Re: Out of Memory: Killed process 16498 (java).

[Jens Axboe <axboe@suse.de>]

On Sat, Jan 21 2006, Andy Chittenden wrote:
> ok - here is the complete dmesg as of today:
> 
> ksize
> usbmon: debugfs is not available
> mice: PS/2 mouse device common for all mice
> NET: Registered protocol family 2
> IP route cache hash table entries: 131072 (order: 8, 1048576 bytes)
> TCP established hash table entries: 524288 (order: 10, 4194304 bytes)
> TCP bind hash table entries: 65536 (order: 7, 524288 bytes)
> TCP: Hash tables configured (established 524288 bind 65536)
> TCP reno registered
> TCP bic registered
> NET: Registered protocol family 1
> NET: Registered protocol family 17
> NET: Registered protocol family 8
> NET: Registered protocol family 20
> ACPI wakeup devices: 
> PCI0 PS2K PS2M UAR2 UAR1 AC97 USB1 USB2 USB3 USB4 EHCI PWRB SLPB 
> ACPI: (supports S0 S1 S3 S4 S5)
> Freeing unused kernel memory: 148k freed
> Uniform Multi-Platform E-IDE driver Revision: 7.00alpha2
> ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
> input: AT Translated Set 2 keyboard as /class/input/input0
> VP_IDE: IDE controller at PCI slot 0000:00:0f.1
> GSI 17 sharing vector 0xB1 and IRQ 17
> ACPI: PCI Interrupt 0000:00:0f.1[A] -> GSI 20 (level, low) -> IRQ 17
> PCI: Via IRQ fixup for 0000:00:0f.1, from 255 to 1
> VP_IDE: chipset revision 6
> VP_IDE: not 100% native mode: will probe irqs later
> VP_IDE: VIA vt8237 (rev 00) IDE UDMA133 controller on pci0000:00:0f.1
>     ide0: BM-DMA at 0xfc00-0xfc07, BIOS settings: hda:DMA, hdb:DMA
>     ide1: BM-DMA at 0xfc08-0xfc0f, BIOS settings: hdc:DMA, hdd:pio
> Probing IDE interface ide0...
> hda: HDS722525VLAT80, ATA DISK drive
> hdb: Maxtor 6Y200P0, ATA DISK drive
> bounce: queue ffff81013fa312d8, setting pfn 1310720, max_low 1310720
> bounce: queue ffff81013fa312d8, setting pfn 1048575, max_low 1310720
> isa bounce pool size: 16 pages
> bounce: queue ffff81013fa31048, setting pfn 1310720, max_low 1310720
> bounce: queue ffff81013fa31048, setting pfn 1048575, max_low 1310720

There's your problem, apparently both of these queues is being set to a
limit lower than blk_max_low_pfn which means the block layer will revert
to the isa dma bounce zone for that queue... There's room for
improvement in the logic that chooses what zone to allocate from. It's a
little problematic since there are two reasons why you would have to
bounce: the hardware cannot support a "high" address, _or_ the driver
cannot support a highmem page. The latter can be fixed, the former is
given of course.

So choosing the gfp mask is a composite of whether the machine has
highmem and the hardware/driver capabilities.

-- 
Jens Axboe