ -87,8 +80,8 @@ static const struct s_p_tab { { SMT_P100D,AC_G, MOFFSS(fddiSMTOpVersionId), "S" } , { SMT_P100E,AC_G, MOFFSS(fddiSMTHiVersionId), "S" } , { SMT_P100F,AC_G, MOFFSS(fddiSMTLoVersionId), "S" } , - { SMT_P1010,AC_G, MOFFSA(fddiSMTManufacturerData), "D" } , - { SMT_P1011,AC_GR, MOFFSA(fddiSMTUserData), "D" } , + { SMT_P1010,AC_G, MOFFSS(fddiSMTManufacturerData), "D" } , + { SMT_P1011,AC_GR, MOFFSS(fddiSMTUserData), "D" } , { SMT_P1012,AC_G, MOFFSS(fddiSMTMIBVersionId), "S" } , /* StationConfigGrp */
@@ -103,7 +96,7 @@ static const struct s_p_tab { { SMT_P101D,AC_GR, MOFFSS(fddiSMTTT_Notify), "wS" } , { SMT_P101E,AC_GR, MOFFSS(fddiSMTStatRptPolicy), "bB" } , { SMT_P101F,AC_GR, MOFFSS(fddiSMTTrace_MaxExpiration),"lL" } , - { SMT_P1020,AC_G, MOFFSA(fddiSMTPORTIndexes), "II" } , + { SMT_P1020,AC_G, MOFFSS(fddiSMTPORTIndexes), "II" } , { SMT_P1021,AC_G, MOFFSS(fddiSMTMACIndexes), "I" } , { SMT_P1022,AC_G, MOFFSS(fddiSMTBypassPresent), "F" } ,
@@ -117,8 +110,8 @@ static const struct s_p_tab {/* MIBOperationGrp */ { SMT_P1032,AC_GROUP } , - { SMT_P1033,AC_G, MOFFSA(fddiSMTTimeStamp),"P" } , - { SMT_P1034,AC_G, MOFFSA(fddiSMTTransitionTimeStamp),"P" } , + { SMT_P1033,AC_G, MOFFSS(fddiSMTTimeStamp),"P" } , + { SMT_P1034,AC_G, MOFFSS(fddiSMTTransitionTimeStamp),"P" } ,/* NOTE : SMT_P1035 is already swapped ! SMT_P_SETCOUNT */ { SMT_P1035,AC_G, MOFFSS(fddiSMTSetCount),"4P" } , { SMT_P1036,AC_G, MOFFSS(fddiSMTLastSetStationId),"8" } ,
@@ -129,7 +122,7 @@ static const struct s_p_tab {* PRIVATE EXTENSIONS * only accessible locally to get/set passwd */ - { SMT_P10F0,AC_GR, MOFFSA(fddiPRPMFPasswd), "8" } , + { SMT_P10F0,AC_GR, MOFFSS(fddiPRPMFPasswd), "8" } , { SMT_P10F1,AC_GR, MOFFSS(fddiPRPMFStation), "8" } , #ifdef ESS { SMT_P10F2,AC_GR, MOFFSS(fddiESSPayload), "lL" } ,
@@ -245,7 +238,7 @@ static const struct s_p_tab { { SMT_P400E,AC_GR, MOFFPS(fddiPORTConnectionPolicies),"bB" } , { SMT_P400F,AC_G, MOFFPS(fddiPORTMacIndicated), "2" } , { SMT_P4010,AC_G, MOFFPS(fddiPORTCurrentPath), "E" } , - { SMT_P4011,AC_GR, MOFFPA(fddiPORTRequestedPaths), "l4" } , + { SMT_P4011,AC_GR, MOFFPS(fddiPORTRequestedPaths), "l4" } , { SMT_P4012,AC_G, MOFFPS(fddiPORTMACPlacement), "S" } , { SMT_P4013,AC_G, MOFFPS(fddiPORTAvailablePaths), "B" } , { SMT_P4016,AC_G, MOFFPS(fddiPORTPMDClass), "E" } ,
drivers/net/smc91x.c +2 -1
@@ -1961,7 +1961,8 @@ static int __init smc_probe(struct net_device *dev, void __iomem *ioaddr,if (dev->dma != (unsigned char )-1 )printk (" DMA %d" , dev->dma); -printk ("%s%s\n" , nowait ?" [nowait]" :"" , +printk ("%s%s\n" , + lp->cfg.flags &SMC91X_NOWAIT ?" [nowait]" :"" ,THROTTLE_TX_PKTS ?" [throttle_tx]" :"" );if (!is_valid_ether_addr (dev->dev_addr)) {
drivers/net/smc91x.h +2 -0
@@ -446,6 +446,8 @@ static inline void LPD7_SMC_outsw (unsigned char* a, int r,#define SMC_CAN_USE_32BIT 1#define SMC_NOWAIT 1 +#define SMC_IO_SHIFT (lp->io_shift) +#define SMC_inb (a, r) readb((a) + (r))#define SMC_inw (a, r) readw((a) + (r))#define SMC_inl (a, r) readl((a) + (r))
drivers/net/tulip/de2104x.c +0 -1
@@ -1418,7 +1418,6 @@ static int de_close (struct net_device *dev)de_free_rings (de);de_adapter_sleep (de); -pci_disable_device (de->pdev);return 0 ; }
drivers/net/usb/hso.c +5 -4
@@ -2211,14 +2211,14 @@ static struct hso_device *hso_create_bulk_serial_device( USB_DIR_IN);if (!serial->in_endp) {dev_err (&interface->dev,"Failed to find BULK IN ep\n" ); - goto exit; + goto exit2; }if (! (serial->out_endp =hso_get_ep (interface,USB_ENDPOINT_XFER_BULK ,USB_DIR_OUT ))) {dev_err (&interface->dev,"Failed to find BULK IN ep\n" ); - goto exit; + goto exit2; } serial->write_data = hso_std_serial_write_data;
@@ -2231,9 +2231,10 @@ static struct hso_device *hso_create_bulk_serial_device(/* done, return it */ return hso_dev; + +exit2: +hso_serial_common_free (serial); exit: -if (hso_dev && serial) -hso_serial_common_free (serial);kfree (serial);hso_free_device (hso_dev);return NULL ;
drivers/net/wan/hdlc_x25.c +5 -3
@@ -163,15 +163,17 @@ static void x25_close(struct net_device *dev)static int x25_rx (struct sk_buff *skb) { +struct net_device *dev = skb->dev; +if ((skb =skb_share_check (skb,GFP_ATOMIC )) ==NULL ) { - skb->dev->stats.rx_dropped++; + dev->stats.rx_dropped++;return NET_RX_DROP ; } -if (lapb_data_received (skb->dev, skb) ==LAPB_OK ) +if (lapb_data_received (dev, skb) ==LAPB_OK )return NET_RX_SUCCESS ; - skb->dev->stats.rx_errors++; + dev->stats.rx_errors++;dev_kfree_skb_any (skb);return NET_RX_DROP ; }
Hi Jeff,
Quoted 2 lines · Jeff Garzik
Vlad Yasevich (1):
bonding: Do not tx-balance some IPv6 packets on ALB/TLB bonds
Regarding this change and include/linux/etherdevice.h
Quoted 29 lines
drivers/net/bonding/bond_alb.c +3
-0@@ -38,6 +38,7 @@#include <linux/in.h> #include <net/ipx.h> #include <net/arp.h> +#include <net/ipv6.h> #include <asm/byteorder.h> #include "bonding.h" #include "bond_alb.h" @@ -81,6 +82,7 @@#define RLB_PROMISC_TIMEOUT 10*ALB_TIMER_TICKS_PER_SECstatic const u8 mac_bcast[ETH_ALEN ] = {0xff ,0xff ,0xff ,0xff ,0xff ,0xff }; +static const u8 mac_v6_allmcast[ETH_ALEN ] = {0x33 ,0x33 ,0x00 ,0x00 ,0x00 ,0x01 };static const int alb_delta_in_ticks =HZ /ALB_TIMER_TICKS_PER_SEC ;#pragma pack(1)@@ -1290,6 +1292,7 @@ int bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev)u32 hash_index =0 ;const u8 *hash_start =NULL ;int res =1 ; +struct ipv6hdr *ip6hdr;skb_reset_mac_header (skb); eth_data =eth_hdr (skb);
drivers/net/bonding/bond_alb.c int bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev) · line 1328
Quoted 7 lines
drivers/net/bonding/bond_alb.c +3
-0@@ -1319,11 +1322,32 @@ int bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev) }break ;case ETH_P_IPV6 : +/* IPv6 doesn't really use broadcast mac address, but leave +* that here just in case. +*/ if (memcmp (eth_data->h_dest, mac_bcast,ETH_ALEN ) ==0 ) {
Didn't someone introduce compare_ether_addr() just for this?
drivers/net/bonding/bond_alb.c int bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev) · line 1339
Quoted 11 lines
drivers/net/bonding/bond_alb.c +7
-0do_tx_balance =0 ;break ; } +/* IPv6 uses all-nodes multicast as an equivalent to +* broadcasts in IPv4. +*/ +if (memcmp (eth_data->h_dest, mac_v6_allmcast,ETH_ALEN ) ==0 ) { + do_tx_balance =0 ; +break ; + }
compare_ether_addr()
Maybe fix this in a followup patch?
Best Regards
Ingo Oeser
David Miller
maintainerFrom: Jeff Garzik <[email protected]>
Date: Sat, 13 Sep 2008 16:35:48 -0400
Quoted 1 line
Includes several fixes from my slice of the akpm patch pie.
Which of these have regression and/or bugzilla entries?
This is too much and I want to see it trimmed down.
Thanks.
Quoted 8 lines · David Miller
From: Jeff Garzik <[email protected]>
Date: Sat, 13 Sep 2008 16:35:48 -0400Quoted 1 line
Includes several fixes from my slice of the akpm patch pie.
Which of these have regression and/or bugzilla entries?
This is too much and I want to see it trimmed down.
What do you not want? The two warning fixes (mlx4, cs89x0) akpm seemed to want in 2.6.27, the 8139too change is cosmetic. The other stuff is critical AFAICS.
Jeff
David Miller
maintainerFrom: Jeff Garzik <[email protected]>
Date: Mon, 15 Sep 2008 16:17:50 -0400
Quoted 6 lines · David Miller
Quoted 5 lines
From: Jeff Garzik <[email protected]>
Date: Sat, 13 Sep 2008 16:35:48 -0400Quoted 1 line
Includes several fixes from my slice of the akpm patch pie.
Quoted 7 lines
Quoted 3 lines
Which of these have regression and/or bugzilla entries?
This is too much and I want to see it trimmed down.
What do you not want? The two warning fixes (mlx4, cs89x0) akpm seemed to want in 2.6.27, the 8139too change is cosmetic. The other stuff is critical AFAICS.
I'll take another look at the queue and try to give you some more specific feedback.
I can think of about 3 or 4 entries on the posted regression list that might be appropriate, so I essentially expected to see that many commits, give or take one or two.
Quoted 17 lines · David Miller
From: Jeff Garzik <[email protected]>
Date: Mon, 15 Sep 2008 16:17:50 -0400Quoted 6 lines · David Miller
Quoted 5 lines
From: Jeff Garzik <[email protected]>
Date: Sat, 13 Sep 2008 16:35:48 -0400Quoted 1 line
Includes several fixes from my slice of the akpm patch pie.
Quoted 7 lines
Quoted 3 lines
Which of these have regression and/or bugzilla entries?
This is too much and I want to see it trimmed down.
What do you not want? The two warning fixes (mlx4, cs89x0) akpm seemed to want in 2.6.27, the 8139too change is cosmetic. The other stuff is critical AFAICS.
Quoted 6 lines
I'll take another look at the queue and try to give you some more specific feedback.
I can think of about 3 or 4 entries on the posted regression list that might be appropriate, so I essentially expected to see that many commits, give or take one or two.
Don't stick strictly to that list... We want an oops fix even if the oops is present in 2.6.26. We want a spinlock fix that makes a driver work properly on SMP, even if that bug is present in 2.6.26. We want a build fix, etc.
Jeff
David Miller
maintainerFrom: Jeff Garzik <[email protected]>
Date: Mon, 15 Sep 2008 16:45:43 -0400
Quoted 4 lines
Don't stick strictly to that list... We want an oops fix even if the oops is present in 2.6.26. We want a spinlock fix that makes a driver work properly on SMP, even if that bug is present in 2.6.26. We want a build fix, etc.
Look at the analysis I just posted.
Most of those changes were complete and utter CRAP.
Build warning fixes, diagnostic message tidyups. You have to be kidding me.
Maybe a OOPS fix for common hardware. But not for something like obscure WAN stuff.
But otherwise, it is in fact strict by the rules regression list stuff. That's all we're fixing now.
And remember it's not you who gets your head chopped off if I try to merge in stuff that isn't appropriate now. And Linus has made it pretty clear what the goals of the non-merge-window period is, and that is to fix regressions. Otherwise the process never comes to closure in a reasonable amount of time.
Krzysztof Halasa
maintainerQuoted 2 lines · David Miller
Maybe a OOPS fix for common hardware. But not for something like obscure WAN stuff.
FYI: the WAN bit was actually a regression, 2.6.26 doesn't have this bug. Nevertheless generic HDLC especially with X.25 is not the mainstream code so I guess it's probably perfectly acceptable to delay the fix to minimize the patch flow.
Signature · 1 line
Krzysztof Halasa
Quoted 1 line · David Miller
Most of those changes were complete and utter CRAP.
thank you, but the fix I've sent (de2104x) isn't crap. The bug breaks the current debian-installer for Cobalt Qube1, when it tries to get an IP address via dhcp.
In my eyes it's a clean and good fix, but I don't care whether it appears in 2.6.27 or 2.6.28, because debian needs to backport it to 2.6.26 anyway. But declaring bugfixes, which just fixes an very old bug, as utter crap isn't really nice.
Thomas.
Signature · 2 lines
Crap can work. Given enough thrust pigs will fly, but it's not necessary a good idea. [ RFC1925, 2.3 ]
David Miller
maintainerFrom: [email protected] (Thomas Bogendoerfer)
Date: Mon, 15 Sep 2008 23:53:08 +0200
Quoted 6 lines · David Miller
Quoted 1 line
Most of those changes were complete and utter CRAP.
thank you, but the fix I've sent (de2104x) isn't crap. The bug breaks the current debian-installer for Cobalt Qube1, when it tries to get an IP address via dhcp.
And why is this bug that has been there for YEARS is appropriate to fix outside of the merge window?
Quoted 1 line
In my eyes it's a clean and good fix,
There are a lot of clean and good fixes, but that is not the criteria for outside-of-the merge-window submissions right now.
Linus has made this clear.
Quoted 4 lines · David Miller
There are a lot of clean and good fixes, but that is not the criteria for outside-of-the merge-window submissions right now.
Linus has made this clear.
He continues to take non-regression fixes and minor cosmetic commits from other trees.
The basic idea is to BALANCE the lateness of -rc with the seriousness of the problem being fixed, and the technical complexity of the fix itself.
The later the -rc, the more you need to take into account how likely it is to create additional problems. But did you READ the patches and extended commit descriptions, to see if additional problems were likely?
Late -rc does not mean "no bug fixes at all except regressions" if you look at the commits going into the kernel tree.
Late -rc does not mean putting off memory corruption fixes for 6 months.
Jeff
David Miller
maintainerFrom: Jeff Garzik <[email protected]>
Date: Mon, 15 Sep 2008 18:38:07 -0400
Quoted 2 lines
He continues to take non-regression fixes and minor cosmetic commits from other trees.
I could complain about that, but he's pushed back hard enough on me that I don't consider that very fruitful.
Feel free if you want.
Quoted 2 lines
Late -rc does not mean "no bug fixes at all except regressions" if you look at the commits going into the kernel tree.
Actually, since you brought this up, he's allowing this largely in obscure subsystems and/or code he happens to be personally working on fixes for.
Quoted 10 lines · David Miller
From: Jeff Garzik <[email protected]>
Date: Mon, 15 Sep 2008 18:38:07 -0400Quoted 2 lines
He continues to take non-regression fixes and minor cosmetic commits from other trees.
I could complain about that, but he's pushed back hard enough on me that I don't consider that very fruitful.
Feel free if you want.
My impression was that Linus was in Wrath of God(tm) mode <grin> and he was reacting strongly for effect not being specific.
But I would rather push my fixes, than complain about others' changes :)
Quoted 6 lines
Quoted 2 lines
Late -rc does not mean "no bug fixes at all except regressions" if you look at the commits going into the kernel tree.
Actually, since you brought this up, he's allowing this largely in obscure subsystems and/or code he happens to be personally working on fixes for.
Well, obscure subsystems like PCI
PCI: Fix printk warnings in setup-bus.c
PCI: Fix printk warnings in probe.c
or libata
sata_inic162x: enable LED blinking
ata: duplicate variable sparse warning
or ia64 (this is not a regression, but a regular bug)
ia64: fix panic during `modprobe -r xpc'
or bugs that have existed for a long time,
bfs: fix Lockdep warning
As I understand and observe, Linus never ever wants to create hard-and-fast rules, in this or anything, but set a general guideline of not putting in changes (fixes or whatever) late in the game that have a ghost of a chance of creating further problems. That is the underlying issue... don't create new headaches, _especially_ for large userbases ;-)
So, I -totally- understand your reaction, but I think this is erring too far on the side of caution. Oh well...
Jeff
David Miller
maintainerFrom: Jeff Garzik <[email protected]>
Date: Mon, 15 Sep 2008 19:25:22 -0400
Quoted 2 lines
So, I -totally- understand your reaction, but I think this is erring too far on the side of caution. Oh well...
Why don't you send me the trimmed down pull request for now, and we can discuss this topic and the specific other fixes as a follow-on thing?
I also think it matters how many fixes I batch up at once to merge, btw. So it's better to toss me 4 or 5 fixes ever could of days, rather than waiting a week and sending 12 or 13.
Quoted 6 lines · David Miller
From: [email protected] (Thomas Bogendoerfer)
Date: Mon, 15 Sep 2008 23:53:08 +0200Quoted 3 lines · David Miller
Quoted 1 line
Most of those changes were complete and utter CRAP.
Quoted 7 lines
Quoted 3 lines
thank you, but the fix I've sent (de2104x) isn't crap. The bug breaks the current debian-installer for Cobalt Qube1, when it tries to get an IP address via dhcp.
And why is this bug that has been there for YEARS is appropriate to fix outside of the merge window?
...
Such a fix is even suitable for -stable.
If something is considered suitable for -stable, but not for -rc with one month to go before the release of the kernel, then there's something wrong in the process.
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 11 lines · Adrian Bunk
Quoted 10 lines · David Miller
From: [email protected] (Thomas Bogendoerfer)
Date: Mon, 15 Sep 2008 23:53:08 +0200Quoted 6 lines · David Miller
Quoted 1 line
Most of those changes were complete and utter CRAP.
thank you, but the fix I've sent (de2104x) isn't crap. The bug breaks the current debian-installer for Cobalt Qube1, when it tries to get an IP address via dhcp.
Quoted 5 lines
Quoted 3 lines
And why is this bug that has been there for YEARS is appropriate to fix outside of the merge window?
...Such a fix is even suitable for -stable.
that was my feeling as well, but if David wants to see it in the next merge window, it's his decision. It's not a important fix as the user base is obviuosly pretty small. And I didn't know about that bug, when the merge windows was open, so I couldn't fix it "in time".
Thomas.
Signature · 2 lines
Crap can work. Given enough thrust pigs will fly, but it's not necessary a good idea. [ RFC1925, 2.3 ]
Quoted 16 lines · Thomas Bogendoerfer
Quoted 15 lines · Adrian Bunk
Quoted 13 lines · David Miller
From: [email protected] (Thomas Bogendoerfer)
Date: Mon, 15 Sep 2008 23:53:08 +0200Quoted 6 lines · David Miller
Quoted 1 line
Most of those changes were complete and utter CRAP.
thank you, but the fix I've sent (de2104x) isn't crap. The bug breaks the current debian-installer for Cobalt Qube1, when it tries to get an IP address via dhcp.
And why is this bug that has been there for YEARS is appropriate to fix outside of the merge window?
...
Quoted 6 lines
Quoted 1 line
Such a fix is even suitable for -stable.
that was my feeling as well, but if David wants to see it in the next merge window, it's his decision. It's not a important fix as the user base is obviuosly pretty small. And I didn't know about that bug, when the merge windows was open, so I couldn't fix it "in time".
And if it goes in during the next merge window it can immediately go into 2.6.27.1 (and even 2.6.26.y if it's still maintained), since the fix fulfills Documentation/stable_kernel_rules.txt .
That doesn't make sense.
Quoted 1 line
Thomas.
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 24 lines · Adrian Bunk
Quoted 16 lines · Thomas Bogendoerfer
Quoted 15 lines · Adrian Bunk
Quoted 13 lines · David Miller
From: [email protected] (Thomas Bogendoerfer)
Date: Mon, 15 Sep 2008 23:53:08 +0200Quoted 6 lines · David Miller
Quoted 1 line
Most of those changes were complete and utter CRAP.
thank you, but the fix I've sent (de2104x) isn't crap. The bug breaks the current debian-installer for Cobalt Qube1, when it tries to get an IP address via dhcp.
And why is this bug that has been there for YEARS is appropriate to fix outside of the merge window?
...Quoted 6 lines
Quoted 1 line
Such a fix is even suitable for -stable.
that was my feeling as well, but if David wants to see it in the next merge window, it's his decision. It's not a important fix as the user base is obviuosly pretty small. And I didn't know about that bug, when the merge windows was open, so I couldn't fix it "in time".
Quoted 5 lines
And if it goes in during the next merge window it can immediately go into 2.6.27.1 (and even 2.6.26.y if it's still maintained), since the fix fulfills
Documentation/stable_kernel_rules.txt.That doesn't make sense.
Note that my emails are not meant against David personally - he just seems to have gotten from pushing too much to pushing too few, with such illogical results.
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 24 lines · Adrian Bunk
Quoted 23 lines · Adrian Bunk
Quoted 10 lines · Thomas Bogendoerfer
Quoted 9 lines · Adrian Bunk
Quoted 8 lines · David Miller
From: [email protected] (Thomas Bogendoerfer)
Date: Mon, 15 Sep 2008 23:53:08 +0200Quoted 5 lines · David Miller
Quoted 1 line
Most of those changes were complete and utter CRAP.
thank you, but the fix I've sent (de2104x) isn't crap. The bug breaks the current debian-installer for Cobalt Qube1, when it tries to get an IP address via dhcp.
Quoted 8 lines
Quoted 4 lines
Quoted 3 lines
And why is this bug that has been there for YEARS is appropriate to fix outside of the merge window?
...Such a fix is even suitable for -stable.
that was my feeling as well, but if David wants to see it in the next merge window, it's his decision. It's not a important fix as the user base is obviuosly pretty small. And I didn't know about that bug, when the merge windows was open, so I couldn't fix it "in time".
And if it goes in during the next merge window it can immediately go into 2.6.27.1 (and even 2.6.26.y if it's still maintained), since the fix fulfills
Documentation/stable_kernel_rules.txt.
Quoted 8 lines
Quoted 1 line
That doesn't make sense.
Note that my emails are not meant against David personally - he just seems to have gotten from pushing too much to pushing too few, with such illogical results.
cu
Adrian
Probably he can read.
Jarek P.
PS:
Quoted 14 lines · Linus Torvalds
Quoted 1 line · Linus Torvalds
What's so hard to understand about this?
Here's a simple rule of thumb:
- if it's not on the regression list
- if it's not a reported security hole
- if it's not on the reported oopses list
then why are people sending it to me?IOW, if it's just another random improvement, and you send it to me outside of the merge window, then what is the point of the merge window?
Linus
Quoted 27 lines · Jarek Poplawski
Quoted 26 lines · Adrian Bunk
Quoted 24 lines · Adrian Bunk
Quoted 10 lines · Thomas Bogendoerfer
Quoted 9 lines · Adrian Bunk
Quoted 8 lines · David Miller
From: [email protected] (Thomas Bogendoerfer)
Date: Mon, 15 Sep 2008 23:53:08 +0200Quoted 5 lines · David Miller
Quoted 1 line
Most of those changes were complete and utter CRAP.
thank you, but the fix I've sent (de2104x) isn't crap. The bug breaks the current debian-installer for Cobalt Qube1, when it tries to get an IP address via dhcp.
Quoted 8 lines
Quoted 4 lines
Quoted 3 lines
And why is this bug that has been there for YEARS is appropriate to fix outside of the merge window?
...Such a fix is even suitable for -stable.
that was my feeling as well, but if David wants to see it in the next merge window, it's his decision. It's not a important fix as the user base is obviuosly pretty small. And I didn't know about that bug, when the merge windows was open, so I couldn't fix it "in time".
And if it goes in during the next merge window it can immediately go into 2.6.27.1 (and even 2.6.26.y if it's still maintained), since the fix fulfills
Documentation/stable_kernel_rules.txt.That doesn't make sense.
Quoted 8 lines
Quoted 6 lines
Note that my emails are not meant against David personally - he just seems to have gotten from pushing too much to pushing too few, with such illogical results.
cu
AdrianProbably he can read.
I'm eagerly awaiting to slap this at Linus the next time he dares to directly apply a patch that only fixes a sparse warning outside of a merge window... ;-)
More seriously, there's a difference between Linus' "another random improvement" and an "is even suitable for -stable".
I'm not reading Linus' (Cc'ed) statement the way that a patch that is appropriate for 2.6.27.1 is not appropriate for -rc now.
If I'm misreading Linus on this it might make sense if they'd discuss it at their current cabal meeting.
Quoted 19 lines
Jarek P.
PS:
Quoted 14 lines · Linus Torvalds
Quoted 1 line · Linus Torvalds
What's so hard to understand about this?
Here's a simple rule of thumb:
- if it's not on the regression list
- if it's not a reported security hole
- if it's not on the reported oopses list
then why are people sending it to me?IOW, if it's just another random improvement, and you send it to me outside of the merge window, then what is the point of the merge window?
Linus
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 6 lines · Adrian Bunk
More seriously, there's a difference between Linus' "another random improvement" and an "is even suitable for -stable".
I'm not reading Linus' (Cc'ed) statement the way that a patch that is appropriate for 2.6.27.1 is not appropriate for -rc now.
Well, remember that patches that get published for -stable do have to go through an extra review process. It's not true that any "obviously correct" bug fix gets automatically published in -stable. Sometimes bug fixes do get rejected for -stable because they are too risky, or require more time for testing in the -rc series before they are deemed suitable for -stable.
Looking at 2.6.26, there is currently about 195 patches queued up, which is certainly not all of the bugs fixed during the 2.6.27 development cycle. Some of it is that some developers aren't as aggressive about nominating patches for backporting to 2.6.26, and some of it is that only the really *important* bugs get this treatment.
But, I bet that if there was a major embarassing problem where someone abused the process by submitting an "obviously correct" patch that ended up breaking things for a others, the end result would be that the -stable team would start cracking down and start using more strict criteria about what would be considered allowable for publishing in -stable.
I'll remind people that the original rule was that new features were only allowed in -rc1, and bug fixes in -rc1 and -rc2. But because that got abused, Linus is now saying that only regression bug fixes are allowed post -rc1. My belief is that part of the reason for this is that Linus is a self-described soft touch, and by stating a draconian rule, he's hoping that subsystem maintainers like David will lay done the law, and prevent the really serious abuses (where people starting submitting "obvious bug fixes" in -rc5 and -rc6 that caused regressions which in turn ends up extending the release cycle).
If it's a really important bug, and it affects a huge number of users, or it's really bad security bug, the reality is that exceptions will be made to the rules. But exceptions need to remain exceptions for extraordinary situations, not everyday occurrences. And of course, if the bug does affect a huge number of users, someone should be asking the question why it wasn't detected sooner, say before the last merge window --- and to ask the question how many users is this bug really going to affect anyway?
At least, that's my take on things,
- Ted
Quoted 11 lines · On Tue, 16 Sep 2008 09:54:08 -0400 Theodore Tso <[email protected]> wrote:
If it's a really important bug, and it affects a huge number of users, or it's really bad security bug, the reality is that exceptions will be made to the rules. But exceptions need to remain exceptions for extraordinary situations, not everyday occurrences. And of course, if the bug does affect a huge number of users, someone should be asking the question why it wasn't detected sooner, say before the last merge window --- and to ask the question how many users is this bug really going to affect anyway?
At least, that's my take on things,
also note that Linus said "regression or on the kerneloops list"; if it has any kind of backtrace, it'll be there if it's a common problem that hits many users.
Signature · 3 lines
Arjan van de Ven Intel Open Source Technology Centre For development, discussion and tips for power savings, visit http://www.lesswatts.org
Quoted 12 lines · Arjan van de Ven
Quoted 9 lines · On Tue, 16 Sep 2008 09:54:08 -0400 Theodore Tso <[email protected]> wrote:
If it's a really important bug, and it affects a huge number of users, or it's really bad security bug, the reality is that exceptions will be made to the rules. But exceptions need to remain exceptions for extraordinary situations, not everyday occurrences. And of course, if the bug does affect a huge number of users, someone should be asking the question why it wasn't detected sooner, say before the last merge window --- and to ask the question how many users is this bug really going to affect anyway?
Quoted 6 lines
Quoted 2 lines
At least, that's my take on things,
also note that Linus said "regression or on the kerneloops list"; if it has any kind of backtrace, it'll be there if it's a common problem that hits many users.
Thomas' patch [1] will fixes a bug of a kind that will likely never make it to your list.
And the same also e.g. goes for bugs where your machine is completely dead (no SysRq possible) with nothing in the logs.
The kerneloops lists are quite valuable, but they can never cover all classes of fatal bugs.
cu
Adrian
[1] http://article.gmane.org/gmane.linux.network/105810
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 17 lines · On Tue, 16 Sep 2008 17:34:10 +0300 Adrian Bunk <[email protected]> wrote:
Quoted 16 lines · Arjan van de Ven
Quoted 12 lines · On Tue, 16 Sep 2008 09:54:08 -0400 Theodore Tso <[email protected]> wrote:
If it's a really important bug, and it affects a huge number of users, or it's really bad security bug, the reality is that exceptions will be made to the rules. But exceptions need to remain exceptions for extraordinary situations, not everyday occurrences. And of course, if the bug does affect a huge number of users, someone should be asking the question why it wasn't detected sooner, say before the last merge window --- and to ask the question how many users is this bug really going to affect anyway?
At least, that's my take on things,
Quoted 6 lines
Quoted 3 lines
also note that Linus said "regression or on the kerneloops list"; if it has any kind of backtrace, it'll be there if it's a common problem that hits many users.
Thomas' patch [1] will fixes a bug of a kind that will likely never make it to your list.
not sure; just we need to catch doing pci_disable_device on a non-enabled device as a WARN_ON.
and the patch looks quite wrong, the real answer should be to do the enable in open() :-)
Quoted 3 lines
And the same also e.g. goes for bugs where your machine is completely dead (no SysRq possible) with nothing in the logs.
that's being worked on, by storing the crash data in some non-volatile piece of memory
Quoted 3 lines
The kerneloops lists are quite valuable, but they can never cover all classes of fatal bugs.
if you want perfection, you're not going to get it.
"Perfect is the enemy of good"
If you want to help make things better (as opposed to perfect), you're very welcome to help out, be it with suggestions on how to improve or with actual code.
Signature · 3 lines
Arjan van de Ven Intel Open Source Technology Centre For development, discussion and tips for power savings, visit http://www.lesswatts.org
Quoted 2 lines · Arjan van de Ven
not sure; just we need to catch doing
pci_disable_deviceon a non-enabled device as aWARN_ON.
the pci_disable_device() is done on an enabled device, so your new WARN_ON wouldn't trigger.
Quoted 2 lines
and the patch looks quite wrong, the real answer should be to do the enable in
open():-)
I thought about the solution, but the pci device is completely setup and enabled in the init function so disabling it at that point and enabling it in open() again sounds silly to me. pci_enable_enable/disable is already done in .init/.remove and .suspend/.resume, so it looked more obviuos to not let close() do the pci_disable_device().
Thomas.
Signature · 2 lines
Crap can work. Given enough thrust pigs will fly, but it's not necessary a good idea. [ RFC1925, 2.3 ]
Quoted 8 lines · Arjan van de Ven
On Tue, 16 Sep 2008 17:34:10 +0300
Adrian Bunk <[email protected]> wrote:
...Quoted 2 lines
And the same also e.g. goes for bugs where your machine is completely dead (no SysRq possible) with nothing in the logs.
that's being worked on, by storing the crash data in some non-volatile piece of memory
The kerneloops list is helpful, and improving it is great, but it will never be able to cover 100% of all problems on 100% of all computers.
Quoted 10 lines
Quoted 2 lines
The kerneloops lists are quite valuable, but they can never cover all classes of fatal bugs.
if you want perfection, you're not going to get it.
"Perfect is the enemy of good"
If you want to help make things better (as opposed to perfect), you're very welcome to help out, be it with suggestions on how to improve or with actual code.
You miss my point:
We agree that the kerneloops list is not perfect.
And usually the bugs with a trace that go into the list are the ones that are easier to debug and fix, while the others might be much harder to debug.
If one takes Linus' "simple rule of thumb" literally then a hard to debug bug without any messages anywhere mustn't be fixed outside of a merge window. This literal reading is what doesn't make sense.
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 8 lines · Theodore Tso
Quoted 6 lines · Adrian Bunk
More seriously, there's a difference between Linus' "another random improvement" and an "is even suitable for -stable".
I'm not reading Linus' (Cc'ed) statement the way that a patch that is appropriate for 2.6.27.1 is not appropriate for -rc now.
Quoted 15 lines
Well, remember that patches that get published for -stable do have to go through an extra review process. It's not true that any "obviously correct" bug fix gets automatically published in -stable. Sometimes bug fixes do get rejected for -stable because they are too risky, or require more time for testing in the -rc series before they are deemed suitable for -stable.
...
If it's a really important bug, and it affects a huge number of users, or it's really bad security bug, the reality is that exceptions will be made to the rules. But exceptions need to remain exceptions for extraordinary situations, not everyday occurrences. And of course, if the bug does affect a huge number of users, someone should be asking the question why it wasn't detected sooner, say before the last merge window --- and to ask the question how many users is this bug really going to affect anyway?
Among the patches David rejected based on what Linus said was Thomas' patch [1].
It is not a "really important bug" and does not affect "huge number of users".
But it passes Documentation/stable_kernel_rules.txt, and can we agree that if it gets submitted for 2.6.27.1 it will likely pass review and get applied?
That's the kind of patches I'm talking about.
Quoted 3 lines
At least, that's my take on things,
- Ted
cu
Adrian
[1] http://article.gmane.org/gmane.linux.network/105810
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedDavid Miller
maintainerFrom: David Miller <[email protected]>
Date: Mon, 15 Sep 2008 13:27:16 -0700 (PDT)
Quoted 2 lines
I'll take another look at the queue and try to give you some more specific feedback.
Thomas Bogendoerfer (1):
tulip: Fix dead 21041 ethernet after ifconfig down
Not a regression, that problem has been there forever.
Adrian Bunk (1):
[netdrvr/usb] hso_create_bulk_serial_device(): fix a double free
This is a new driver in 2.6.27, so OK.
Andrew Morton (1):
drivers/net/mlx4/alloc.c needs mm.h
Build bustage, so ok.
Breno Leitao (1):
s2io: Fix enabling VLAN tag stripping at driver initialization
Not a 2.6.27 regression, 2.6.26 has this problem too.
Chris Snook (1):
MAINTAINERS: add Atheros maintainer for atlx
Doc change, so fine I guess.
David Fries (1):
[netdrvr] ne: Fix suspend and resume for ISA PnP cards.
Definitely not a 2.6.27 regression.
Francois Romieu (1):
r8169: fix RxMissed register access
Has a bugzilla entry, etc., so OK.
Hannes Hering (1):
ehea: Fix DLPAR memory handling
No kernel bugzilla, no regression list entry, no way.
Krzysztof Halasa (1):
wan/hdlc_x25.c: fix a NULL dereference
Not a 2.6.27 regression, this problem has been there forever.
Magnus Damm (2):
smc91x: fix nowait printout
This one is good but the commit message needs to be fixed up. It references a raw SHA1 ID but does not give the commit message header line as well, which is a requirement.
smc91x: SMC_IO_SHIFT platform data support for default case
Ok, but needs commit header line text for referenced commit.
Martin Gebert (1):
[netdrvr] au1000_eth: Spinlock initialisation fix
This problem has existed forever, not a 2.6.27 regression, and therefore not appropriate.
Mats Erik Andersson (1):
8139too: [cosmetic] fix incorrect register for flash-rom
Cosmetic, indeed. Not appropriate for 2.6.27
Matthew Wilcox (1):
[netdrvr] atl1e: Don't take the mdio_lock in atl1e_probe
This is just lockep noise, and definitely not a 2.6.27 regression as far as I can tell. So this one has to go.
Sebastien Dugue (1):
ehea: fix phyp debugging typo
DEBUG build fix, not a regression, not appropriate.
Takashi Iwai (1):
drivers/net/skfp/pmf.c: Use offsetof() macro
Cleanup, not appropriate. The code is actually fine but emits some warnings during the build.
Vlad Yasevich (1):
bonding: Do not tx-balance some IPv6 packets on ALB/TLB bonds
Not a 2.6.27 regression, not appropriate.
Yinghai Lu (1): forcedeth: call restore mac addr in nv_shutdown path
This appears to fix an entry in the regression list, so OK.
Yoichi Yuasa (1):
cs89x0: fix warning release_irq label
Warning fix, not appropriate for 2.6.27
So, as I said, the vast majority of this stuff is absolulte not appropriate. Most of it can totally wait for 2.6.28, and not be merged now.
You can accelerate this process by making sure that every commit message has a reference to the regression list entry or a kernel bugzilla that the patch fixes.
So please fix up this patch queue and resubmit, thanks.
Quoted 21 lines · David Miller
From: David Miller <[email protected]>
Date: Mon, 15 Sep 2008 13:27:16 -0700 (PDT)Quoted 2 lines
I'll take another look at the queue and try to give you some more specific feedback.
Thomas Bogendoerfer (1):
tulip: Fix dead 21041 ethernet after ifconfig downNot a regression, that problem has been there forever.
Hannes Hering (1):
ehea: Fix DLPAR memory handlingNo kernel bugzilla, no regression list entry, no way.
Krzysztof Halasa (1):
wan/hdlc_x25.c: fix a NULL dereferenceNot a 2.6.27 regression, this problem has been there forever.
Quoted 7 lines
So, as I said, the vast majority of this stuff is absolulte not appropriate. Most of it can totally wait for 2.6.28, and not be merged now.
You can accelerate this process by making sure that every commit message has a reference to the regression list entry or a kernel bugzilla that the patch fixes.
So, my takeaway from this is...
1) Creating a bugzilla entry magically makes a bug fix acceptable?
2) We no longer want "this kills the driver" fixes?
I disagree with that logic, and I seriously doubt Linus wants to turn away serious fixes to serious problems. Many of these are clearly needed, just read the extended patch description and the patch itself.
You just rejected patches that (a) fixed dead ethernet [de2104x], (b) fixed an oops [WAN], and (c) fixed memory corruption [ehea].
I feel like I have just stepped into Odd World, if you don't want fixes as serious as these.
Jeff
David Miller
maintainerFrom: Jeff Garzik <[email protected]>
Date: Mon, 15 Sep 2008 17:24:59 -0400
Quoted 3 lines
So, my takeaway from this is...
1) Creating a bugzilla entry magically makes a bug fix acceptable?
It adds more credence to the bug. Because it's usually reported by a user.
Quoted 1 line
2) We no longer want "this kills the driver" fixes?
I don't take that away from what I said.
Why is it so important to fix something "RIGHT NOW" that has existed forever? Where were these fixes during the merge window?
Quoted 2 lines
I disagree with that logic, and I seriously doubt Linus wants to turn away serious fixes to serious problems.
He wants the time outside of the merge window to work on fixing regressions and exploitable problems. This is so that the process actually ends at some point.
Quoted 1 line
You just rejected patches that (a) fixed dead ethernet [de2104x],
And it's getting submitted now, because?!?
(b) fixed an
Quoted 1 line
oops [WAN], and
It's WAN, it's been there forever, who cares and why now?
Quoted 1 line
(c) fixed memory corruption [ehea].
Again, it's been there for quite some time, why is it "now" all of sudden now important to integrate this fix?
Quoted 5 lines · David Miller
Quoted 1 line
You just rejected patches that (a) fixed dead ethernet [de2104x],
And it's getting submitted now, because?!?
(b) fixed an
Quoted 8 lines
Quoted 1 line
oops [WAN], and
It's WAN, it's been there forever, who cares and why now?
Quoted 1 line
(c) fixed memory corruption [ehea].
Again, it's been there for quite some time, why is it "now" all of sudden now important to integrate this fix?
Because users should not be forced to wait 3 months for fixes, for admittedly serious bugs, which are ready for them right now. Small, self-contained, tested, obvious fixes.
There is just something horribly wrong if you are not taking two-line fixes for memory corruption, even if the problem existed since kernel 1.2.13.
July 13: 2.6.26 released, merge window opens July 29: 2.6.27-rc1 released, merge window closes Sept 20? 2.6.27 release (wild guess) Dec 25? 2.6.28 release (wild guess)
So if a small, easily-mergable fix for a major bug like an oops appears on July 30, being strict to your plan has this fix waiting until almost 2009 (2.6.28 release), a wait of over five months.
That is not serving Linux users, our primary customers.
Complaining about the 8139too change, I can see that (even though Linus takes that stuff via libata all the time). But about memory corruption and oopses? Come on.
Jeff
David Miller
maintainerFrom: Jeff Garzik <[email protected]>
Date: Mon, 15 Sep 2008 18:34:07 -0400
Quoted 5 lines
So if a small, easily-mergable fix for a major bug like an oops appears on July 30, being strict to your plan has this fix waiting until almost 2009 (2.6.28 release), a wait of over five months.
That is not serving Linux users, our primary customers.
I personally partly agree with you, but Linus put his foot down and I'm implementing the stated policy by doing the same.
On Mon, Sep 15, 2008 at 02:11:59PM -0700, David Miller wrote:
[snip]
Quoted 6 lines
Not a 2.6.27 regression, this problem has been there forever.
Magnus Damm (2):
smc91x: fix nowait printout
Quoted 7 lines
This one is good but the commit message needs to be fixed up. It references a raw SHA1 ID but does not give the commit message header line as well, which is a requirement.
smc91x:
SMC_IO_SHIFTplatform data support for default caseOk, but needs commit header line text for referenced commit.
[snip]
Hi Dave,
can you confirm that the commit message requirement stands even if the SHA1 ID references Linus' tree? I personally think that is reasonable, I just want to confirm that is the requirement.
Signature · 3 lines
Simon Horman VA Linux Systems Japan K.K., Sydney, Australia Satellite Office H: www.vergenet.net/~horms/ W: www.valinux.co.jp/en
David Miller
maintainerFrom: Simon Horman <[email protected]>
Date: Tue, 16 Sep 2008 12:01:18 +1000
Quoted 13 lines · David Miller
[snip]
Quoted 10 lines
Not a 2.6.27 regression, this problem has been there forever.
Magnus Damm (2):
smc91x: fix nowait printoutThis one is good but the commit message needs to be fixed up. It references a raw SHA1 ID but does not give the commit message header line as well, which is a requirement.
Quoted 7 lines
Quoted 3 lines
smc91x:
SMC_IO_SHIFTplatform data support for default caseOk, but needs commit header line text for referenced commit.
can you confirm that the commit message requirement stands even if the SHA1 ID references Linus' tree? I personally think that is reasonable, I just want to confirm that is the requirement.
Yes, because the commit messages will propagate back usually verbatim for changes backported into the -stable branches.