regression: CD burning (k3b) went broke

Greetings,

K3b recently (9a4c854..5d9c4a7 pull) began terminally griping about buffer underrun upon every attempt to burn a CD. I can't fully bisect the problem because intervening kernels hang soft during boot. Using git bisect visualize, and converting to postable text:

bisect/bad block: add request->raw_data_len (6b00769fe1502b4ad97bb327ef7ac971b208bfb5)
bisect block: update bio according to DMA alignment padding (40b01b9bbdf51ae543a04744283bf2d56c4a6afa)
libata: update ATAPI overflow draining bisect/good-e164094964e6e20fe7fce418e06a9dce952bb7a4

Serial console log of hung kernel 40b01b9bbdf51ae543a04744283bf2d56c4a6afa below

[    0.000000] Linux version 2.6.25-rc2-smp (root@homer) (gcc version 4.2.1 (SUSE Linux)) #14 SMP PREEMPT Thu Feb 21 08:49:51 CET 2008
[    0.000000] BIOS-provided physical RAM map:
[    0.000000]  BIOS-e820: 0000000000000000 - 000000000009fc00 (usable)
[    0.000000]  BIOS-e820: 000000000009fc00 - 00000000000a0000 (reserved)
[    0.000000]  BIOS-e820: 00000000000f0000 - 0000000000100000 (reserved)
[    0.000000]  BIOS-e820: 0000000000100000 - 000000003fff0000 (usable)
[    0.000000]  BIOS-e820: 000000003fff0000 - 000000003fff3000 (ACPI NVS)
[    0.000000]  BIOS-e820: 000000003fff3000 - 0000000040000000 (ACPI data)
[    0.000000]  BIOS-e820: 00000000fec00000 - 0000000100000000 (reserved)
[    0.000000] 0MB HIGHMEM available.
[    0.000000] 1023MB LOWMEM available.
[    0.000000] Scan SMP from b0000000 for 1024 bytes.
[    0.000000] Scan SMP from b009fc00 for 1024 bytes.
[    0.000000] Scan SMP from b00f0000 for 65536 bytes.
[    0.000000] found SMP MP-table at [b00f5320] 000f5320
[    0.000000] Zone PFN ranges:
[    0.000000]   DMA             0 ->     4096
[    0.000000]   Normal       4096 ->   262128
[    0.000000]   HighMem    262128 ->   262128
[    0.000000] Movable zone start PFN for each node
[    0.000000] early_node_map[1] active PFN ranges
[    0.000000]     0:        0 ->   262128
[    0.000000] DMI 2.3 present.
[    0.000000] ACPI: RSDP 000F6CC0, 0014 (r0 IntelR)
[    0.000000] ACPI: RSDT 3FFF3000, 002C (r1 IntelR AWRDACPI 42302E31 AWRD        0)
[    0.000000] ACPI: FACP 3FFF3040, 0074 (r1 IntelR AWRDACPI 42302E31 AWRD        0)
[    0.000000] ACPI: DSDT 3FFF30C0, 4139 (r1 INTELR AWRDACPI     1000 MSFT  100000E)
[    0.000000] ACPI: FACS 3FFF0000, 0040
[    0.000000] ACPI: APIC 3FFF7200, 0068 (r1 IntelR AWRDACPI 42302E31 AWRD        0)
[    0.000000] ACPI: PM-Timer IO Port: 0x408
[    0.000000] ACPI: LAPIC (acpi_id[0x00] lapic_id[0x00] enabled)
[    0.000000] Processor #0 15:2 APIC version 20
[    0.000000] ACPI: LAPIC (acpi_id[0x01] lapic_id[0x01] enabled)
[    0.000000] Processor #1 15:2 APIC version 20
[    0.000000] WARNING: maxcpus limit of 1 reached. Processor ignored.
[    0.000000] ACPI: LAPIC_NMI (acpi_id[0x00] high edge lint[0x1])
[    0.000000] ACPI: LAPIC_NMI (acpi_id[0x01] high edge lint[0x1])
[    0.000000] ACPI: IOAPIC (id[0x02] address[0xfec00000] gsi_base[0])
[    0.000000] IOAPIC[0]: apic_id 2, version 32, address 0xfec00000, GSI 0-23
[    0.000000] ACPI: INT_SRC_OVR (bus 0 bus_irq 0 global_irq 2 dfl dfl)
[    0.000000] ACPI: INT_SRC_OVR (bus 0 bus_irq 9 global_irq 9 high level)
[    0.000000] Enabling APIC mode:  Flat.  Using 1 I/O APICs
[    0.000000] Using ACPI (MADT) for SMP configuration information
[    0.000000] Allocating PCI resources starting at 50000000 (gap: 40000000:bec00000)
[    0.000000] PM: Registered nosave memory: 000000000009f000 - 00000000000a0000
[    0.000000] PM: Registered nosave memory: 00000000000a0000 - 00000000000f0000
[    0.000000] PM: Registered nosave memory: 00000000000f0000 - 0000000000100000
[    0.000000] Built 1 zonelists in Zone order, mobility grouping on.  Total pages: 260081
[    0.000000] Kernel command line: root=/dev/sdb3 rootflags=data=writeback vga=0x314 resume=/dev/sdb2 console=ttyS0,115200n8 console=tty splash=silent PROFILE=default 1 maxcpus=1
[    0.000000] Enabling fast FPU save and restore... done.
[    0.000000] Enabling unmasked SIMD FPU exception support... done.
[    0.000000] Initializing CPU#0
[    0.000000] Preemptible RCU implementation.
[    0.000000] CPU 0 irqstacks, hard=b0427000 soft=b0425000
[    0.000000] PID hash table entries: 4096 (order: 12, 16384 bytes)
[    0.000000] Detected 2992.603 MHz processor.
[    0.000999] Console: colour dummy device 80x25
[    0.000999] console [tty0] enabled
[    0.000999] console [ttyS0] enabled
[    0.000999] Dentry cache hash table entries: 131072 (order: 7, 524288 bytes)
[    0.000999] Inode-cache hash table entries: 65536 (order: 6, 262144 bytes)
[    0.000999] Memory: 1028968k/1048512k available (1998k kernel code, 18904k reserved, 955k data, 236k init, 0k highmem)
[    0.000999] virtual kernel memory layout:
[    0.000999]     fixmap  : 0xfff9b000 - 0xfffff000   ( 400 kB)
[    0.000999]     pkmap   : 0xff800000 - 0xffc00000   (4096 kB)
[    0.000999]     vmalloc : 0xf0800000 - 0xff7fe000   ( 239 MB)
[    0.000999]     lowmem  : 0xb0000000 - 0xefff0000   (1023 MB)
[    0.000999]       .init : 0xb03e7000 - 0xb0422000   ( 236 kB)
[    0.000999]       .data : 0xb02f3b26 - 0xb03e29a8   ( 955 kB)
[    0.000999]       .text : 0xb0100000 - 0xb02f3b26   (1998 kB)
[    0.000999] Checking if this processor honours the WP bit even in supervisor mode...Ok.
[    0.060993] Calibrating delay using timer specific routine.. 5987.55 BogoMIPS (lpj=2993775)
[    0.063022] Security Framework initialized
[    0.064010] Mount-cache hash table entries: 512
[    0.065129] CPU: Trace cache: 12K uops, L1 D cache: 8K
[    0.066992] CPU: L2 cache: 512K
[    0.067992] CPU: Physical Processor ID: 0
[    0.068993] Intel machine check architecture supported.
[    0.069994] Intel machine check reporting enabled on CPU#0.
[    0.070991] CPU0: Intel P4/Xeon Extended MCE MSRs (12) available
[    0.071992] CPU0: Thermal monitoring enabled
[    0.072993] Compat vDSO mapped to ffffe000.
[    0.073996] Checking 'hlt' instruction... OK.
[    0.079743] SMP alternatives: switching to UP code
[    0.079998] Freeing SMP alternatives: 9k freed
[    0.080991] ACPI: Core revision 20070126
[    0.091025] CPU0: Intel(R) Pentium(R) 4 CPU 3.00GHz stepping 09
[    0.094019] Total of 1 processors activated (5987.55 BogoMIPS).
[    0.095110] ENABLING IO-APIC IRQs
[    0.096152] ..TIMER: vector=0x31 apic1=0 pin1=2 apic2=-1 pin2=-1
[    0.107983] Brought up 1 CPUs
[    0.108208] net_namespace: 552 bytes
[    0.110058] NET: Registered protocol family 16
[    0.111195] ACPI: bus type pci registered
[    0.114122] PCI: PCI BIOS revision 2.10 entry at 0xfb980, last bus=2
[    0.114984] PCI: Using configuration type 1
[    0.115983] Setting up standard PCI resources
[    0.139635] ACPI: Interpreter enabled
[    0.139985] ACPI: (supports S0 S3 S4 S5)
[    0.142261] ACPI: Using IOAPIC for interrupt routing
[    0.148393] ACPI: PCI Root Bridge [PCI0] (0000:00)
[    0.149406] pci 0000:00:1f.0: quirk: region 0400-047f claimed by ICH4 ACPI/GPIO/TCO
[    0.149981] pci 0000:00:1f.0: quirk: region 0480-04bf claimed by ICH4 GPIO
[    0.151395] PCI: Transparent bridge - 0000:00:1e.0
[    0.160727] ACPI: PCI Interrupt Link [LNKA] (IRQs 3 4 *5 7 9 10 11 12 14 15)
[    0.164091] ACPI: PCI Interrupt Link [LNKB] (IRQs *3 4 5 7 9 10 11 12 14 15)
[    0.168383] ACPI: PCI Interrupt Link [LNKC] (IRQs 3 4 *5 7 9 10 11 12 14 15)
[    0.171666] ACPI: PCI Interrupt Link [LNKD] (IRQs *3 4 5 7 9 10 11 12 14 15)
[    0.175277] ACPI: PCI Interrupt Link [LNKE] (IRQs 3 4 5 7 9 10 *11 12 14 15)
[    0.179656] ACPI: PCI Interrupt Link [LNKF] (IRQs 3 4 5 7 *9 10 11 12 14 15)
[    0.183050] ACPI: PCI Interrupt Link [LNK0] (IRQs 3 4 5 7 9 10 11 12 14 15) *0, disabled.
[    0.188275] ACPI: PCI Interrupt Link [LNK1] (IRQs 3 4 5 7 9 10 *11 12 14 15)
[    0.192014] Linux Plug and Play Support v0.97 (c) Adam Belay
[    0.193005] pnp: PnP ACPI init
[    0.193979] ACPI: bus type pnp registered
[    0.199104] pnpacpi: exceeded the max number of mem resources: 12
[    0.200035] pnp: PnP ACPI: found 13 devices
[    0.200971] ACPI: ACPI bus type pnp unregistered
[    0.202306] PCI: Using ACPI for IRQ routing
[    0.202977] PCI: If a device doesn't work, try "pci=routeirq".  If it helps, post a report
[    0.226967] NetLabel: Initializing
[    0.226969] NetLabel:  domain hash size = 128
[    0.227968] NetLabel:  protocols = UNLABELED CIPSOv4
[    0.228981] NetLabel:  unlabeled traffic allowed by default
[    0.230055] hpet0: at MMIO 0xfed00000, IRQs 2, 8, 11
[    0.232158] hpet0: 3 64-bit timers, 14318180 Hz
[    0.235002] ACPI: RTC can wake from S4
[    0.235966] Time: tsc clocksource has been installed.
[    0.243966] system 00:01: ioport range 0xb78-0xb7b has been reserved
[    0.243969] system 00:01: ioport range 0xf78-0xf7b has been reserved
[    0.244971] system 00:01: ioport range 0xa78-0xa7b has been reserved
[    0.245968] system 00:01: ioport range 0xe78-0xe7b has been reserved
[    0.246967] system 00:01: ioport range 0xbbc-0xbbf has been reserved
[    0.247967] system 00:01: ioport range 0xfbc-0xfbf has been reserved
[    0.248968] system 00:01: ioport range 0x4d0-0x4d1 has been reserved
[    0.249967] system 00:01: ioport range 0x200-0x200 has been reserved
[    0.250967] system 00:01: ioport range 0x202-0x208 has been reserved
[    0.251967] system 00:01: ioport range 0x320-0x32f has been reserved
[    0.252966] system 00:01: ioport range 0x295-0x296 has been reserved
[    0.254970] system 00:0b: ioport range 0x400-0x4bf could not be reserved
[    0.255973] system 00:0c: iomem range 0xf0000-0xf3fff could not be reserved
[    0.256966] system 00:0c: iomem range 0xf4000-0xf7fff could not be reserved
[    0.257966] system 00:0c: iomem range 0xf8000-0xfbfff could not be reserved
[    0.258966] system 00:0c: iomem range 0xfc000-0xfffff could not be reserved
[    0.259966] system 00:0c: iomem range 0x3fff0000-0x3fffffff could not be reserved
[    0.260965] system 00:0c: iomem range 0x0-0x9ffff could not be reserved
[    0.261965] system 00:0c: iomem range 0x100000-0x3ffeffff could not be reserved
[    0.262965] system 00:0c: iomem range 0xfec00000-0xfec00fff could not be reserved
[    0.263965] system 00:0c: iomem range 0xfec01000-0xfed8ffff could not be reserved
[    0.264965] system 00:0c: iomem range 0xfee00000-0xfee00fff could not be reserved
[    0.265965] system 00:0c: iomem range 0xffb00000-0xffbfffff could not be reserved
[    0.266964] system 00:0c: iomem range 0xfff00000-0xffffffff could not be reserved
[    0.298911] PCI: Bridge: 0000:00:01.0
[    0.298960]   IO window: a000-afff
[    0.299963]   MEM window: 0xf8000000-0xf9ffffff
[    0.300961]   PREFETCH window: 0x00000000e8000000-0x00000000f7ffffff
[    0.301963] PCI: Bridge: 0000:00:1e.0
[    0.302959]   IO window: b000-bfff
[    0.303961]   MEM window: 0xfa000000-0xfa0fffff
[    0.304960]   PREFETCH window: disabled.
[    0.305994] NET: Registered protocol family 2
[    0.325957] IP route cache hash table entries: 32768 (order: 5, 131072 bytes)
[    0.326233] TCP established hash table entries: 131072 (order: 8, 1048576 bytes)
[    0.328385] TCP bind hash table entries: 65536 (order: 7, 524288 bytes)
[    0.329394] TCP: Hash tables configured (established 131072 bind 65536)
[    0.329963] TCP reno registered
[    0.337961] Unpacking initramfs... done
[    0.574326] Freeing initrd memory: 6128k freed
[    0.576000] Machine check exception polling timer started.
[    0.577268] audit: initializing netlink socket (disabled)
[    0.577937] type=2000 audit(1203584121.788:1): initialized
[    0.579060] Total HugeTLB memory allocated, 0
[    0.579988] VFS: Disk quotas dquot_6.5.1
[    0.580947] Dquot-cache hash table entries: 1024 (order 0, 4096 bytes)
[    0.582937] Block layer SCSI generic (bsg) driver version 0.4 loaded (major 253)
[    0.583930] io scheduler noop registered
[    0.584925] io scheduler anticipatory registered
[    0.585924] io scheduler deadline registered
[    0.586932] io scheduler cfq registered (default)
[    0.589129] vesafb: framebuffer at 0xe8000000, mapped to 0xf0880000, using 1875k, total 16384k
[    0.589925] vesafb: mode is 800x600x16, linelength=1600, pages=16
[    0.590924] vesafb: protected mode interface info at c000:b544
[    0.591926] vesafb: pmi: set display start = b00cb5d2, set palette = b00cb612
[    0.592923] vesafb: scrolling: redraw
[    0.593924] vesafb: Truecolor: size=0:5:6:5, shift=0:11:5:0
[    0.612485] Console: switching to colour frame buffer device 100x37
[    0.627919] fb0: VESA VGA frame buffer device
[    0.638247] Serial: 8250/16550 driver $Revision: 1.90 $ 4 ports, IRQ sharing enabled
[    0.639029] serial8250: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A
[    0.641591] 00:07: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A
[    0.642200] PNP: PS/2 Controller [PNP0303:PS2K] at 0x60,0x64 irq 1
[    0.642917] PNP: PS/2 appears to have AUX port disabled, if this is incorrect please boot with i8042.nopnp
[    0.644280] serio: i8042 KBD port at 0x60,0x64 irq 1
[    0.645234] mice: PS/2 mouse device common for all mice
[    0.675685] input: AT Translated Set 2 keyboard as /devices/platform/i8042/serio0/input/input0
[    0.687684] rtc_cmos 00:03: rtc core: registered rtc_cmos as rtc0
[    0.687709] rtc0: alarms up to one month
[    0.688747] cpuidle: using governor ladder
[    0.689689] cpuidle: using governor menu
[    0.690751] oprofile: using NMI interrupt.
[    0.693003] NET: Registered protocol family 1
[    0.693759] p4-clockmod: P4/Xeon(TM) CPU On-Demand Clock Modulation available
[    0.694686] Using IPI No-Shortcut mode
[    0.695814] registered taskstats version 1
[    0.696813] rtc_cmos 00:03: setting system clock to 2008-02-21 08:55:23 UTC (1203584123)
[    0.698738] Freeing unused kernel memory: 236k freed
[    0.699701] Write protecting the kernel text: 2000k
[    0.700690] Write protecting the kernel read-only data: 792k
[    0.762206] ACPI: ACPI0007:00 is registered as cooling_device0
[    0.768027] ACPI: LNXTHERM:01 is registered as thermal_zone0
[    0.768870] ACPI: Thermal Zone [THRM] (40 C)
[    0.785027] SCSI subsystem initialized
[    0.806767] ACPI: PCI Interrupt 0000:00:1f.1[A] -> GSI 18 (level, low) -> IRQ 18
[    0.808835] scsi0 : ata_piix
[    0.809721] scsi1 : ata_piix
[    0.812130] ata1: PATA max UDMA/100 cmd 0x1f0 ctl 0x3f6 bmdma 0xf000 irq 14
[    0.812702] ata2: PATA max UDMA/100 cmd 0x170 ctl 0x376 bmdma 0xf008 irq 15
[    1.290553] ata1.00: ATA-6: ST3160021A, 3.04, max UDMA/100
[    1.290558] ata1.00: 312581808 sectors, multi 16: LBA48 
[    1.291580] ata1.01: ATAPI: BENQ    DVD DD DW1625, BBIA, max UDMA/33
[    1.314801] ata1.00: configured for UDMA/100
[    1.473512] ata1.01: configured for UDMA/33
[    3.788245] ata2.00: ATA-6: ST3120022A, 3.06, max UDMA/100
[    3.788248] ata2.00: 234441648 sectors, multi 16: LBA48 
[    3.811575] ata2.00: configured for UDMA/100
[    3.823087] scsi 0:0:0:0: Direct-Access     ATA      ST3160021A       3.04 PQ: 0 ANSI: 5
[    3.825453] scsi 0:0:1:0: CD-ROM            BENQ     DVD DD DW1625    BBIA PQ: 0 ANSI: 5
[    3.825587] scsi 1:0:0:0: Direct-Access     ATA      ST3120022A       3.06 PQ: 0 ANSI: 5
[    3.831138] ACPI: PNP0C0B:00 is registered as cooling_device1
[    3.831481] ACPI: Fan [FAN] (on)
[    3.856766] BIOS EDD facility v0.16 2004-Jun-25, 6 devices found
[    3.999333] Driver 'sd' needs updating - please use bus_type methods
[    3.999551] sd 0:0:0:0: [sda] 312581808 512-byte hardware sectors (160042 MB)
[    4.000464] sd 0:0:0:0: [sda] Write Protect is off
[    4.001459] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[    4.002528] sd 0:0:0:0: [sda] 312581808 512-byte hardware sectors (160042 MB)
[    4.003475] sd 0:0:0:0: [sda] Write Protect is off
[    4.004458] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[    4.005586]  sda: sda1 sda2 < sda5 sda6 >
[    4.048569] sd 0:0:0:0: [sda] Attached SCSI disk
[    4.049519] sd 1:0:0:0: [sdb] 234441648 512-byte hardware sectors (120034 MB)
[    4.050439] sd 1:0:0:0: [sdb] Write Protect is off
[    4.051473] sd 1:0:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[    4.052505] sd 1:0:0:0: [sdb] 234441648 512-byte hardware sectors (120034 MB)
[    4.053436] sd 1:0:0:0: [sdb] Write Protect is off
[    4.054450] sd 1:0:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[    4.055498]  sdb: sdb1 sdb2 sdb3
[    4.063528] sd 1:0:0:0: [sdb] Attached SCSI disk
[   37.473476] SysRq : Show State
[   37.474338]   task                PC stack   pid father
[   37.474338] init          S ef836eac     0     1      0
[   37.474338]        ef836f00 00000082 ffffffff ef836eac b01224d3 00000000 00000000 00000001 
[   37.474338]        e80cb97b 00000000 b0421180 b0421180 b0421180 ef835020 ef83527c b180b180 
[   37.474338]        00000000 ef836000 ee6e0c80 b1031ae0 00000000 00000000 ee5b9148 ef836f14 
[   37.474338] Call Trace:
[   37.474338]  [<b01224d3>] ? hrtick_set+0x7f/0xf7
[   37.474338]  [<b016de52>] ? do_wp_page+0x2a0/0x3e5
[   37.474338]  [<b01d218e>] ? security_task_wait+0xf/0x11
[   37.474338]  [<b0129e1f>] do_wait+0x470/0xa1a
[   37.474338]  [<b01249e9>] ? wake_up_new_task+0x77/0x91
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b012a431>] sys_wait4+0x68/0x9f
[   37.474338]  [<b012a48f>] sys_waitpid+0x27/0x29
[   37.474338]  [<b0104d52>] syscall_call+0x7/0xb
[   37.474338]  [<b02f0000>] ? relay_hotcpu_callback+0x51/0xb1
[   37.474338]  =======================
[   37.474338] kthreadd      S 00000000     0     2      0
[   37.474338]        ef83bfc8 00000046 ef8b50a0 00000000 00000092 ef83bf80 00000000 00000000 
[   37.474338]        ee5f776c 00000000 b0421180 b0421180 b0421180 ef83a0a0 ef83a2fc b180b180 
[   37.474338]        00000000 ef83b000 ee564d00 00000000 00000ae7 00000000 ee5f6e1c ee5f6e20 
[   37.474338] Call Trace:
[   37.474338]  [<b011d295>] ? complete+0x43/0x4b
[   37.474338]  [<b0139450>] kthreadd+0x13a/0x13f
[   37.474338]  [<b0139316>] ? kthreadd+0x0/0x13f
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] migration/0   S 0102c262     0     3      2
[   37.474338]        ef83df98 00000046 ef83c120 0102c262 00000246 ef83df4c 00000000 ef83df4c 
[   37.474338]        066fb2fc 00000000 b0421180 b0421180 b0421180 ef83c120 ef83c37c b180b180 
[   37.474338]        00000000 ef83d000 b03ba200 00000000 00000000 00000000 00000000 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b0122f2a>] ? migration_thread+0x0/0x210
[   37.474338]  [<b012304e>] migration_thread+0x124/0x210
[   37.474338]  [<b0122f2a>] ? migration_thread+0x0/0x210
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] ksoftirqd/0   R running      0     4      2
[   37.474338]        f4744940 00000008 ef841f14 b01424d9 f4650f99 00000008 00000000 01808638 
[   37.474338]        b18086c0 b1808638 b1808600 ef841f50 b013ca83 b1808600 b1808604 f4744940 
[   37.474338]        00000008 f465092e 00000008 f465092e 00000046 b1807120 00000000 00000246 
[   37.474338] Call Trace:
[   37.474338]  [<b01424d9>] ? tick_program_event+0x3f/0x61
[   37.474338]  [<b013ca83>] ? hrtimer_interrupt+0x13f/0x164
[   37.474338]  [<b012c055>] ? irq_exit+0x3f/0x79
[   37.474338]  [<b0113103>] ? smp_apic_timer_interrupt+0x5c/0x89
[   37.474338]  [<b012c055>] ? irq_exit+0x3f/0x79
[   37.474338]  [<b01057bc>] ? apic_timer_interrupt+0x28/0x30
[   37.474338]  [<b01079d9>] ? do_softirq+0x35/0x8c
[   37.474338]  [<b012c515>] ? ksoftirqd+0xad/0x17f
[   37.474338]  [<b012c468>] ? ksoftirqd+0x0/0x17f
[   37.474338]  [<b01392f4>] ? kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] ? kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] events/0      R running      0     5      2
[   37.474338]        ef843fa0 00000046 ef8112c0 ef843f44 b0136d0e 00000a75 00000000 00000a75 
[   37.474338]        60e9afad 00000008 b0421180 b0421180 b0421180 ef8420a0 ef8422fc b180b180 
[   37.474338]        00000000 ef843000 ee6e0880 b18089e0 3399ec88 00000002 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b0136d0e>] ? queue_delayed_work+0x40/0x48
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] khelper       S b02f12e5     0     6      2
[   37.474338]        ef845fa0 00000046 ef845f3c b02f12e5 ef836d40 ef836d44 00000000 ef845f44 
[   37.474338]        2a9ba799 00000000 b0421180 b0421180 b0421180 ef844120 ef84437c b180b180 
[   37.474338]        00000000 ef845000 ee6e0880 ef822480 00000000 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b02f12e5>] ? preempt_schedule+0x40/0x55
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] kblockd/0     S ef887f4c     0    35      2
[   37.474338]        ef887fa0 00000046 00000246 ef887f4c b01224ed ef887f4c 00000000 00000000 
[   37.474338]        0800f107 00000000 b0421180 b0421180 b0421180 ef886120 ef88637c b180b180 
[   37.474338]        00000000 ef887000 b03ba200 08006f94 00000c39 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b01224ed>] ? hrtick_set+0x99/0xf7
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] kacpid        S ef88cf4c     0    37      2
[   37.474338]        ef88cfa0 00000046 00000246 ef88cf4c b01224ed ef88cf4c 00000000 00000000 
[   37.474338]        0801a506 00000000 b0421180 b0421180 b0421180 ef88b0a0 ef88b2fc b180b180 
[   37.474338]        00000000 ef88c000 b03ba200 08018266 00000000 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b01224ed>] ? hrtick_set+0x99/0xf7
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] kacpi_notify  S ef88ff4c     0    38      2
[   37.474338]        ef88ffa0 00000046 00000246 ef88ff4c b01224ed ef88ff4c 00000000 00000000 
[   37.474338]        0886f6cd 00000000 b0421180 b0421180 b0421180 ef88e120 ef88e37c b180b180 
[   37.474338]        00000000 ef88f000 b03ba200 0801c44e 00000000 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b01224ed>] ? hrtick_set+0x99/0xf7
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] cqueue/0      S ef8e0f4c     0   111      2
[   37.474338]        ef8e0fa0 00000046 00000246 ef8e0f4c b01224ed ef8e0f4c 00000000 00000000 
[   37.474338]        0c0b0eaf 00000000 b0421180 b0421180 b0421180 ef8bd0a0 ef8bd2fc b180b180 
[   37.474338]        00000000 ef8e0000 b03ba200 0c0abebc 00000cf2 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b01224ed>] ? hrtick_set+0x99/0xf7
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] kseriod       S b011d321     0   115      2
[   37.474338]        ef8e4f8c 00000046 ef8e4f34 b011d321 00000000 00000000 00000000 ef88d440 
[   37.474338]        28fd3c39 00000000 b0421180 b0421180 b0421180 ef874120 ef87437c b180b180 
[   37.474338]        00000000 ef8e4000 b03ba200 00000000 00000000 00000000 b03cc440 ef88d43c 
[   37.474338] Call Trace:
[   37.474338]  [<b011d321>] ? __wake_up+0x3a/0x42
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0255c33>] serio_thread+0xc2/0x32f
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0255b71>] ? serio_thread+0x0/0x32f
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] pdflush       S 00000000     0   148      2
[   37.474338]        ee52cfa4 00000046 b180b180 00000000 ee52c000 ee52cf68 00000000 000001fa 
[   37.474338]        22833768 00000000 b0421180 b0421180 b0421180 ef864120 ef86437c b180b180 
[   37.474338]        00000000 ee52c000 b03ba200 b012229c 00000000 00000000 ef864120 b180b180 
[   37.474338] Call Trace:
[   37.474338]  [<b012229c>] ? set_cpus_allowed+0x50/0xb8
[   37.474338]  [<b02f2c57>] ? _spin_unlock_irqrestore+0x1f/0x21
[   37.474338]  [<b011eb19>] ? set_user_nice+0xcf/0xdf
[   37.474338]  [<b0167dd1>] ? pdflush+0x0/0x1b4
[   37.474338]  [<b0167e82>] pdflush+0xb1/0x1b4
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] pdflush       S 00000286     0   149      2
[   37.474338]        ee52dfa4 00000046 b03c16e0 00000286 ee52df54 b0130184 00000000 00000000 
[   37.474338]        60e9f95c 00000008 b0421180 b0421180 b0421180 ef8620a0 ef8622fc b180b180 
[   37.474338]        00000000 ee52d000 ee6e0680 00000000 00000000 00000000 00000000 00000000 
[   37.474338] Call Trace:
[   37.474338]  [<b0130184>] ? __mod_timer+0xa0/0xaf
[   37.474338]  [<b0167dd1>] ? pdflush+0x0/0x1b4
[   37.474338]  [<b0167e82>] pdflush+0xb1/0x1b4
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] kswapd0       S 00000000     0   150      2
[   37.474338]        ee52ef2c 00000046 00000000 00000000 00000000 00000000 00000000 000004fa 
[   37.474338]        22918b1f 00000000 b0421180 b0421180 b0421180 ef85e020 ef85e27c b180b180 
[   37.474338]        00000000 ee52e000 b03ba200 b012229c 00000000 00000000 b1807b00 b180b180 
[   37.474338] Call Trace:
[   37.474338]  [<b012229c>] ? set_cpus_allowed+0x50/0xb8
[   37.474338]  [<b011ab2c>] ? __dequeue_entity+0x31/0x35
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b016aefe>] ? kswapd+0x0/0x4a2
[   37.474338]  [<b016b38e>] kswapd+0x490/0x4a2
[   37.474338]  [<b02f0d71>] ? schedule+0x34e/0x82e
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b011d295>] ? complete+0x43/0x4b
[   37.474338]  [<b016aefe>] ? kswapd+0x0/0x4a2
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] aio/0         S ee530f4c     0   151      2
[   37.474338]        ee530fa0 00000046 00000246 ee530f4c b01224ed ee530f4c 00000000 00000000 
[   37.474338]        22bec65c 00000000 b0421180 b0421180 b0421180 ef85c120 ef85c37c b180b180 
[   37.474338]        00000000 ee530000 b03ba200 2291af16 00000000 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b01224ed>] ? hrtick_set+0x99/0xf7
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] kpsmoused     S ee653f4c     0   378      2
[   37.474338]        ee653fa0 00000046 00000246 ee653f4c b01224ed ee653f4c 00000000 00000000 
[   37.474338]        26811a10 00000000 b0421180 b0421180 b0421180 ee6ba020 ee6ba27c b180b180 
[   37.474338]        00000000 ee653000 b03ba200 2680eca0 00000000 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b01224ed>] ? hrtick_set+0x99/0xf7
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] kondemand/0   S ee5b0f4c     0   384      2
[   37.474338]        ee5b0fa0 00000046 00000246 ee5b0f4c b01224ed ee5b0f4c 00000000 00000000 
[   37.474338]        292b36fc 00000000 b0421180 b0421180 b0421180 ee5cc120 ee5cc37c b180b180 
[   37.474338]        00000000 ee5b0000 b03ba200 290d4c43 00000b94 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b01224ed>] ? hrtick_set+0x99/0xf7
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] blogd         S 00000000     0   423      1
[   37.474338]        ee6d0b10 00000082 000004b3 00000000 b1807120 15b2b2c0 00000000 ef8420a0 
[   37.474338]        60e9cf9d 00000008 b0421180 b0421180 b0421180 ee5e60a0 ee5e62fc b180b180 
[   37.474338]        00000000 ee6d0000 ee6e0680 b0130059 339a08bf 00000002 ee6d0b20 00000286 
[   37.474338] Call Trace:
[   37.474338]  [<b0130059>] ? lock_timer_base+0x1f/0x40
[   37.474338]  [<b0130184>] ? __mod_timer+0xa0/0xaf
[   37.474338]  [<b02f14ba>] schedule_timeout+0x44/0xa4
[   37.474338]  [<b01397ad>] ? add_wait_queue+0x2f/0x36
[   37.474338]  [<b012fcda>] ? process_timeout+0x0/0xa
[   37.474338]  [<b02f14b5>] ? schedule_timeout+0x3f/0xa4
[   37.474338]  [<b018dce0>] do_select+0x4b6/0x53c
[   37.474338]  [<b018e25a>] ? __pollwait+0x0/0xcb
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b01224d3>] ? hrtick_set+0x7f/0xf7
[   37.474338]  [<b02f0d71>] ? schedule+0x34e/0x82e
[   37.474338]  [<b020e60c>] ? cfb_fillrect+0x138/0x2bd
[   37.474338]  [<b013c088>] ? enqueue_hrtimer+0x75/0xf3
[   37.474338]  [<b0109fbc>] ? read_tsc+0x8/0xa
[   37.474338]  [<b013e472>] ? getnstimeofday+0x34/0xdf
[   37.474338]  [<b013c90f>] ? ktime_get_ts+0x44/0x49
[   37.474338]  [<b013c925>] ? ktime_get+0x11/0x30
[   37.474338]  [<b011b82b>] ? hrtick_start_fair+0x10d/0x144
[   37.474338]  [<b011b900>] ? enqueue_task_fair+0x52/0x56
[   37.474338]  [<b011a3f1>] ? enqueue_task+0x4c/0x58
[   37.474338]  [<b011eb76>] ? try_to_wake_up+0x4d/0x1be
[   37.474338]  [<b018df10>] core_sys_select+0x1aa/0x2bb
[   37.474338]  [<b013c6ed>] ? hrtimer_start+0xc7/0x13c
[   37.474338]  [<b01224d3>] ? hrtick_set+0x7f/0xf7
[   37.474338]  [<b01042b4>] ? do_notify_resume+0x55/0x79e
[   37.474338]  [<b0127ad4>] ? release_console_sem+0x1c4/0x1d4
[   37.474338]  [<b011d321>] ? __wake_up+0x3a/0x42
[   37.474338]  [<b0232d6a>] ? tty_ldisc_deref+0x55/0x6e
[   37.474338]  [<b018e3fc>] sys_select+0xd7/0x1a2
[   37.474338]  [<b0104d52>] syscall_call+0x7/0xb
[   37.474338]  =======================
[   37.474338] blogd         R running      0   424      1
[   37.474338]        ee6cbdf8 00000082 b18086c0 ee6cbda0 b01e99b7 b1808640 00000000 b180b5f4 
[   37.474338]        60e8dd8e 00000008 b0421180 b0421180 b0421180 ee5d70a0 ee5d72fc b180b180 
[   37.474338]        00000000 ee6cb000 ee6e0680 ee6cbdf8 375f44e1 00000003 ee6cbdec b041e600 
[   37.474338] Call Trace:
[   37.474338]  [<b01e99b7>] ? rb_insert_color+0x77/0xd8
[   37.474338]  [<b0143bfe>] futex_wait+0x285/0x2d3
[   37.474338]  [<b013c31c>] ? hrtimer_wakeup+0x0/0x1c
[   37.474338]  [<b0143b3d>] ? futex_wait+0x1c4/0x2d3
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b0144c2d>] do_futex+0x20d/0xa5f
[   37.474338]  [<b0168667>] ? pagevec_lookup_tag+0x25/0x2e
[   37.474338]  [<b016158a>] ? wait_on_page_writeback_range+0x5b/0xf0
[   37.474338]  [<b0109fbc>] ? read_tsc+0x8/0xa
[   37.474338]  [<b013e472>] ? getnstimeofday+0x34/0xdf
[   37.474338]  [<b013c90f>] ? ktime_get_ts+0x44/0x49
[   37.474338]  [<b013c925>] ? ktime_get+0x11/0x30
[   37.474338]  [<b0145502>] sys_futex+0x83/0xe8
[   37.474338]  [<b01ec504>] ? copy_to_user+0x2a/0x36
[   37.474338]  [<b0104d52>] syscall_call+0x7/0xb
[   37.474338]  =======================
[   37.474338] ata/0         S ee7b3f8c     0   438      2
[   37.474338]        ee7b3fa0 00000046 00000002 ee7b3f8c ee7b3f80 ee7b3f44 00000000 002dae6f 
[   37.474338]        e3f3f21f 00000000 b0421180 b0421180 b0421180 ee6be120 ee6be37c b180b180 
[   37.474338]        00000000 ee7b3000 ee6e0a80 ef906480 00000000 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] ata_aux       S ee7b2f4c     0   439      2
[   37.474338]        ee7b2fa0 00000046 ffffffff ee7b2f4c b01224d3 00000000 00000000 00000000 
[   37.474338]        2fc09ed2 00000000 b0421180 b0421180 b0421180 ee5dd020 ee5dd27c b180b180 
[   37.474338]        00000000 ee7b2000 ee6e0280 2f81ef36 00000000 00000000 b0421180 b0421180 
[   37.474338] Call Trace:
[   37.474338]  [<b01224d3>] ? hrtick_set+0x7f/0xf7
[   37.474338]  [<b0139711>] ? prepare_to_wait+0x48/0x4d
[   37.474338]  [<b0136df5>] worker_thread+0xb9/0xd7
[   37.474338]  [<b01395a9>] ? autoremove_wake_function+0x0/0x38
[   37.474338]  [<b0136d3c>] ? worker_thread+0x0/0xd7
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] scsi_eh_0     S ef906520     0   445      2
[   37.474338]        ee585f64 00000046 ef906524 ef906520 00000000 00000092 00000000 b011d321 
[   37.474338]        e4331293 00000000 b0421180 b0421180 b0421180 ee5e4020 ee5e427c b180b180 
[   37.474338]        00000000 ee585000 ee6e0480 ee702008 00108a29 00000000 f084c295 ee702000 
[   37.474338] Call Trace:
[   37.474338]  [<b011d321>] ? __wake_up+0x3a/0x42
[   37.474338]  [<f084c295>] ? __scsi_iterate_devices+0x5d/0x6b [scsi_mod]
[   37.474338]  [<f0850303>] ? scsi_error_handler+0x0/0x4eb [scsi_mod]
[   37.474338]  [<f0850303>] ? scsi_error_handler+0x0/0x4eb [scsi_mod]
[   37.474338]  [<f085033a>] scsi_error_handler+0x37/0x4eb [scsi_mod]
[   37.474338]  [<b011d295>] ? complete+0x43/0x4b
[   37.474338]  [<f0850303>] ? scsi_error_handler+0x0/0x4eb [scsi_mod]
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] scsi_eh_1     S ee622520     0   446      2
[   37.474338]        ee5acf64 00000046 ee622524 ee622520 00000000 00000092 00000000 b011d321 
[   37.474338]        e4333e8f 00000000 b0421180 b0421180 b0421180 ee5d10a0 ee5d12fc b180b180 
[   37.474338]        00000000 ee5ac000 ee6e0480 ee702808 00000000 00000000 f084c295 ee702800 
[   37.474338] Call Trace:
[   37.474338]  [<b011d321>] ? __wake_up+0x3a/0x42
[   37.474338]  [<f084c295>] ? __scsi_iterate_devices+0x5d/0x6b [scsi_mod]
[   37.474338]  [<f0850303>] ? scsi_error_handler+0x0/0x4eb [scsi_mod]
[   37.474338]  [<f0850303>] ? scsi_error_handler+0x0/0x4eb [scsi_mod]
[   37.474338]  [<f085033a>] scsi_error_handler+0x37/0x4eb [scsi_mod]
[   37.474338]  [<b011d295>] ? complete+0x43/0x4b
[   37.474338]  [<f0850303>] ? scsi_error_handler+0x0/0x4eb [scsi_mod]
[   37.474338]  [<b01392f4>] kthread+0x37/0x59
[   37.474338]  [<b01392bd>] ? kthread+0x0/0x59
[   37.474338]  [<b010593f>] kernel_thread_helper+0x7/0x18
[   37.474338]  =======================
[   37.474338] udevd         S 00000000     0   461      1
[   37.474338]        ee5d8b10 00000082 00000000 00000000 00000000 00000000 00000000 ee5d8ab4 
[   37.474338]        93a637c5 00000001 b0421180 b0421180 b0421180 ee6c20a0 ee6c22fc b180b180 
[   37.474338]        00000000 ee5d8000 ee6e0480 b1807980 00000000 00000000 000f41a9 00000000 
[   37.474338] Call Trace:
[   37.474338]  [<b011b920>] ? __update_rq_clock+0x1c/0x157
[   37.474338]  [<b02f14e0>] schedule_timeout+0x6a/0xa4
[   37.474338]  [<b01397ad>] ? add_wait_queue+0x2f/0x36
[   37.474338]  [<b018e2c1>] ? __pollwait+0x67/0xcb
[   37.474338]  [<b0186ef4>] ? pipe_poll+0x29/0x8f
[   37.474338]  [<b018dce0>] do_select+0x4b6/0x53c
[   37.474338]  [<b013fdbc>] ? clocksource_get_next+0x3d/0x44
[   37.474338]  [<b018e25a>] ? __pollwait+0x0/0xcb
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b01414ad>] ? clockevents_program_event+0x93/0x108
[   37.474338]  [<b01424d9>] ? tick_program_event+0x3f/0x61
[   37.474338]  [<b01eab55>] ? number+0x2a3/0x2b5
[   37.474338]  [<b01057bc>] ? apic_timer_interrupt+0x28/0x30
[   37.474338]  [<b019007b>] ? fcntl_getlk64+0x4e/0x159
[   37.474338]  [<b01eb3a7>] ? vsnprintf+0x2e8/0x5ea
[   37.474338]  [<b0119234>] ? kmap_atomic_prot+0x47/0xa8
[   37.474338]  [<b0165eb1>] ? get_page_from_freelist+0x25f/0x443
[   37.474338]  [<b016625f>] ? __alloc_pages+0x57/0x32d
[   37.474338]  [<b016c409>] ? __inc_zone_page_state+0x18/0x1a
[   37.474338]  [<b018df10>] core_sys_select+0x1aa/0x2bb
[   37.474338]  [<b0165c04>] ? free_hot_page+0xa/0xc
[   37.474338]  [<b0168bf3>] ? put_page+0x2d/0xac
[   37.474338]  [<b0176340>] ? free_page_and_swap_cache+0x1e/0x3e
[   37.474338]  [<b016e44a>] ? unmap_vmas+0x317/0x54b
[   37.474338]  [<b0117410>] ? pgd_dtor+0x0/0x4a
[   37.474338]  [<b011740e>] ? check_pgt_cache+0x1e/0x20
[   37.474338]  [<b01711bf>] ? unmap_region+0xdc/0x12f
[   37.474338]  [<b018e35e>] sys_select+0x39/0x1a2
[   37.474338]  [<b0104d52>] syscall_call+0x7/0xb
[   37.474338]  =======================
[   37.474338] udevsettle    R running      0   465      1
[   37.474338]        ee524f1c 00000082 ee524ed4 00000000 00010542 b1808640 00000000 ee6cbe14 
[   37.474338]        60e8f831 00000008 b0421180 b0421180 b0421180 ee5ce020 ee5ce27c b180b180 
[   37.474338]        00000000 ee524000 ee6e0880 ee524f1c 33bd800d 00000003 00000008 b041e600 
[   37.474338] Call Trace:
[   37.474338]  [<b02f1a50>] do_nanosleep+0x70/0x9a
[   37.474338]  [<b013c7ae>] hrtimer_nanosleep+0x4c/0xaf
[   37.474338]  [<b013c31c>] ? hrtimer_wakeup+0x0/0x1c
[   37.474338]  [<b02f1a3d>] ? do_nanosleep+0x5d/0x9a
[   37.474338]  [<b013c868>] sys_nanosleep+0x57/0x5b
[   37.474338]  [<b0104d52>] syscall_call+0x7/0xb
[   37.474338]  [<b02f0000>] ? relay_hotcpu_callback+0x51/0xb1
[   37.474338]  =======================
[   37.474338] udevd         S 00000000     0   873    461
[   37.474338]        e7c7eb10 00000082 00000000 00000000 00000000 00000000 00000000 00000000 
[   37.474338]        9385214a 00000001 b0421180 b0421180 b0421180 ef8d40a0 ef8d42fc b180b180 
[   37.474338]        00000000 e7c7e000 ee6e0280 00000000 00000000 00000000 00000000 00000000 
[   37.474338] Call Trace:
[   37.474338]  [<b02f14e0>] schedule_timeout+0x6a/0xa4
[   37.474338]  [<b01397ad>] ? add_wait_queue+0x2f/0x36
[   37.474338]  [<b018e2c1>] ? __pollwait+0x67/0xcb
[   37.474338]  [<b0186ef4>] ? pipe_poll+0x29/0x8f
[   37.474338]  [<b018dce0>] do_select+0x4b6/0x53c
[   37.474338]  [<b018e25a>] ? __pollwait+0x0/0xcb
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b01920af>] ? __d_lookup+0xba/0x124
[   37.474338]  [<b01920af>] ? __d_lookup+0xba/0x124
[   37.474338]  [<b0119234>] ? kmap_atomic_prot+0x47/0xa8
[   37.474338]  [<b0165eb1>] ? get_page_from_freelist+0x25f/0x443
[   37.474338]  [<b018df10>] core_sys_select+0x1aa/0x2bb
[   37.474338]  [<b013c088>] ? enqueue_hrtimer+0x75/0xf3
[   37.474338]  [<b013c6ed>] ? hrtimer_start+0xc7/0x13c
[   37.474338]  [<b01224d3>] ? hrtick_set+0x7f/0xf7
[   37.474338]  [<b02f0d71>] ? schedule+0x34e/0x82e
[   37.474338]  [<b016de52>] ? do_wp_page+0x2a0/0x3e5
[   37.474338]  [<b016f0e7>] ? handle_mm_fault+0x442/0x5d4
[   37.474338]  [<b02f12e5>] ? preempt_schedule+0x40/0x55
[   37.474338]  [<b018e35e>] sys_select+0x39/0x1a2
[   37.474338]  [<b017fc6f>] ? filp_close+0x43/0x69
[   37.474338]  [<b01ec504>] ? copy_to_user+0x2a/0x36
[   37.474338]  [<b0104d52>] syscall_call+0x7/0xb
[   37.474338]  =======================
[   37.474338] udevd         S 00000000     0   876    461
[   37.474338]        ee780b10 00000086 00000000 00000000 00000000 00000000 00000000 00000000 
[   37.474338]        93a2c2db 00000001 b0421180 b0421180 b0421180 ef8a1120 ef8a137c b180b180 
[   37.474338]        00000000 ee780000 ee564100 00000000 00000000 00000000 00000000 00000000 
[   37.474338] Call Trace:
[   37.474338]  [<b02f14e0>] schedule_timeout+0x6a/0xa4
[   37.474338]  [<b01397ad>] ? add_wait_queue+0x2f/0x36
[   37.474338]  [<b018e2c1>] ? __pollwait+0x67/0xcb
[   37.474338]  [<b0186ef4>] ? pipe_poll+0x29/0x8f
[   37.474338]  [<b018dce0>] do_select+0x4b6/0x53c
[   37.474338]  [<b018e25a>] ? __pollwait+0x0/0xcb
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b011a8a0>] ? update_curr+0x12f/0x136
[   37.474338]  [<b012381e>] ? task_tick_fair+0x59/0x86
[   37.474338]  [<b01227b3>] ? scheduler_tick+0x268/0x3cc
[   37.474338]  [<b0109fbc>] ? read_tsc+0x8/0xa
[   37.474338]  [<b013e472>] ? getnstimeofday+0x34/0xdf
[   37.474338]  [<b01414ad>] ? clockevents_program_event+0x93/0x108
[   37.474338]  [<b01424d9>] ? tick_program_event+0x3f/0x61
[   37.474338]  [<b013ca83>] ? hrtimer_interrupt+0x13f/0x164
[   37.474338]  [<b012c055>] ? irq_exit+0x3f/0x79
[   37.474338]  [<b0113103>] ? smp_apic_timer_interrupt+0x5c/0x89
[   37.474338]  [<b01920af>] ? __d_lookup+0xba/0x124
[   37.474338]  [<b01920af>] ? __d_lookup+0xba/0x124
[   37.474338]  [<b0119234>] ? kmap_atomic_prot+0x47/0xa8
[   37.474338]  [<b0165eb1>] ? get_page_from_freelist+0x25f/0x443
[   37.474338]  [<b018df10>] core_sys_select+0x1aa/0x2bb
[   37.474338]  [<b013c088>] ? enqueue_hrtimer+0x75/0xf3
[   37.474338]  [<b013c6ed>] ? hrtimer_start+0xc7/0x13c
[   37.474338]  [<b01224d3>] ? hrtick_set+0x7f/0xf7
[   37.474338]  [<b02f0d71>] ? schedule+0x34e/0x82e
[   37.474338]  [<b016de52>] ? do_wp_page+0x2a0/0x3e5
[   37.474338]  [<b016f0e7>] ? handle_mm_fault+0x442/0x5d4
[   37.474338]  [<b02f12e5>] ? preempt_schedule+0x40/0x55
[   37.474338]  [<b018e35e>] sys_select+0x39/0x1a2
[   37.474338]  [<b017fc6f>] ? filp_close+0x43/0x69
[   37.474338]  [<b01ec504>] ? copy_to_user+0x2a/0x36
[   37.474338]  [<b0104d52>] syscall_call+0x7/0xb
[   37.474338]  =======================
[   37.474338] scsi_id       D 00000000     0   878    873
[   37.474338]        ee60bb7c 00000086 00000046 00000000 00000000 ef904000 00000000 ee66d910 
[   37.474338]        93796336 00000001 b0421180 b0421180 b0421180 ee5d3120 ee5d337c b180b180 
[   37.474338]        00000000 ee60b000 ee564700 ee5c2800 0000065a 00000000 ee60bb5c b024e1b8 
[   37.474338] Call Trace:
[   37.474338]  [<b024e1b8>] ? put_device+0xf/0x11
[   37.474338]  [<f0852878>] ? scsi_request_fn+0x218/0x345 [scsi_mod]
[   37.474338]  [<b02f14e0>] schedule_timeout+0x6a/0xa4
[   37.474338]  [<b01d8b08>] ? elv_insert+0xf3/0x20b
[   37.474338]  [<b01302da>] ? mod_timer+0x26/0x3d
[   37.474338]  [<b01db15b>] ? blk_plug_device+0x42/0x9a
[   37.474338]  [<b02f08e5>] wait_for_common+0x74/0x12e
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b02f0a21>] wait_for_completion+0x12/0x14
[   37.474338]  [<b01dd237>] blk_execute_rq+0x5c/0x9c
[   37.474338]  [<b01dd277>] ? blk_end_sync_rq+0x0/0x29
[   37.474338]  [<b01a3fe2>] ? bio_add_pc_page+0x24/0x2a
[   37.474338]  [<b01d9672>] ? blk_rq_bio_prep+0x9e/0xb2
[   37.474338]  [<b01dceef>] ? blk_rq_append_bio+0x17/0x49
[   37.474338]  [<b01dd022>] ? blk_rq_map_user+0x101/0x1b4
[   37.474338]  [<b01e01f6>] sg_io+0x18c/0x2e1
[   37.474338]  [<b0165eb1>] ? get_page_from_freelist+0x25f/0x443
[   37.474338]  [<b01e05df>] scsi_cmd_ioctl+0x294/0x3d5
[   37.474338]  [<b01920af>] ? __d_lookup+0xba/0x124
[   37.474338]  [<b0188770>] ? do_lookup+0x5a/0x162
[   37.474338]  [<f08799b5>] sd_ioctl+0x82/0xc9 [sd_mod]
[   37.474338]  [<b01ddf93>] blkdev_driver_ioctl+0x55/0x5e
[   37.474338]  [<b01de1bf>] blkdev_ioctl+0x223/0x814
[   37.474338]  [<b01e7574>] ? kobject_get+0x12/0x17
[   37.474338]  [<b0160fc3>] ? find_lock_page+0x72/0x8d
[   37.474338]  [<b0163210>] ? filemap_fault+0x240/0x449
[   37.474338]  [<b01395a5>] ? wake_up_bit+0x17/0x1b
[   37.474338]  [<b0160e75>] ? unlock_page+0x25/0x28
[   37.474338]  [<b016d9b5>] ? __do_fault+0x17a/0x377
[   37.474338]  [<b01a57b6>] ? blkdev_open+0x28/0x58
[   37.474338]  [<b016eda8>] ? handle_mm_fault+0x103/0x5d4
[   37.474338]  [<b01a4c39>] block_ioctl+0x1b/0x21
[   37.474338]  [<b01a4c1e>] ? block_ioctl+0x0/0x21
[   37.474338]  [<b018cc12>] vfs_ioctl+0x22/0x71
[   37.474338]  [<b018cea9>] do_vfs_ioctl+0x248/0x290
[   37.474338]  [<b0117ba2>] ? do_page_fault+0x13d/0x5fb
[   37.474338]  [<b0189bfc>] ? putname+0x25/0x30
[   37.474338]  [<b01800e0>] ? do_sys_open+0xb1/0xc7
[   37.474338]  [<b018cf43>] sys_ioctl+0x52/0x63
[   37.474338]  [<b0104d52>] syscall_call+0x7/0xb
[   37.474338]  =======================
[   37.474338] scsi_id       D ee621bd4     0   879    876
[   37.474338]        ee636b7c 00000086 ee621bac ee621bd4 00000000 f084c4f9 00000000 ee66d0c8 
[   37.474338]        9394eaa1 00000001 b0421180 b0421180 b0421180 ef888020 ef88827c b180b180 
[   37.474338]        00000000 ee636000 ee564300 ee6e8c00 00000000 00000000 ee636b5c b024e1b8 
[   37.474338] Call Trace:
[   37.474338]  [<f084c4f9>] ? scsi_done+0x0/0x19 [scsi_mod]
[   37.474338]  [<b024e1b8>] ? put_device+0xf/0x11
[   37.474338]  [<f0852878>] ? scsi_request_fn+0x218/0x345 [scsi_mod]
[   37.474338]  [<b02f14e0>] schedule_timeout+0x6a/0xa4
[   37.474338]  [<b01d8b08>] ? elv_insert+0xf3/0x20b
[   37.474338]  [<b01302da>] ? mod_timer+0x26/0x3d
[   37.474338]  [<b01db15b>] ? blk_plug_device+0x42/0x9a
[   37.474338]  [<b02f08e5>] wait_for_common+0x74/0x12e
[   37.474338]  [<b011ece7>] ? default_wake_function+0x0/0xd
[   37.474338]  [<b02f0a21>] wait_for_completion+0x12/0x14
[   37.474338]  [<b01dd237>] blk_execute_rq+0x5c/0x9c
[   37.474338]  [<b01dd277>] ? blk_end_sync_rq+0x0/0x29
[   37.474338]  [<b01a3fe2>] ? bio_add_pc_page+0x24/0x2a
[   37.474338]  [<b01d9672>] ? blk_rq_bio_prep+0x9e/0xb2
[   37.474338]  [<b01dceef>] ? blk_rq_append_bio+0x17/0x49
[   37.474338]  [<b01dd022>] ? blk_rq_map_user+0x101/0x1b4
[   37.474338]  [<b01e01f6>] sg_io+0x18c/0x2e1
[   37.474338]  [<b0165eb1>] ? get_page_from_freelist+0x25f/0x443
[   37.474338]  [<b01e05df>] scsi_cmd_ioctl+0x294/0x3d5
[   37.474338]  [<b01920af>] ? __d_lookup+0xba/0x124
[   37.474338]  [<b0188770>] ? do_lookup+0x5a/0x162
[   37.474338]  [<f08799b5>] sd_ioctl+0x82/0xc9 [sd_mod]
[   37.474338]  [<b01ddf93>] blkdev_driver_ioctl+0x55/0x5e
[   37.474338]  [<b01de1bf>] blkdev_ioctl+0x223/0x814
[   37.474338]  [<b0168bbe>] ? activate_page+0xb1/0xb9
[   37.474338]  [<b0168cce>] ? mark_page_accessed+0x27/0x2e
[   37.474338]  [<b0163210>] ? filemap_fault+0x240/0x449
[   37.474338]  [<b01395a5>] ? wake_up_bit+0x17/0x1b
[   37.474338]  [<b0160e75>] ? unlock_page+0x25/0x28
[   37.474338]  [<b016d9b5>] ? __do_fault+0x17a/0x377
[   37.474338]  [<b01a57b6>] ? blkdev_open+0x28/0x58
[   37.474338]  [<b016eda8>] ? handle_mm_fault+0x103/0x5d4
[   37.474338]  [<b01a4c39>] block_ioctl+0x1b/0x21
[   37.474338]  [<b01a4c1e>] ? block_ioctl+0x0/0x21
[   37.474338]  [<b018cc12>] vfs_ioctl+0x22/0x71
[   37.474338]  [<b018cea9>] do_vfs_ioctl+0x248/0x290
[   37.474338]  [<b0117ba2>] ? do_page_fault+0x13d/0x5fb
[   37.474338]  [<b0189bfc>] ? putname+0x25/0x30
[   37.474338]  [<b01800e0>] ? do_sys_open+0xb1/0xc7
[   37.474338]  [<b018cf43>] sys_ioctl+0x52/0x63
[   37.474338]  [<b0104d52>] syscall_call+0x7/0xb
[   37.474338]  =======================
[   37.474338] Sched Debug Version: v0.07, 2.6.25-rc2-smp #14
[   37.474338] now at 44900.194289 msecs
[   37.474338]   .sysctl_sched_latency                    : 20.000000
[   37.474338]   .sysctl_sched_min_granularity            : 4.000000
[   37.474338]   .sysctl_sched_wakeup_granularity         : 10.000000
[   37.474338]   .sysctl_sched_batch_wakeup_granularity   : 10.000000
[   37.474338]   .sysctl_sched_child_runs_first           : 0.000001
[   37.474338]   .sysctl_sched_features                   : 39
[   37.474338] 
[   37.474338] cpu#0, 2992.603 MHz
[   37.474338]   .nr_running                    : 4
[   37.474338]   .load                          : 2048
[   37.474338]   .nr_switches                   : 3363
[   37.474338]   .nr_load_updates               : 34557
[   37.474338]   .nr_uninterruptible            : 2
[   37.474338]   .jiffies                       : 4294705811
[   37.474338]   .next_balance                  : 4294.705968
[   37.474338]   .curr->pid                     : 4
[   37.474338]   .clock                         : 37474.338672
[   37.474338]   .idle_clock                    : 3065.714769
[   37.474338]   .prev_clock_raw                : 78772.978492
[   37.474338]   .clock_warps                   : 0
[   37.474338]   .clock_overflows               : 3996
[   37.474338]   .clock_underflows              : 31781
[   37.474338]   .clock_deep_idle_events        : 1
[   37.474338]   .clock_max_delta               : 0.999848
[   37.474338]   .cpu_load[0]                   : 2048
[   37.474338]   .cpu_load[1]                   : 2048
[   37.474338]   .cpu_load[2]                   : 2048
[   37.474338]   .cpu_load[3]                   : 2048
[   37.474338]   .cpu_load[4]                   : 2048
[   37.474338] 
[   37.474338] cfs_rq
[   37.474338]   .exec_clock                    : 34293.783916
[   37.474338]   .MIN_vruntime                  : 0.000001
[   37.474338]   .min_vruntime                  : 17146.893105
[   37.474338]   .max_vruntime                  : 0.000001
[   37.474338]   .spread                        : 0.000000
[   37.474338]   .spread0                       : 0.000000
[   37.474338]   .nr_running                    : 1
[   37.474338]   .load                          : 2048
[   37.474338]   .bkl_count                     : 405
[   37.474338]   .nr_spread_over                : 0
[   37.474338] 
[   37.474338] cfs_rq
[   37.474338]   .exec_clock                    : 34293.783916
[   37.474338]   .MIN_vruntime                  : 13830.833996
[   37.474338]   .min_vruntime                  : 17146.893105
[   37.474338]   .max_vruntime                  : 13830.833996
[   37.474338]   .spread                        : 0.000000
[   37.474338]   .spread0                       : 0.000000
[   37.474338]   .nr_running                    : 4
[   37.474338]   .load                          : 8290
[   37.474338]   .bkl_count                     : 405
[   37.474338]   .nr_spread_over                : 6
[   37.474338] 
[   37.474338] runnable tasks:
[   37.474338]             task   PID         tree-key  switches  prio     exec-runtime         sum-exec        sum-sleep
[   37.474338] ----------------------------------------------------------------------------------------------------------
[   37.474338] R    ksoftirqd/0     4     14329.115724        37   115     14329.115724     33248.417353      4113.418770
[   37.474338]         events/0     5     13830.833996        35   115     13830.833996         0.284125      4265.748410
[   37.474338]            blogd   424     13830.833996       135   120     13830.833996         0.460905      1680.558420
[   37.474338]       udevsettle   465     13830.833996        17   120     13830.833996         0.838894       643.594622
[   37.474338] 

Quoted 7 lines · Mike Galbraith

Greetings,

K3b recently (9a4c854..5d9c4a7 pull) began terminally griping about buffer underrun upon every attempt to burn a CD. I can't fully bisect the problem because intervening kernels hang soft during boot. Using git bisect visualize, and converting to postable text:

Quoted 4 lines

bisect/bad block: add request->raw_data_len (6b00769fe1502b4ad97bb327ef7ac971b208bfb5)
bisect block: update bio according to DMA alignment padding (40b01b9bbdf51ae543a04744283bf2d56c4a6afa)
libata: update ATAPI overflow draining bisect/good-e164094964e6e20fe7fce418e06a9dce952bb7a4

Tejun?

Meanwhile back at the ranch, reverting
6b00769fe1502b4ad97bb327ef7ac971b208bfb5 40b01b9bbdf51ae543a04744283bf2d56c4a6afa and the one entangled line from dde2020754aeb14e17052d61784dcb37f252aac2 did restore my burner.

Greetings,

I straced both a good and a bad kernel (good being .git with attached revert patch applied) and filtered/diffed/merged the output. Scroll down to "HERE" to see the problem (resid).

I'm poking around, but not having much luck.

Quoted 7 lines · Mike Galbraith

Greetings,

I straced both a good and a bad kernel (good being .git with attached revert patch applied) and filtered/diffed/merged the output. Scroll down to "HERE" to see the problem (resid).

I'm poking around, but not having much luck.

Seems the problem is data_len changes, but raw_data_len doesn't. I've not the foggiest IO-land clue, but k3b works again, so the below may have some diagnostic value.

drivers/scsi/scsi_lib.c +0 -0

index ba21d97..7a6f784 100644

Quoted 7 lines · On Tue, 26 Feb 2008 14:36:43 +0100 Mike Galbraith <[email protected]> wrote:

Quoted 6 lines · Mike Galbraith

Greetings,

I straced both a good and a bad kernel (good being .git with attached revert patch applied) and filtered/diffed/merged the output. Scroll down to "HERE" to see the problem (resid).

Quoted 1 line

Quoted 1 line

I'm poking around, but not having much luck.

cc's added.

I'm told this is part of "Tejun's DMA drain handling".

Quoted 3 lines

Seems the problem is data_len changes, but raw_data_len doesn't. I've not the foggiest IO-land clue, but k3b works again, so the below may have some diagnostic value.

So this change fixes a bug? Can we have a recap of how it does this?

drivers/scsi/scsi_lib.c void scsi_io_completion(struct scsi_cmnd *cmd, unsigned int good_bytes) · line 875

Quoted 11 lines

drivers/scsi/scsi_lib.c +1 -1

index ba21d97..7a6f784 100644
@@ -871,7 +871,7 @@ void scsi_io_completion(struct scsi_cmnd *cmd, unsigned int good_bytes)
 			scsi_end_bidi_request(cmd);
 			return;
 		}
-		req->data_len = scsi_get_resid(cmd);
+		req->data_len = req->raw_data_len = scsi_get_resid(cmd);
 	}

I would love to get an answer as to what data_len (and of course raw_data_len) should be set to AFTER the command completes, which is what is going on here.

Quoted 3 lines · On Tue, 26 Feb 2008 14:36:43 +0100 Mike Galbraith <[email protected]> wrote:

Seems the problem is data_len changes, but raw_data_len doesn't. I've not the foggiest IO-land clue, but k3b works again, so the below may have some diagnostic value.

So this change fixes a bug? Can we have a recap of how it does this?

Quoted 1 line · Andrew Morton

So this change fixes a bug? Can we have a recap of how it does this?

Yeah, it fixes the problem. (wrt recap, if I could write it, it would be a changelog;)

(patchlet takes wing, and... goes splat?)

Hello, all.

Sorry about the delay. Was buried under other stuff. Mike, thanks a lot for reporting and analyzing the problem; however, the patch is slightly incorrect. rq->data_len is rq->data_len + extra stuff for alignment and padding, so the correct thing to do is...

  req->raw_data_len -= req->data_len - scsi_get_resid(cmd);
  req->data_len = scsi_get_resid(cmd);

which is ugly and error-prone. In addition, this isn't the only place where resid is set. Other block drivers do this too. This definitely should be done in block layer.

With rq->data_len and rq->raw_data_len, it's impossible to translate resid of rq->data_len to resid of rq->raw_data_len as block layer doesn't know how much was extra data after rq->data_len is modified. The attached patch substitutes rq->raw_data_len w/ rq->extra_len and adds blk_rq_raw_data_len(). Things look cleaner this way and the resid problem should be solved with this.

Can you please verify the attached patch fixes the problem?

Thanks.

plain text document attachment (patch)

block/blk-core.c +1 -2

index 775c851..929ab61 100644
@@ -127,7 +127,7 @@ void rq_init(struct request_queue *q, struct request *rq)
 	rq->nr_hw_segments = 0;
 	rq->ioprio = 0;
 	rq->special = NULL;
-	rq->raw_data_len = 0;
+	rq->extra_len = 0;
 	rq->buffer = NULL;
 	rq->tag = -1;
 	rq->errors = 0;

@@ -2016,7 +2016,6 @@ void blk_rq_bio_prep(struct request_queue *q, struct request *rq,
 	rq->hard_cur_sectors = rq->current_nr_sectors;
 	rq->hard_nr_sectors = rq->nr_sectors = bio_sectors(bio);
 	rq->buffer = bio_data(bio);
-	rq->raw_data_len = bio->bi_size;
 	rq->data_len = bio->bi_size;
 
 	rq->bio = rq->biotail = bio;

block/blk-map.c +1 -1

index 09f7fd0..c67a75f 100644
@@ -19,7 +19,6 @@ int blk_rq_append_bio(struct request_queue *q, struct request *rq,
 		rq->biotail->bi_next = bio;
 		rq->biotail = bio;
 
-		rq->raw_data_len += bio->bi_size;
 		rq->data_len += bio->bi_size;
 	}
 	return 0;

@@ -156,6 +155,7 @@ int blk_rq_map_user(struct request_queue *q, struct request *rq,
 		bio->bi_io_vec[bio->bi_vcnt - 1].bv_len += pad_len;
 		bio->bi_size += pad_len;
 		rq->data_len += pad_len;
+		rq->extra_len += pad_len;
 	}
 
 	rq->buffer = rq->data = NULL;

block/blk-merge.c +1 -0

index 7506c4f..efb5b4d 100644
@@ -232,6 +232,7 @@ new_segment:
 			    (PAGE_SIZE - 1));
 		nsegs++;
 		rq->data_len += q->dma_drain_size;
+		rq->extra_len += q->dma_drain_size;
 	}
 
 	if (sg)

block/bsg.c +4 -4

index 7f3c095..81b2133 100644
@@ -437,14 +437,14 @@ static int blk_complete_sgv4_hdr_rq(struct request *rq, struct sg_io_v4 *hdr,
 	}
 
 	if (rq->next_rq) {
-		hdr->dout_resid = rq->raw_data_len;
-		hdr->din_resid = rq->next_rq->raw_data_len;
+		hdr->dout_resid = blk_rq_raw_data_len(rq);
+		hdr->din_resid = blk_rq_raw_data_len(rq->next_rq);
 		blk_rq_unmap_user(bidi_bio);
 		blk_put_request(rq->next_rq);
 	} else if (rq_data_dir(rq) == READ)
-		hdr->din_resid = rq->raw_data_len;
+		hdr->din_resid = blk_rq_raw_data_len(rq);
 	else
-		hdr->dout_resid = rq->raw_data_len;
+		hdr->dout_resid = blk_rq_raw_data_len(rq);
 
 	/*
 	 * If the request generated a negative error number, return it

block/scsi_ioctl.c +2 -2

index e993cac..32424b3 100644
@@ -266,7 +266,7 @@ static int blk_complete_sghdr_rq(struct request *rq, struct sg_io_hdr *hdr,
 	hdr->info = 0;
 	if (hdr->masked_status || hdr->host_status || hdr->driver_status)
 		hdr->info |= SG_INFO_CHECK;
-	hdr->resid = rq->raw_data_len;
+	hdr->resid = blk_rq_raw_data_len(rq);
 	hdr->sb_len_wr = 0;
 
 	if (rq->sense_len && hdr->sbp) {

@@ -528,8 +528,8 @@ static int __blk_send_generic(struct request_queue *q, struct gendisk *bd_disk,
 	rq = blk_get_request(q, WRITE, __GFP_WAIT);
 	rq->cmd_type = REQ_TYPE_BLOCK_PC;
 	rq->data = NULL;
-	rq->raw_data_len = 0;
 	rq->data_len = 0;
+	rq->extra_len = 0;
 	rq->timeout = BLK_DEFAULT_SG_TIMEOUT;
 	memset(rq->cmd, 0, sizeof(rq->cmd));
 	rq->cmd[0] = cmd;

drivers/ata/libata-scsi.c +2 -1

index 0562b0a..5cab84c 100644
@@ -2539,7 +2539,8 @@ static unsigned int atapi_xlat(struct ata_queued_cmd *qc)
 	 * want to set it properly, and for DMA where it is
 	 * effectively meaningless.
 	 */
-	nbytes = min(scmd->request->raw_data_len, (unsigned int)63 * 1024);
+	nbytes = min(blk_rq_raw_data_len(scmd->request),
+		     (unsigned int)63 * 1024);
 
 	/* Most ATAPI devices which honor transfer chunk size don't
 	 * behave according to the spec when odd chunk size which

drivers/block/cciss.c +0 -0

drivers/scsi/scsi_lib.c +0 -0

include/linux/blkdev.h +6 -1

index 6fe67d1..57e2a9e 100644
@@ -216,8 +216,8 @@ struct request {
 	unsigned int cmd_len;
 	unsigned char cmd[BLK_MAX_CDB];
 
-	unsigned int raw_data_len;
 	unsigned int data_len;
+	unsigned int extra_len;
 	unsigned int sense_len;
 	void *data;
 	void *sense;

@@ -477,6 +477,11 @@ enum {
 
 #define rq_data_dir(rq)		((rq)->cmd_flags & 1)
 
+static inline unsigned int blk_rq_raw_data_len(struct request *rq)
+{
+	return rq->data_len - min(rq->extra_len, rq->data_len);
+}
+
 /*
  * We regard a request as sync, if it's a READ or a SYNC write.
  */

rq->raw_data_len introduced for block layer padding and draining (commit 6b00769fe1502b4ad97bb327ef7ac971b208bfb5) broke residual byte count handling. Block drivers modify rq->data_len to notify residual byte count to the block layer which blindly reported unmodified rq->raw_data_len to userland.

To keep block drivers dealing only with rq->data_len, this should be handled inside block layer. However, how much extra buffer was appened is lost after rq->data_len is modified.

This patch replaces rq->raw_data_len with rq->extra_len and add blk_rq_raw_data_len() helper to calculate raw data size from rq->data_len and rq->extra_len. The helper returns correct raw residual byte count when called on a rq whose data_len is modified to carry residual byte count.

This problem was reported and diagnosed by Mike Galbraith.

Quoted 1 line

This problem was reported and diagnosed by Mike Galbraith.

Tejun, this patch isn't much cleaner at all. It really shows the pain of these two seperate, yet related, variables.

Quoted 4 lines · Jens Axboe

Quoted 1 line

This problem was reported and diagnosed by Mike Galbraith.

Tejun, this patch isn't much cleaner at all. It really shows the pain of these two seperate, yet related, variables.

Not much cleaner compared to what? I think padding stuff is bound to be somewhat complex. It's a nasty thing in nature. I think ->extra_len is better than ->raw_data_len because ->extra_len only needs to be updated where the dirty jobs are done and extra buffer areas are added. Any better suggestions?

Quoted 11 lines · Tejun Heo

Quoted 4 lines · Jens Axboe

Quoted 1 line

This problem was reported and diagnosed by Mike Galbraith.

Tejun, this patch isn't much cleaner at all. It really shows the pain of these two seperate, yet related, variables.

Not much cleaner compared to what? I think padding stuff is bound to be somewhat complex. It's a nasty thing in nature. I think ->extra_len is better than ->raw_data_len because ->extra_len only needs to be updated where the dirty jobs are done and extra buffer areas are added. Any better suggestions?

Well, I just investigated a bug report in the SCSI transport class. Our SMP handler is broken in exactly the same way. We rely on the incoming reported request lengths to size our request data, and they've blown up from the true length to 512 bytes (the size of our alignment).

With the original patch, I have to run through the whole of libsas and scsi_transport_sas doing

s/data_len/raw_data_len/

With your update it looks like I have to run through them all doing

s/data_len/data_len - extra_len/

which is even worse. Can't we put things back to a point where data_len means exactly that and extra_len means how much we have spare on the end, so you know you can DMA up to data_len + extra_len if need be?

That way we don't have to sweep through every block driver altering the way it uses data_len.

Quoted 8 lines

With the original patch, I have to run through the whole of libsas and scsi_transport_sas doing

s/data_len/raw_data_len/

With your update it looks like I have to run through them all doing

s/data_len/data_len - extra_len/

Quoted 6 lines

which is even worse. Can't we put things back to a point where data_len means exactly that and extra_len means how much we have spare on the end, so you know you can DMA up to data_len + extra_len if need be?

That way we don't have to sweep through every block driver altering the way it uses data_len.

Fully agree. The reason why I think it's so ugly is that you have to keep these two seperate variables in sync. The burning was just one bug, there will be others...

Hello, Jens, James.

Quoted 3 lines · Jens Axboe

Quoted 3 lines

With the original patch, I have to run through the whole of libsas and scsi_transport_sas doing

Quoted 5 lines

Quoted 5 lines

s/data_len/raw_data_len/

With your update it looks like I have to run through them all doing

s/data_len/data_len - extra_len/

blk_rq_raw_data_len() should do.

Quoted 6 lines

Quoted 6 lines

which is even worse. Can't we put things back to a point where data_len means exactly that and extra_len means how much we have spare on the end, so you know you can DMA up to data_len + extra_len if need be?

That way we don't have to sweep through every block driver altering the way it uses data_len.

If SMP is broken because it needs start address alignment but not padding to align the size, what should be done is to make that exact requirement visible to the block layer. Say, blk_queue_dma_start_alignment() or maybe change blk_queue_dma_alignment() such that it only indicates start address alignment and add blk_queue_dma_size_alignment() for drivers which require size to be aligned too. I think those are few.

I think the decision which value rq->data_len represents comes down to which size is used more in low level drivers because no matter which way we choose we'll have to update some of the drivers which expects the other thing from rq->data_len.

blk_rq_raw_data_len() is needed iff a driver needs dummy buffers attached at the end and still needs to know the original request size which isn't the common case.

Quoted 3 lines

Fully agree. The reason why I think it's so ugly is that you have to keep these two seperate variables in sync. The burning was just one bug, there will be others...

The posted modification isn't too bad as the maintenance of the two variables is at places where the nasty things happen. I think what rq->data_len should represent when seen from LLDs is more important and please note that if SMP is broken because it simply doesn't require 512byte size alignment, it's a different issue.

As long as both raw_data_len and data_len are accessible, I'm okay either way. My biggest reluctance is against breaking sum(sg) == rq->data_len. I think this can lead to much more subtle problems such as programming the controller w/ wrong bytes count and wrapped-around resid calculation.

Hello, Jens, James.

Quoted 8 lines · Jens Axboe

Quoted 8 lines

With the original patch, I have to run through the whole of libsas and scsi_transport_sas doing

s/data_len/raw_data_len/

With your update it looks like I have to run through them all doing

s/data_len/data_len - extra_len/

blk_rq_raw_data_len() should do.

Quoted 6 lines

Quoted 6 lines

which is even worse. Can't we put things back to a point where data_len means exactly that and extra_len means how much we have spare on the end, so you know you can DMA up to data_len + extra_len if need be?

That way we don't have to sweep through every block driver altering the way it uses data_len.

If SMP is broken because it needs start address alignment but not padding to align the size, what should be done is to make that exact requirement visible to the block layer. Say, blk_queue_dma_start_alignment() or maybe change blk_queue_dma_alignment() such that it only indicates start address alignment and add blk_queue_dma_size_alignment() for drivers which require size to be aligned too. I think those are few.

I think the decision which value rq->data_len represents comes down to which size is used more in low level drivers because no matter which way we choose we'll have to update some of the drivers which expects the other thing from rq->data_len.

blk_rq_raw_data_len() is needed iff a driver needs dummy buffers attached at the end and still needs to know the original request size which isn't the common case.

Quoted 3 lines

Fully agree. The reason why I think it's so ugly is that you have to keep these two seperate variables in sync. The burning was just one bug, there will be others...

The posted modification isn't too bad as the maintenance of the two variables is at places where the nasty things happen. I think what rq->data_len should represent when seen from LLDs is more important and please note that if SMP is broken because it simply doesn't require 512byte size alignment, it's a different issue.

As long as both raw_data_len and data_len are accessible, I'm okay either way. My biggest reluctance is against breaking sum(sg) == rq->data_len. I think this can lead to much more subtle problems such as programming the controller w/ wrong bytes count and wrapped-around resid calculation.

sum(sg) == rq->data_len is already broken; sg sends such requests (though it would be nice if it doesn't).

sum(sg) == rq->data_len is already broken; sg sends such requests (though it would be nice if it doesn't).

I've not followed the earlier discussion (because I thought the drain buffer stuff affected only libata but seems it doesn't ...). Why did we need to change the meaning of rq->data_len?

From the POV of the block layer, as now it might extend the sg list, it

rq->data_len meant the true data length and the patch to change it doesn't look to make anything simple. Can we revert the meaning of rq->data_len? I'm not sure that we need to add rq->extra_len but it's fine as long as it's only for drivers that want to use it.

This is only compile tested.

Quoted 6 lines · FUJITA Tomonori

sum(sg) == rq->data_len is already broken; sg sends such requests (though it would be nice if it doesn't).

I've not followed the earlier discussion (because I thought the drain buffer stuff affected only libata but seems it doesn't ...). Why did we need to change the meaning of rq->data_len?

At this point, it's not clear what the original meaning of rq->data_len is because before moving alignment and padding to block layer, rq->data_len equaled both the requested data length and the length of sg list. AFAIK, it's SCSI midlayer which makes sg list and data length mismatch not block layer.

From the POV of the block layer, as now it might extend the sg list, it has to decide what rq->data_len means in this case - the requested transfer length from userland or the length of mapped sg list.

I think that currently the biggest problem is that drivers which don't require request size adjustment are getting it because alignment setting doesn't distinguish between start address alignment and size alignment. For drivers which don't require data size adjustment from block layer, data_len or raw_data_len doesn't matter. They're equal anyway. I'm prepping a patch for this now.

For drivers which do require request size adjustments, I think it's better to keep rq->data_len in line with the size of mapped sg list. The rationales are...

- Those are dumb controllers which want to see requests which meet certain size requirements and they're likely to care more about actual data buffer size than user requested buffer size. IOW, they wanna see sizes which meet certain requirements, so give them those values.

- I think bugs caused by using raw_data_len instead of data_len are more subtle than the other way around. Using data_len instead of raw_data_len usually affects the application layer while using raw_data_len instead of data_len affects the DMA engine and transport layer.

Quoted 6 lines

rq->data_len meant the true data length and the patch to change it doesn't look to make anything simple. Can we revert the meaning of rq->data_len? I'm not sure that we need to add rq->extra_len but it's fine as long as it's only for drivers that want to use it.

This is only compile tested.

If we're gonna go this way, we'll need blk_rq_total_data_len() and use it for drivers which requires request size adjustments.

Quoted 4 lines

- I think bugs caused by using raw_data_len instead of data_len are more subtle than the other way around. Using data_len instead of raw_data_len usually affects the application layer while using raw_data_len instead of data_len affects the DMA engine and transport layer.

If we add extra_len, we can get what raw_data_len and data_len provide.

I can't see what changing the meaning of rq->data_len (and investigating all the block drivers) gives us.

Block layer alignment was used for two different purposes - memory alignment and padding. This causes problems in lower layers because drivers which only require memory alignment ends up with adjusted rq->data_len. Separate out padding such that padding occurs iff driver explicitly requests it.

Quoted 5 lines · FUJITA Tomonori

Quoted 4 lines

- I think bugs caused by using raw_data_len instead of data_len are more subtle than the other way around. Using data_len instead of raw_data_len usually affects the application layer while using raw_data_len instead of data_len affects the DMA engine and transport layer.

Quoted 5 lines

If we add extra_len, we can get what raw_data_len and data_len provide.

I can't see what changing the meaning of rq->data_len (and investigating all the block drivers) gives us.

No matter which way you go, you change the meaning of rq->data_len and you MUST inspect rq->data_len usage whichever way you go.

Apply your patch and try to do sg IO on IDE cdrom w/ various transfer lengths.

Quoted 4 lines

Quoted 2 lines

I can't see what changing the meaning of rq->data_len (and investigating all the block drivers) gives us.

No matter which way you go, you change the meaning of rq->data_len and you MUST inspect rq->data_len usage whichever way you go.

The patch doens't change that rq->data_len means the true data length. But yeah, it breaks rq->data_len == sum(sg). So it might break some drivers.

Quoted 2 lines

Apply your patch and try to do sg IO on IDE cdrom w/ various transfer lengths.

I've just tried the patch with both ata and libata and it seems to work.

For anyone hitting this problem, please try the following patch:

http://lkml.org/lkml/2008/3/2/218

Quoted 5 lines · FUJITA Tomonori

Quoted 4 lines

Quoted 2 lines

I can't see what changing the meaning of rq->data_len (and investigating all the block drivers) gives us.

No matter which way you go, you change the meaning of rq->data_len and you MUST inspect rq->data_len usage whichever way you go.

Quoted 3 lines

The patch doens't change that rq->data_len means the true data length. But yeah, it breaks rq->data_len == sum(sg). So it might break some drivers.

Yeah, that's what I was saying. You end up breaking one of the two assumptions. As sglist is getting modified for any driver if it has DMA alignment set, whether rq->data_len is adjusted together or not, sglist and data_len usages have to be audited.

Quoted 5 lines · On Mon, 03 Mar 2008 18:21:13 +0900 Tejun Heo <[email protected]> wrote:

Quoted 4 lines · FUJITA Tomonori

Quoted 4 lines

Quoted 2 lines

I can't see what changing the meaning of rq->data_len (and investigating all the block drivers) gives us.

No matter which way you go, you change the meaning of rq->data_len and you MUST inspect rq->data_len usage whichever way you go.

Quoted 7 lines

Quoted 3 lines

The patch doens't change that rq->data_len means the true data length. But yeah, it breaks rq->data_len == sum(sg). So it might break some drivers.

Yeah, that's what I was saying. You end up breaking one of the two assumptions. As sglist is getting modified for any driver if it has DMA alignment set, whether rq->data_len is adjusted together or not, sglist and data_len usages have to be audited.

My patch (well, James' original approach) doesn't affect drivers that don't use drain buffer. rq->data_len still means the true data length and rq->data_len is equal to sum(sg) for them. So right now we need to audit only libata.

But your patch changes the meaning of rq->data_len. It affects all the drivers. So it breaks non libata stuff, like the SMP handler. We need to audit all the drivers.

Quoted 16 lines · FUJITA Tomonori

Quoted 12 lines · On Mon, 03 Mar 2008 18:21:13 +0900 Tejun Heo <[email protected]> wrote:

Quoted 7 lines · FUJITA Tomonori

Quoted 4 lines

Quoted 2 lines

I can't see what changing the meaning of rq->data_len (and investigating all the block drivers) gives us.

No matter which way you go, you change the meaning of rq->data_len and you MUST inspect rq->data_len usage whichever way you go.

The patch doens't change that rq->data_len means the true data length. But yeah, it breaks rq->data_len == sum(sg). So it might break some drivers.

Yeah, that's what I was saying. You end up breaking one of the two assumptions. As sglist is getting modified for any driver if it has DMA alignment set, whether rq->data_len is adjusted together or not, sglist and data_len usages have to be audited.

Quoted 4 lines

My patch (well, James' original approach) doesn't affect drivers that don't use drain buffer. rq->data_len still means the true data length and rq->data_len is equal to sum(sg) for them. So right now we need to audit only libata.

Your patch does change sglist for any driver which sets DMA alignment.

Quoted 12 lines

Quoted 12 lines

Quoted 12 lines

Quoted 7 lines · FUJITA Tomonori

Quoted 4 lines

Quoted 2 lines

I can't see what changing the meaning of rq->data_len (and investigating all the block drivers) gives us.

No matter which way you go, you change the meaning of rq->data_len and you MUST inspect rq->data_len usage whichever way you go.

The patch doens't change that rq->data_len means the true data length. But yeah, it breaks rq->data_len == sum(sg). So it might break some drivers.

Yeah, that's what I was saying. You end up breaking one of the two assumptions. As sglist is getting modified for any driver if it has DMA alignment set, whether rq->data_len is adjusted together or not, sglist and data_len usages have to be audited.

Quoted 5 lines

Quoted 4 lines

My patch (well, James' original approach) doesn't affect drivers that don't use drain buffer. rq->data_len still means the true data length and rq->data_len is equal to sum(sg) for them. So right now we need to audit only libata.

Your patch does change sglist for any driver which sets DMA alignment.

I overlook it. Where does it changes sglist?

Quoted 16 lines · FUJITA Tomonori

Quoted 16 lines

Quoted 16 lines

Quoted 5 lines

Quoted 4 lines · FUJITA Tomonori

Quoted 4 lines

Quoted 2 lines

I can't see what changing the meaning of rq->data_len (and investigating all the block drivers) gives us.

No matter which way you go, you change the meaning of rq->data_len and you MUST inspect rq->data_len usage whichever way you go.

Quoted 7 lines

Quoted 3 lines

The patch doens't change that rq->data_len means the true data length. But yeah, it breaks rq->data_len == sum(sg). So it might break some drivers.

Yeah, that's what I was saying. You end up breaking one of the two assumptions. As sglist is getting modified for any driver if it has DMA alignment set, whether rq->data_len is adjusted together or not, sglist and data_len usages have to be audited.

My patch (well, James' original approach) doesn't affect drivers that don't use drain buffer. rq->data_len still means the true data length and rq->data_len is equal to sum(sg) for them. So right now we need to audit only libata.

Quoted 3 lines

Quoted 1 line

Your patch does change sglist for any driver which sets DMA alignment.

I overlook it. Where does it changes sglist?

At the end of blk_rq_map_user() together with data_len / extra_len mangling or were you talking about James' original patch?

Quoted 2 lines

At the end of blk_rq_map_user() together with data_len / extra_len mangling or were you talking about James' original patch?

With my patch, at the end of blk_rq_map_user, we have:

	if (len & queue_dma_alignment(q)) {
		unsigned int pad_len = (queue_dma_alignment(q) & ~len) + 1;

		rq->extra_len += pad_len;
	}

So no change as compared with 2.6.24?

Quoted 8 lines · FUJITA Tomonori

Quoted 2 lines

At the end of blk_rq_map_user() together with data_len / extra_len mangling or were you talking about James' original patch?

With my patch, at the end of blk_rq_map_user, we have:

	if (len & queue_dma_alignment(q)) {
		unsigned int pad_len = (queue_dma_alignment(q) & ~len) + 1;

Quoted 5 lines

		rq->extra_len += pad_len;
	}

So no change as compared with 2.6.24?

Oh.. you killed sg list manipulation. Many controllers do allow odd bytes as the last sg entry but not all. Also, if you append drain

buffer after it, it ends up with unaligned sg entry in the middle and rq->data_len + rq->extra_len will overrun the sg entry after the drain page which is really dangerous.

Quoted 12 lines · On Mon, 03 Mar 2008 23:22:46 +0900 Tejun Heo <[email protected]> wrote:

Quoted 11 lines · FUJITA Tomonori

Quoted 2 lines

At the end of blk_rq_map_user() together with data_len / extra_len mangling or were you talking about James' original patch?

With my patch, at the end of blk_rq_map_user, we have:

	if (len & queue_dma_alignment(q)) {
		unsigned int pad_len = (queue_dma_alignment(q) & ~len) + 1;

		rq->extra_len += pad_len;
	}

Quoted 3 lines

Quoted 1 line

So no change as compared with 2.6.24?

Oh.. you killed sg list manipulation. Many controllers do allow odd bytes as the last sg entry but not all. Also, if you append drain

Until 2.6.24, these drivers have taken care about the issue by themselves. There is no change as compared with 2.6.24.

Quoted 3 lines

buffer after it, it ends up with unaligned sg entry in the middle and rq->data_len + rq->extra_len will overrun the sg entry after the drain page which is really dangerous.

The drivers know that they use drain buffer. They can take care about themselves on this too. If we want to do explicitly, we could have rq->pad_len and rq->drain_len instead of rq->extra_len, though I think that we are fine without these values because these drivers already tell the block layer what they want and know that the block layer gives it.

Quoted 19 lines · FUJITA Tomonori

Quoted 15 lines · On Mon, 03 Mar 2008 23:22:46 +0900 Tejun Heo <[email protected]> wrote:

Quoted 4 lines · FUJITA Tomonori

Quoted 2 lines

At the end of blk_rq_map_user() together with data_len / extra_len mangling or were you talking about James' original patch?

With my patch, at the end of blk_rq_map_user, we have:

Quoted 8 lines

	if (len & queue_dma_alignment(q)) {
		unsigned int pad_len = (queue_dma_alignment(q) & ~len) + 1;

		rq->extra_len += pad_len;
	}

So no change as compared with 2.6.24?

Oh.. you killed sg list manipulation. Many controllers do allow odd bytes as the last sg entry but not all. Also, if you append drain

Quoted 2 lines

Until 2.6.24, these drivers have taken care about the issue by themselves. There is no change as compared with 2.6.24.

Yeah, libata did its own padding and needed to add draining. Private implementation was complex as hell and James suggested moving them to block layer. Are you suggesting moving them back to drivers?

Quoted 10 lines

Quoted 3 lines

buffer after it, it ends up with unaligned sg entry in the middle and rq->data_len + rq->extra_len will overrun the sg entry after the drain page which is really dangerous.

The drivers know that they use drain buffer. They can take care about themselves on this too. If we want to do explicitly, we could have rq->pad_len and rq->drain_len instead of rq->extra_len, though I think that we are fine without these values because these drivers already tell the block layer what they want and know that the block layer gives it.

So, if a driver has requested aligning and draining, the driver should extend the sg entry before the last one by the alignment if draining was used for the request and extent the last sg if the draining wasn't used.
I'd rather just implement them in the drivers.

Quoted 3 lines

Yeah, libata did its own padding and needed to add draining. Private implementation was complex as hell and James suggested moving them to block layer. Are you suggesting moving them back to drivers?

No, I'm not. I've been working on the IOMMUs to remove such workarounds in LLDs.

What drivers need to do on this is just adding a padding length, that is, drivers don't need to change the structure of the sg list (like splitting a sg entry), right? And it doesn't break the SAS drivers that support SATAPI, does it?

But I agree that drivers want to get a complete sglist so I'm fine with adjusting sglist entries in the block layer with your secode patch (separate out padding from alignment). As we discussed, I'm fine with breaking sum(sg) == rq->data_len as long as rq->data_len means the true data length.

Quoted 9 lines

Quoted 9 lines

Quoted 3 lines

buffer after it, it ends up with unaligned sg entry in the middle and rq->data_len + rq->extra_len will overrun the sg entry after the drain page which is really dangerous.

The drivers know that they use drain buffer. They can take care about themselves on this too. If we want to do explicitly, we could have rq->pad_len and rq->drain_len instead of rq->extra_len, though I think that we are fine without these values because these drivers already tell the block layer what they want and know that the block layer gives it.

Quoted 4 lines

So, if a driver has requested aligning and draining, the driver should extend the sg entry before the last one by the alignment if draining was used for the request and extent the last sg if the draining wasn't used.
I'd rather just implement them in the drivers.

The block layer extends the sg entry? The drivers just adjust sg->length?

Quoted 12 lines · FUJITA Tomonori

Quoted 3 lines

Yeah, libata did its own padding and needed to add draining. Private implementation was complex as hell and James suggested moving them to block layer. Are you suggesting moving them back to drivers?

No, I'm not. I've been working on the IOMMUs to remove such workarounds in LLDs.

What drivers need to do on this is just adding a padding length, that is, drivers don't need to change the structure of the sg list (like splitting a sg entry), right? And it doesn't break the SAS drivers that support SATAPI, does it?

Quoted 5 lines

But I agree that drivers want to get a complete sglist so I'm fine with adjusting sglist entries in the block layer with your secode patch (separate out padding from alignment). As we discussed, I'm fine with breaking sum(sg) == rq->data_len as long as rq->data_len means the true data length.

As long as the second patch is in, what value rq->data_len indicates doesn't matter to drivers which don't use explicit padding or draining, so the situation is much more controlled. I don't care which value rq->data_len would indicate. I'd prefer it equal sum(sg) as that value is what IDE and libata which will be the major users of padding and/or draining expect in rq->data_len but fixing up that shouldn't be too difficult. I guess this can be determined by Jens. If Jens likes rq->data_len to contain requested transfer size, I'll post updated patches.

Quoted 14 lines

Quoted 13 lines

Quoted 9 lines

Quoted 3 lines

buffer after it, it ends up with unaligned sg entry in the middle and rq->data_len + rq->extra_len will overrun the sg entry after the drain page which is really dangerous.

The drivers know that they use drain buffer. They can take care about themselves on this too. If we want to do explicitly, we could have rq->pad_len and rq->drain_len instead of rq->extra_len, though I think that we are fine without these values because these drivers already tell the block layer what they want and know that the block layer gives it.

So, if a driver has requested aligning and draining, the driver should extend the sg entry before the last one by the alignment if draining was used for the request and extent the last sg if the draining wasn't used.
I'd rather just implement them in the drivers.

Quoted 2 lines

The block layer extends the sg entry? The drivers just adjust sg->length?

Still, do you really wanna force such things into low level drivers? That will be one extremely fragile API and will be really difficult to tell when things go wrong.

Quoted 19 lines · On Tue, 04 Mar 2008 11:32:56 +0900 Tejun Heo <[email protected]> wrote:

Quoted 12 lines · FUJITA Tomonori

Quoted 3 lines

Yeah, libata did its own padding and needed to add draining. Private implementation was complex as hell and James suggested moving them to block layer. Are you suggesting moving them back to drivers?

No, I'm not. I've been working on the IOMMUs to remove such workarounds in LLDs.

What drivers need to do on this is just adding a padding length, that is, drivers don't need to change the structure of the sg list (like splitting a sg entry), right? And it doesn't break the SAS drivers that support SATAPI, does it?

Quoted 5 lines

But I agree that drivers want to get a complete sglist so I'm fine with adjusting sglist entries in the block layer with your secode patch (separate out padding from alignment). As we discussed, I'm fine with breaking sum(sg) == rq->data_len as long as rq->data_len means the true data length.

Quoted 8 lines

As long as the second patch is in, what value rq->data_len indicates doesn't matter to drivers which don't use explicit padding or draining, so the situation is much more controlled. I don't care which value rq->data_len would indicate. I'd prefer it equal sum(sg) as that value is what IDE and libata which will be the major users of padding and/or draining expect in rq->data_len but fixing up that shouldn't be too difficult. I guess this can be determined by Jens. If Jens likes rq->data_len to contain requested transfer size, I'll post updated patches.

OK, I prefer rq->data_len means the true data length though you prefer rq->data_len means the allocated buffer length (the true data length plus padding and drain). We agree on other things. We can live with either way.

Jens, what's your preference?

Quoted 36 lines · FUJITA Tomonori

Quoted 19 lines · On Tue, 04 Mar 2008 11:32:56 +0900 Tejun Heo <[email protected]> wrote:

Quoted 12 lines · FUJITA Tomonori

Quoted 3 lines

Yeah, libata did its own padding and needed to add draining. Private implementation was complex as hell and James suggested moving them to block layer. Are you suggesting moving them back to drivers?

No, I'm not. I've been working on the IOMMUs to remove such workarounds in LLDs.

What drivers need to do on this is just adding a padding length, that is, drivers don't need to change the structure of the sg list (like splitting a sg entry), right? And it doesn't break the SAS drivers that support SATAPI, does it?

Quoted 5 lines

But I agree that drivers want to get a complete sglist so I'm fine with adjusting sglist entries in the block layer with your secode patch (separate out padding from alignment). As we discussed, I'm fine with breaking sum(sg) == rq->data_len as long as rq->data_len means the true data length.

Quoted 8 lines

As long as the second patch is in, what value rq->data_len indicates doesn't matter to drivers which don't use explicit padding or draining, so the situation is much more controlled. I don't care which value rq->data_len would indicate. I'd prefer it equal sum(sg) as that value is what IDE and libata which will be the major users of padding and/or draining expect in rq->data_len but fixing up that shouldn't be too difficult. I guess this can be determined by Jens. If Jens likes rq->data_len to contain requested transfer size, I'll post updated patches.

OK, I prefer rq->data_len means the true data length though you prefer rq->data_len means the allocated buffer length (the true data length plus padding and drain). We agree on other things. We can live with either way.

Quoted 1 line

Jens, what's your preference?

I completely agree with you, ->data_len meaning true data length is way cleaner imho. Only the driver should care for the padded length, all other parts of the kernel only need to know what they actually got.

Quoted 37 lines · On Tue, 4 Mar 2008 09:59:46 +0100 Jens Axboe <[email protected]> wrote:

Quoted 36 lines · FUJITA Tomonori

Quoted 19 lines · On Tue, 04 Mar 2008 11:32:56 +0900 Tejun Heo <[email protected]> wrote:

Quoted 12 lines · FUJITA Tomonori

Quoted 3 lines

Yeah, libata did its own padding and needed to add draining. Private implementation was complex as hell and James suggested moving them to block layer. Are you suggesting moving them back to drivers?

No, I'm not. I've been working on the IOMMUs to remove such workarounds in LLDs.

What drivers need to do on this is just adding a padding length, that is, drivers don't need to change the structure of the sg list (like splitting a sg entry), right? And it doesn't break the SAS drivers that support SATAPI, does it?

Quoted 5 lines

But I agree that drivers want to get a complete sglist so I'm fine with adjusting sglist entries in the block layer with your secode patch (separate out padding from alignment). As we discussed, I'm fine with breaking sum(sg) == rq->data_len as long as rq->data_len means the true data length.

Quoted 8 lines

As long as the second patch is in, what value rq->data_len indicates doesn't matter to drivers which don't use explicit padding or draining, so the situation is much more controlled. I don't care which value rq->data_len would indicate. I'd prefer it equal sum(sg) as that value is what IDE and libata which will be the major users of padding and/or draining expect in rq->data_len but fixing up that shouldn't be too difficult. I guess this can be determined by Jens. If Jens likes rq->data_len to contain requested transfer size, I'll post updated patches.

OK, I prefer rq->data_len means the true data length though you prefer rq->data_len means the allocated buffer length (the true data length plus padding and drain). We agree on other things. We can live with either way.

Quoted 5 lines

Quoted 1 line

Jens, what's your preference?

I completely agree with you, ->data_len meaning true data length is way cleaner imho. Only the driver should care for the padded length, all other parts of the kernel only need to know what they actually got.

OK, now we can fix the whole SG_IO (and bsg handler) mess.

Here's my patch with a proper description. which several people have already tested (thanks!). Then we need an updated version of Tejun's separate out padding from alignment patch.

OK, I've updated his patch. Tejun, can you audit this?

Quoted 46 lines · On Tue, 04 Mar 2008 18:06:48 +0900 FUJITA Tomonori…

Quoted 42 lines · On Tue, 4 Mar 2008 09:59:46 +0100 Jens Axboe <[email protected]> wrote:

Quoted 31 lines · FUJITA Tomonori

Quoted 19 lines · On Tue, 04 Mar 2008 11:32:56 +0900 Tejun Heo <[email protected]> wrote:

Quoted 12 lines · FUJITA Tomonori

Quoted 3 lines

Yeah, libata did its own padding and needed to add draining. Private implementation was complex as hell and James suggested moving them to block layer. Are you suggesting moving them back to drivers?

No, I'm not. I've been working on the IOMMUs to remove such workarounds in LLDs.

What drivers need to do on this is just adding a padding length, that is, drivers don't need to change the structure of the sg list (like splitting a sg entry), right? And it doesn't break the SAS drivers that support SATAPI, does it?

Quoted 5 lines

But I agree that drivers want to get a complete sglist so I'm fine with adjusting sglist entries in the block layer with your secode patch (separate out padding from alignment). As we discussed, I'm fine with breaking sum(sg) == rq->data_len as long as rq->data_len means the true data length.

Quoted 8 lines

As long as the second patch is in, what value rq->data_len indicates doesn't matter to drivers which don't use explicit padding or draining, so the situation is much more controlled. I don't care which value rq->data_len would indicate. I'd prefer it equal sum(sg) as that value is what IDE and libata which will be the major users of padding and/or draining expect in rq->data_len but fixing up that shouldn't be too difficult. I guess this can be determined by Jens. If Jens likes rq->data_len to contain requested transfer size, I'll post updated patches.

Quoted 6 lines

OK, I prefer rq->data_len means the true data length though you prefer rq->data_len means the allocated buffer length (the true data length plus padding and drain). We agree on other things. We can live with either way.

Jens, what's your preference?

I completely agree with you, ->data_len meaning true data length is way cleaner imho. Only the driver should care for the padded length, all other parts of the kernel only need to know what they actually got.

Quoted 5 lines

OK, now we can fix the whole SG_IO (and bsg handler) mess.

Here's my patch with a proper description. which several people have already tested (thanks!). Then we need an updated version of Tejun's separate out padding from alignment patch.

OK, I've updated his patch. Tejun, can you audit this?

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

Yeah, the other version which added extra_len to data_len has been verified to work. The only difference is now libata is adding extra_len, so this one should be safe.

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

Quoted 8 lines · Mike Galbraith

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

can you please try git as of this morning without any patches applied, and then pull

git://git.kernel.dk/linux-2.6-block.git for-linus

into that and see if that works?

Quoted 6 lines

can you please try git as of this morning without any patches applied, and then pull

git://git.kernel.dk/linux-2.6-block.git for-linus

into that and see if that works?

I'll give it a shot in a bit.

Quoted 7 lines · Mike Galbraith

Quoted 6 lines

can you please try git as of this morning without any patches applied, and then pull

git://git.kernel.dk/linux-2.6-block.git for-linus

into that and see if that works?

Quoted 1 line

I'll give it a shot in a bit.

Aw poo, so many choices.
I did:
git add remote block-for-linus git://git.kernel.dk/linux-2.6-block.git
git remote update
Now, which one do I check out? block-for-linus/master maybe, or block-for-linus/for-linus?

Re-read my original mail! It states that you should just pull:

git://git.kernel.dk/linux-2.6-block.git for-linus

into your linus branch, or just create a test branch off linus' master and pull into that. IOW, it's the for-linus branch that you should pull, nothing else.

Quoted 4 lines · Jens Axboe

Re-read my original mail! It states that you should just pull:

git://git.kernel.dk/linux-2.6-block.git for-linus

Quoted 3 lines

into your linus branch, or just create a test branch off linus' master and pull into that. IOW, it's the for-linus branch that you should pull, nothing else.

Well, I had a good reason. You know how to un-pull, I know how to un-remote to get back to pristine after I'm done testing... guaranteed without whimpering pathetically on the git list ;-)

Anyway, I checked out the one with the big-fat-hint in it's name (block-for-linus/for-linus).
Same error. Git this morning with patches...
restore_meaning_of_data_len.diff seperate_out_padding_from_alignment.diff
...reverted restored me to the originally reported k3b error, nothing new noted.

If I tested the wrong branch, whack me upside the head, and I'll follow your pull destructions, and figure out how to un-pull later.

Quoted 7 lines · Mike Galbraith

Quoted 4 lines · Jens Axboe

Re-read my original mail! It states that you should just pull:

git://git.kernel.dk/linux-2.6-block.git for-linus

Quoted 7 lines

Quoted 3 lines

into your linus branch, or just create a test branch off linus' master and pull into that. IOW, it's the for-linus branch that you should pull, nothing else.

Well, I had a good reason. You know how to un-pull, I know how to un-remote to get back to pristine after I'm done testing... guaranteed without whimpering pathetically on the git list ;-)

OK, if you're on master, it's pretty easy:

$ git branch test-branch
$ git checkout test-branch
$ git pull git://git.kernel.dk/linux-2.6-block.git for-linus

[build, boot, test]

$ git checkout master
$ git branch -D test-branch

Quoted 8 lines

Anyway, I checked out the one with the big-fat-hint in it's name (block-for-linus/for-linus).
Same error. Git this morning with patches...
restore_meaning_of_data_len.diff seperate_out_padding_from_alignment.diff
...reverted restored me to the originally reported k3b error, nothing new noted.

Quoted 2 lines

If I tested the wrong branch, whack me upside the head, and I'll follow your pull destructions, and figure out how to un-pull later.

for-linus is the right branch, but I'm just a little worried that you didn't test what you think you tested. What does cat .git/HEAD say? If that is a ref to a file (eg refs/heads/master), what does that file contain?

Quoted 7 lines · Mike Galbraith

Quoted 4 lines · Jens Axboe

Re-read my original mail! It states that you should just pull:

git://git.kernel.dk/linux-2.6-block.git for-linus

Quoted 7 lines

Quoted 3 lines

into your linus branch, or just create a test branch off linus' master and pull into that. IOW, it's the for-linus branch that you should pull, nothing else.

Well, I had a good reason. You know how to un-pull, I know how to un-remote to get back to pristine after I'm done testing... guaranteed without whimpering pathetically on the git list ;-)

OK, if you're on master, it's pretty easy:

$ git branch test-branch
$ git checkout test-branch
$ git pull git://git.kernel.dk/linux-2.6-block.git for-linus

[build, boot, test]

$ git checkout master
$ git branch -D test-branch

Quoted 8 lines

Anyway, I checked out the one with the big-fat-hint in it's name (block-for-linus/for-linus).
Same error. Git this morning with patches...
restore_meaning_of_data_len.diff seperate_out_padding_from_alignment.diff
...reverted restored me to the originally reported k3b error, nothing new noted.

Quoted 2 lines

If I tested the wrong branch, whack me upside the head, and I'll follow your pull destructions, and figure out how to un-pull later.

for-linus is the right branch, but I'm just a little worried that you didn't test what you think you tested. What does cat .git/HEAD say? If that is a ref to a file (eg refs/heads/master), what does that file contain?

Quoted 18 lines · Jens Axboe

Quoted 7 lines · Mike Galbraith

Quoted 4 lines · Jens Axboe

Re-read my original mail! It states that you should just pull:

git://git.kernel.dk/linux-2.6-block.git for-linus

Quoted 7 lines

Quoted 3 lines

into your linus branch, or just create a test branch off linus' master and pull into that. IOW, it's the for-linus branch that you should pull, nothing else.

Well, I had a good reason. You know how to un-pull, I know how to un-remote to get back to pristine after I'm done testing... guaranteed without whimpering pathetically on the git list ;-)

OK, if you're on master, it's pretty easy:

Quoted 7 lines

$ git branch test-branch
$ git checkout test-branch
$ git pull git://git.kernel.dk/linux-2.6-block.git for-linus

[build, boot, test]

$ git checkout master
$ git branch -D test-branch

Hm, that's simple enough. I'll do this for the edification. Thanks. Maybe some day, I'll cease to be so paranoid that my test setup may become compromised. (at which time...)

Quoted 11 lines

Quoted 10 lines

Anyway, I checked out the one with the big-fat-hint in it's name (block-for-linus/for-linus).
Same error. Git this morning with patches...
restore_meaning_of_data_len.diff seperate_out_padding_from_alignment.diff
...reverted restored me to the originally reported k3b error, nothing new noted.

If I tested the wrong branch, whack me upside the head, and I'll follow your pull destructions, and figure out how to un-pull later.

Quoted 4 lines

for-linus is the right branch, but I'm just a little worried that you didn't test what you think you tested. What does cat .git/HEAD say? If that is a ref to a file (eg refs/heads/master), what does that file contain?

That wouldn't surprise me one bit. (ergo...)

It says cc66b4512cae8df4ed1635483210aabf7690ec27... kewpie doll?

Quoted 1 line

It says cc66b4512cae8df4ed1635483210aabf7690ec27... kewpie doll?

That looks right, then perhaps there's still an issue there :/ Logs?

Quoted 8 lines · Tejun Heo

Quoted 3 lines · FUJITA Tomonori

Quoted 2 lines

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Tejun, I thought that libata needs a fix for sum(sg) != rq->data_len. No?

The extra_len you added to qc->nbytes should be it. The only other place to pay attention is the ATAPI transfer chunk size and your patch seems to get it right.

Quoted 4 lines

Quoted 2 lines

Now Jens' git tree should work with all the non libata stuff, ide, firewire, bsg, etc. But I'm not sure about libata.

With the second patch, all others should be fine no matter what. I'll go check libata part again.

I can reproduce the problem here and it's very weird. I'll report back when I know more.

Quoted 4 lines · Jens Axboe

Quoted 1 line

It says cc66b4512cae8df4ed1635483210aabf7690ec27... kewpie doll?

That looks right, then perhaps there's still an issue there :/ Logs?

Tejuns patchlet (below) fixed it here.

Quoted 8 lines · Mike Galbraith

Quoted 4 lines · Jens Axboe

Quoted 1 line

It says cc66b4512cae8df4ed1635483210aabf7690ec27... kewpie doll?

That looks right, then perhaps there's still an issue there :/ Logs?

Quoted 1 line

Tejuns patchlet (below) fixed it here.

OK, can you try changing that to

        good_bytes = scsi_bufflen(cmd) + cmd->request->extra_len;

and retest?

Quoted 11 lines · Jens Axboe

Quoted 9 lines · Mike Galbraith

Quoted 4 lines · Jens Axboe

Quoted 1 line

It says cc66b4512cae8df4ed1635483210aabf7690ec27... kewpie doll?

That looks right, then perhaps there's still an issue there :/ Logs?

Tejuns patchlet (below) fixed it here.

Quoted 5 lines

OK, can you try changing that to

        good_bytes = scsi_bufflen(cmd) + cmd->request->extra_len;

and retest?

Yup, disk #42 is happily burning away.

Quoted 8 lines · Mike Galbraith

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

can you please try git as of this morning without any patches applied, and then pull

git://git.kernel.dk/linux-2.6-block.git for-linus

into that and see if that works?

Quoted 13 lines · Jens Axboe

Quoted 8 lines · Mike Galbraith

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

can you please try git as of this morning without any patches applied, and then pull

Quoted 3 lines

git://git.kernel.dk/linux-2.6-block.git for-linus

into that and see if that works?

Works for me with the SAS SMP handler. Both input request and output response frame sizes are picked up and returned with the correct residues.

Quoted 6 lines · Mike Galbraith

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

can you please try git as of this morning without any patches applied, and then pull

git://git.kernel.dk/linux-2.6-block.git for-linus

into that and see if that works?

Quoted 8 lines · Jens Axboe

Quoted 6 lines · Mike Galbraith

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

Quoted 6 lines

can you please try git as of this morning without any patches applied, and then pull

git://git.kernel.dk/linux-2.6-block.git for-linus

into that and see if that works?

Unfortunately this doesn't fix a problem I've discussed off-list with Kiyoshi Ueda, who suggested that I should follow this thread and try any patches posted here.

Here is what happens when I try to mount a CD (before and after I pull 'for-linus'):

hdc: ide_cd_check_ireason: wrong transfer direction!
cdrom: failed setting lba address space
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status timeout: status=0xd0 { Busy }
ide: failed opcode was: unknown
hdc: DMA disabled
hdc: drive not ready for command
hdc: ATAPI reset complete
ISO 9660 Extensions: Microsoft Joliet Level 3
ISOFS: changing to secondary root
VFS: busy inodes on changed media.

The mount can take from 5 seconds on up to a minute or so before the CD can be accessed.

Quoted 56 lines · walt

Quoted 5 lines · Jens Axboe

Quoted 4 lines · Mike Galbraith

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

Quoted 8 lines

Quoted 2 lines

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

can you please try git as of this morning without any patches applied, and then pull

git://git.kernel.dk/linux-2.6-block.git for-linus

into that and see if that works?

Unfortunately this doesn't fix a problem I've discussed off-list with Kiyoshi Ueda, who suggested that I should follow this thread and try any patches posted here.

Here is what happens when I try to mount a CD (before and after I pull 'for-linus'):

hdc: ide_cd_check_ireason: wrong transfer direction!
cdrom: failed setting lba address space
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status error: status=0x58 { DriveReady SeekComplete DataRequest }
ide: failed opcode was: unknown
hdc: drive not ready for command
hdc: status timeout: status=0xd0 { Busy }
ide: failed opcode was: unknown
hdc: DMA disabled
hdc: drive not ready for command
hdc: ATAPI reset complete
ISO 9660 Extensions: Microsoft Joliet Level 3
ISOFS: changing to secondary root
VFS: busy inodes on changed media.

Quoted 2 lines

The mount can take from 5 seconds on up to a minute or so before the CD can be accessed.

Which version did you try? There was a recent IDE bug fix which affected CD recording. Commit bcd88ac3b2ff2eae3d0fa57a6b02d4fce5392f32 which is included in 2.6.25-rc3.

Also, did 2.6.24 work?

Quoted 11 lines · Jens Axboe

Quoted 6 lines · Mike Galbraith

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

can you please try git as of this morning without any patches applied, and then pull

Quoted 3 lines

git://git.kernel.dk/linux-2.6-block.git for-linus

into that and see if that works?

Unfortunately this doesn't fix a problem I've discussed off-list with Kiyoshi Ueda, who suggested that I should follow this thread and try any patches posted here.

cdrecord: Warning: controller returns zero sized CD capabilities page.
cdrecord: Warning: controller returns wrong page 0 for CD capabilities page (2A).

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

[  136.440021] ata1.01: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 frozen
[  136.440043] ata1.01: cmd a0/00:00:00:00:00/00:00:00:00:00/b0 tag 0
[  136.440045]          cdb 51 00 00 00 00 00 00 00  02 00 00 00 00 00 00 00
[  136.440047]          res 58/00:02:00:02:00/00:00:00:00:00/b0 Emask 0x2 (HSM violation)
[  136.440053] ata1.01: status: { DRDY DRQ }
[  136.440086] ata1: soft resetting link
[  165.327627] ata1.01: qc timeout (cmd 0xa1)
[  165.327627] ata1.01: failed to IDENTIFY (I/O error, err_mask=0x4)
[  165.327627] ata1.01: revalidation failed (errno=-5)
[  165.327627] ata1: failed to recover some devices, retrying in 5 secs
[  177.272373] ata1: port is slow to respond, please be patient (Status 0x80)
[  180.388879] ata1: device not ready (errno=-16), forcing hardreset
[  180.388879] ata1: soft resetting link
[  210.832471] ata1.01: qc timeout (cmd 0xa1)
[  210.832471] ata1.01: failed to IDENTIFY (I/O error, err_mask=0x4)
[  210.832471] ata1.01: revalidation failed (errno=-5)

[  210.832471] ata1: failed to recover some devices, retrying in 5 secs
[  223.392899] ata1: port is slow to respond, please be patient (Status 0x80)
[  225.920376] ata1: device not ready (errno=-16), forcing hardreset
[  225.920376] ata1: soft resetting link
[  256.542565] ata1.01: qc timeout (cmd 0xa1)
[  256.542565] ata1.01: failed to IDENTIFY (I/O error, err_mask=0x4)
[  256.542565] ata1.01: revalidation failed (errno=-5)
[  256.542565] ata1.01: disabled

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Quoted 24 lines · Mike Galbraith

Quoted 2 lines · Jens Axboe

Looks excellent to me, has a variant of this been tested as OK by the users reporting the regression?

K3b burning seems to be a nogo here. This is git pulled this morning though, so it's a somewhat different tree than previously tested fwtw.

[  136.440021] ata1.01: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 frozen
[  136.440043] ata1.01: cmd a0/00:00:00:00:00/00:00:00:00:00/b0 tag 0
[  136.440045]          cdb 51 00 00 00 00 00 00 00  02 00 00 00 00 00 00 00
[  136.440047]          res 58/00:02:00:02:00/00:00:00:00:00/b0 Emask 0x2 (HSM violation)
[  136.440053] ata1.01: status: { DRDY DRQ }
[  136.440086] ata1: soft resetting link
[  165.327627] ata1.01: qc timeout (cmd 0xa1)
[  165.327627] ata1.01: failed to IDENTIFY (I/O error, err_mask=0x4)
[  165.327627] ata1.01: revalidation failed (errno=-5)
[  165.327627] ata1: failed to recover some devices, retrying in 5 secs
[  177.272373] ata1: port is slow to respond, please be patient (Status 0x80)
[  180.388879] ata1: device not ready (errno=-16), forcing hardreset
[  180.388879] ata1: soft resetting link
[  210.832471] ata1.01: qc timeout (cmd 0xa1)
[  210.832471] ata1.01: failed to IDENTIFY (I/O error, err_mask=0x4)
[  210.832471] ata1.01: revalidation failed (errno=-5)

Quoted 8 lines

[  210.832471] ata1: failed to recover some devices, retrying in 5 secs
[  223.392899] ata1: port is slow to respond, please be patient (Status 0x80)
[  225.920376] ata1: device not ready (errno=-16), forcing hardreset
[  225.920376] ata1: soft resetting link
[  256.542565] ata1.01: qc timeout (cmd 0xa1)
[  256.542565] ata1.01: failed to IDENTIFY (I/O error, err_mask=0x4)
[  256.542565] ata1.01: revalidation failed (errno=-5)
[  256.542565] ata1.01: disabled

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Quoted 2 lines

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Tejun, I thought that libata needs a fix for sum(sg) != rq->data_len. No?

Now Jens' git tree should work with all the non libata stuff, ide, firewire, bsg, etc. But I'm not sure about libata.

Quoted 3 lines · FUJITA Tomonori

Quoted 2 lines

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Tejun, I thought that libata needs a fix for sum(sg) != rq->data_len. No?

The extra_len you added to qc->nbytes should be it. The only other place to pay attention is the ATAPI transfer chunk size and your patch seems to get it right.

Quoted 2 lines

Now Jens' git tree should work with all the non libata stuff, ide, firewire, bsg, etc. But I'm not sure about libata.

With the second patch, all others should be fine no matter what. I'll go check libata part again.

Quoted 8 lines · Tejun Heo

Quoted 3 lines · FUJITA Tomonori

Quoted 2 lines

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Tejun, I thought that libata needs a fix for sum(sg) != rq->data_len. No?

The extra_len you added to qc->nbytes should be it. The only other place to pay attention is the ATAPI transfer chunk size and your patch seems to get it right.

Quoted 4 lines

Quoted 2 lines

Now Jens' git tree should work with all the non libata stuff, ide, firewire, bsg, etc. But I'm not sure about libata.

With the second patch, all others should be fine no matter what. I'll go check libata part again.

I can reproduce the problem here and it's very weird. I'll report back when I know more.

Quoted 9 lines · Tejun Heo

Quoted 8 lines · Tejun Heo

Quoted 3 lines · FUJITA Tomonori

Quoted 2 lines

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Tejun, I thought that libata needs a fix for sum(sg) != rq->data_len. No?

The extra_len you added to qc->nbytes should be it. The only other place to pay attention is the ATAPI transfer chunk size and your patch seems to get it right.

Quoted 6 lines

Quoted 4 lines

Quoted 2 lines

Now Jens' git tree should work with all the non libata stuff, ide, firewire, bsg, etc. But I'm not sure about libata.

With the second patch, all others should be fine no matter what. I'll go check libata part again.

I can reproduce the problem here and it's very weird. I'll report back when I know more.

Okay, I got it. Heh, it turns out SCSI and/or block layer is not ready for rq->data_len != sum(sg). When adjusted command completes, SCSI midlayer completes the command with rq->data_len for PC commands which eventually ends up in __end_that_request_first(). As there are extra sg area left after completing rq->data_len, blk layer says so to SCSI layer and SCSI layer retries the command only with the appended area.

The following patch gets the writing going. I really think it's a serious mistake to break rq->data_len == sum(sg). If we break rq->data_len == requested size, the worst bugs are giving wrong size when issuing commands to application layer of devices which is relatively easy to spot and not all that command anyway. Breaking rq->data_len == sum(sg), bugs will be in internal mechanics, DMA engine programming and transport layer. Oh well...

drivers/scsi/scsi.c +0 -0

index fecba05..32439ac 100644

drivers/scsi/scsi.c void scsi_finish_command(struct scsi_cmnd *cmd) · line 763

Quoted 5 lines

         if (cmd->request->cmd_type != REQ_TYPE_BLOCK_PC) {

drivers/scsi/scsi.c +0 -0

 		drv = scsi_cmd_to_driver(cmd);
 		if (drv->done)

I hate this patch. I wish you could maybe take the extra_len into account inside blk_end_request. The padding should be transparent to all concerned but the requesting LLD and the internals of the block layer. If block layer added padding it should take that into account on completion. My $0.2.

Quoted 9 lines · Tejun Heo

Quoted 8 lines · Tejun Heo

Quoted 3 lines · FUJITA Tomonori

Quoted 2 lines

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Tejun, I thought that libata needs a fix for sum(sg) != rq->data_len. No?

The extra_len you added to qc->nbytes should be it. The only other place to pay attention is the ATAPI transfer chunk size and your patch seems to get it right.

Quoted 7 lines

Quoted 4 lines

Quoted 2 lines

Now Jens' git tree should work with all the non libata stuff, ide, firewire, bsg, etc. But I'm not sure about libata.

With the second patch, all others should be fine no matter what. I'll go check libata part again.

I can reproduce the problem here and it's very weird. I'll report back when I know more.

Okay, I got it. Heh, it turns out SCSI and/or block layer is not ready for rq->data_len != sum(sg). When adjusted command completes, SCSI midlayer completes the command with rq->data_len for PC commands which eventually ends up in __end_that_request_first(). As there are extra sg area left after completing rq->data_len, blk layer says so to SCSI layer and SCSI layer retries the command only with the appended area.

The following patch gets the writing going. I really think it's a serious mistake to break rq->data_len == sum(sg). If we break rq->data_len == requested size, the worst bugs are giving wrong size when issuing commands to application layer of devices which is relatively easy to spot and not all that command anyway. Breaking rq->data_len == sum(sg), bugs will be in internal mechanics, DMA engine programming and transport layer. Oh well...

drivers/scsi/scsi.c +0 -0

index fecba05..32439ac 100644

Quoted 40 lines · Tejun Heo

Quoted 9 lines · Tejun Heo

Quoted 8 lines · Tejun Heo

Quoted 3 lines · FUJITA Tomonori

Quoted 2 lines

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Tejun, I thought that libata needs a fix for sum(sg) != rq->data_len. No?

The extra_len you added to qc->nbytes should be it. The only other place to pay attention is the ATAPI transfer chunk size and your patch seems to get it right.

Quoted 6 lines

Quoted 4 lines

Quoted 2 lines

Now Jens' git tree should work with all the non libata stuff, ide, firewire, bsg, etc. But I'm not sure about libata.

With the second patch, all others should be fine no matter what. I'll go check libata part again.

I can reproduce the problem here and it's very weird. I'll report back when I know more.

Okay, I got it. Heh, it turns out SCSI and/or block layer is not ready for rq->data_len != sum(sg). When adjusted command completes, SCSI midlayer completes the command with rq->data_len for PC commands which eventually ends up in __end_that_request_first(). As there are extra sg area left after completing rq->data_len, blk layer says so to SCSI layer and SCSI layer retries the command only with the appended area.

The following patch gets the writing going. I really think it's a serious mistake to break rq->data_len == sum(sg). If we break rq->data_len == requested size, the worst bugs are giving wrong size when issuing commands to application layer of devices which is relatively easy to spot and not all that command anyway. Breaking rq->data_len == sum(sg), bugs will be in internal mechanics, DMA engine programming and transport layer. Oh well...

drivers/scsi/scsi.c +0 -0

index fecba05..32439ac 100644
@@ -757,7 +757,7 @@ void scsi_finish_command(struct scsi_cmnd *cmd)
 				"Notifying upper driver of completion "
 				"(result %x)\n", cmd->result));
 

drivers/scsi/scsi.c void scsi_finish_command(struct scsi_cmnd *cmd) · line 760

Quoted 2 lines

drivers/scsi/scsi.c +1 -1

-	good_bytes = scsi_bufflen(cmd);
+	good_bytes = scsi_bufflen(cmd) + cmd->request->data_len;

This doesn't look right. scsi_bufflen(cmd) is req->data_len for PC commands ... did you mean to add extra_len here?

The following patch gets the writing going.

Quoted 1 line · Tejun Heo

The following patch gets the writing going.

Bingo.

Quoted 6 lines · Mike Galbraith

Quoted 1 line · Tejun Heo

The following patch gets the writing going.

Bingo.

Pretty please test this on top of current -git?

Quoted 4 lines · Jens Axboe

Quoted 3 lines · Mike Galbraith

On Wed, 2008-03-05 at 01:42 +0900, Tejun Heo wrote:

Quoted 5 lines

Quoted 3 lines

Quoted 1 line

The following patch gets the writing going.

Bingo.

Pretty please test this on top of current -git?

? That's the patch I just tested, and the tree. Oh.. 976dde0..87baa2b just a sec....

Quoted 9 lines · Tejun Heo

Quoted 8 lines · Tejun Heo

Quoted 3 lines · FUJITA Tomonori

Quoted 2 lines

Aiee... device going down after timing out on READ_DISC_INFO. That's gruesome. Can you please try the other patches?

Tejun, I thought that libata needs a fix for sum(sg) != rq->data_len. No?

The extra_len you added to qc->nbytes should be it. The only other place to pay attention is the ATAPI transfer chunk size and your patch seems to get it right.

Quoted 7 lines

Quoted 4 lines

Quoted 2 lines

Now Jens' git tree should work with all the non libata stuff, ide, firewire, bsg, etc. But I'm not sure about libata.

With the second patch, all others should be fine no matter what. I'll go check libata part again.

I can reproduce the problem here and it's very weird. I'll report back when I know more.

Okay, I got it. Heh, it turns out SCSI and/or block layer is not ready for rq->data_len != sum(sg). When adjusted command completes, SCSI midlayer completes the command with rq->data_len for PC commands which eventually ends up in __end_that_request_first(). As there are extra sg area left after completing rq->data_len, blk layer says so to SCSI layer and SCSI layer retries the command only with the appended area.

The following patch gets the writing going. I really think it's a serious mistake to break rq->data_len == sum(sg). If we break rq->data_len == requested size, the worst bugs are giving wrong size when issuing commands to application layer of devices which is relatively easy to spot and not all that command anyway. Breaking rq->data_len == sum(sg), bugs will be in internal mechanics, DMA engine programming and transport layer. Oh well...

drivers/scsi/scsi.c +0 -0

index fecba05..32439ac 100644

Hmm, does SCSI mid-layer need to care about how many bytes the block layer allocates? I don't think that extra_len is NOT good_bytes.

I think that the block layer had better take care about it (fix __end_that_request_first?).

Quoted 5 lines · FUJITA Tomonori

Hmm, does SCSI mid-layer need to care about how many bytes the block layer allocates? I don't think that extra_len is NOT good_bytes.

I think that the block layer had better take care about it (fix __end_that_request_first?).

Yeah, probably calling completion functions w/o bytes count is the right thing to do but what I was talking about was what could break when the semantics of rq->data_len changed. If we keep rq->data_len() == sum(sg), we keep it business as usual for all the rest except for the device application layer if we don't we do the reverse and SCSI midlayer completion was a good example, I think.

Things going the other way is fine with me but I at least want to hear a valid rationale. Till now all I got is "because that's the true size" which doesn't really make much sense to me.

Quoted 5 lines · FUJITA Tomonori

Hmm, does SCSI mid-layer need to care about how many bytes the block layer allocates? I don't think that extra_len is NOT good_bytes.

I think that the block layer had better take care about it (fix __end_that_request_first?).

Yeah, probably calling completion functions w/o bytes count is the right thing to do but what I was talking about was what could break when the semantics of rq->data_len changed. If we keep rq->data_len() == sum(sg), we keep it business as usual for all the rest except for the device application layer if we don't we do the reverse and SCSI midlayer completion was a good example, I think.

Things going the other way is fine with me but I at least want to hear a valid rationale. Till now all I got is "because that's the true size" which doesn't really make much sense to me.

sglist is a low-level I/O representation for device drivers. SCSI midlayer should not care about sglist. We should not fix SCSI midlayer for rq->data_len != sum(sg) change (so I can't agree with your diagrams in another mail).

When if we change a rule, we need to fix something.

If we keep rq->data_len == sum(sg), we need to fix the device application layer. If we keep rq->data_len == the true data length, we need to fix the low-level drivers.

Now I'm fine with the commit e97a294ef6938512b655b1abf17656cf2b26f709 since we are in -rc stages. But I plan to send a patch to revert it and fix this issue in the block layer. I'd like to test it in -mm for a while.

Only sglist stuff in SCSI midlayer is scsi_req_map_sg now. As you know, we really want to remove it.

Quoted 3 lines

Things going the other way is fine with me but I at least want to hear a valid rationale. Till now all I got is "because that's the true size" which doesn't really make much sense to me.

Most of users of request structure care about only the real data length, don't care about padding and drain length. Why do they bother to use a helper function to get the real data length?

Quoted 4 lines

Quoted 3 lines

Things going the other way is fine with me but I at least want to hear a valid rationale. Till now all I got is "because that's the true size" which doesn't really make much sense to me.

Quoted 3 lines

Most of users of request structure care about only the real data length, don't care about padding and drain length. Why do they bother to use a helper function to get the real data length?

I think this is where the difference comes from. To me it seems internal usage seems more wide-spread and more delicate and not too many care about the true size and when they do only in well defined places. Maybe it comes from the difference between your most and my most.

Quoted 10 lines

Quoted 6 lines

Quoted 3 lines

Things going the other way is fine with me but I at least want to hear a valid rationale. Till now all I got is "because that's the true size" which doesn't really make much sense to me.

Most of users of request structure care about only the real data length, don't care about padding and drain length. Why do they bother to use a helper function to get the real data length?

I think this is where the difference comes from. To me it seems internal usage seems more wide-spread and more delicate and not too many care about the true size and when they do only in well defined places. Maybe it comes from the difference between your most and my most.

I don't think that they only in well defined places.

If you see scsi mid-layer (and LLDs), you can find several places that use rq->data_len as the true data length.

Breaking rq->data_len == the true data length theoretically wrong. Even if it affects only libata now, it will hurt us, I think.

Quoted 4 lines · On Wed, 05 Mar 2008 08:33:05 +0900 Tejun Heo <[email protected]> wrote:

Quoted 3 lines · FUJITA Tomonori

Hmm, does SCSI mid-layer need to care about how many bytes the block layer allocates? I don't think that extra_len is NOT good_bytes.

Quoted 8 lines

Quoted 2 lines

I think that the block layer had better take care about it (fix __end_that_request_first?).

Yeah, probably calling completion functions w/o bytes count is the right thing to do but what I was talking about was what could break when the semantics of rq->data_len changed. If we keep rq->data_len() == sum(sg), we keep it business as usual for all the rest except for the device application layer if we don't we do the reverse and SCSI midlayer completion was a good example, I think.

sglist is a low-level I/O representation for device drivers. SCSI midlayer should not care about sglist. We should not fix SCSI midlayer for rq->data_len != sum(sg) change (so I can't agree with your diagrams in another mail).

When if we change a rule, we need to fix something.

If we keep rq->data_len == sum(sg), we need to fix the device application layer. If we keep rq->data_len == the true data length, we need to fix the low-level drivers.

Now I'm fine with the commit e97a294ef6938512b655b1abf17656cf2b26f709 since we are in -rc stages. But I plan to send a patch to revert it and fix this issue in the block layer. I'd like to test it in -mm for a while.

Only sglist stuff in SCSI midlayer is scsi_req_map_sg now. As you know, we really want to remove it.

Quoted 3 lines

Things going the other way is fine with me but I at least want to hear a valid rationale. Till now all I got is "because that's the true size" which doesn't really make much sense to me.

Most of users of request structure care about only the real data length, don't care about padding and drain length. Why do they bother to use a helper function to get the real data length?
--