ide-cd: Handle strange interrupt on the Intel ESB2
PATCH 1/1 ide-cd: Handle strange interrupt on the Intel ESB2
The ESB2 appears to emit spurious DMA interrupts when configured for native mode and handling ATAPI devices. Stratus were able to pin this bug down and produce a patch. This is a rework which applies the fixup only to the ESB2 (for now). We can apply it to other chips later if the same problem is found.
This code has been tested and confirmed to fix the problem on the tested systems.
(Most of the hard work done by Stratus however)
drivers/ide/ide-cd.c +7 -0
@@ -687,8 +687,15 @@static int cdrom_decode_status (ide_drive_t *drive,int good_stat,int *stat_ret) {struct request *rq =HWGROUP (drive)->rq; +ide_hwif_t *hwif =HWIF (drive);int stat, err, sense_key; +/* We may have bogus DMA interrupts in PIO state here */ +if (HWIF (drive)->dma_status && hwif->atapi_irq_bogon) { + stat = hwif->INB (hwif->dma_status); +/* Should we force the bit as well ? */ + hwif->OUTB (stat, hwif->dma_status); + }/* Check for errors. */ stat =HWIF (drive)->INB (IDE_STATUS_REG );if (stat_ret)
drivers/ide/pci/piix.c +4 -0
@@ -473,6 +473,10 @@/* This is a painful system best to let it self tune for now */ return ; } +/* ESB2 appears to generate spurious DMA interrupts in PIO mode +when in native mode */ +if (hwif->pci_dev->device ==PCI_DEVICE_ID_INTEL_ESB2_18 ) + hwif->atapi_irq_bogon =1 ; hwif->autodma =0 ; hwif->tuneproc = &piix_tune_drive;
include/linux/ide.h +1 -0
@@ -796,6 +796,7 @@ unsigned sg_mapped : 1; /* sg_table and sg_nents are ready */ unsigned no_io_32bit : 1; /* 1 = can not do 32-bit IO ops */ unsigned err_stops_fifo : 1; /* 1=data FIFO is cleared by an error */ + unsigned atapi_irq_bogon : 1; /* Generates spurious DMA interrupts in PIO mode */struct device gendev;struct completion gendev_rel_comp;/* To deal with device release() */
[ added Jeff and Tejun to cc: ]
Quoted 5 lines · Alan
The ESB2 appears to emit spurious DMA interrupts when configured for native mode and handling ATAPI devices. Stratus were able to pin this bug down and produce a patch. This is a rework which applies the fixup only to the ESB2 (for now). We can apply it to other chips later if the same problem is found.
Looks good but aren't this trying to fix the same ICH issue that is fixed in libata by ap->ops->irq_clear(ap)?
[ please see Tejun's mail: http://lkml.org/lkml/2006/11/15/94 ]
If so shouldn't we apply this fix for all ICH5/6/7/8 chipsets?
Quoted 2 lines
This code has been tested and confirmed to fix the problem on the tested systems.
Also shouldn't the fix be in IRQ handler? Currently the fix is limited to ide-cd driver which seems to be the wrong place as the problem is supposed to happen also when using other IDE device drivers or/and other ATA/ATAPI devices?
Quoted 47 lines
Signed-off-by Alan Cox <[email protected]>(Most of the hard work done by Stratus however)
diff -u --new-file --recursive --exclude-from
/usr/src/excludelinux.vanilla-2.6.19-rc6-mm1/drivers/ide/ide-cd.clinux-2.6.19-rc6-mm1/drivers/ide/ide-cd.cdrivers/ide/ide-cd.c 2006-12-01 19:24:58.000000000 +0000 renamed from drivers/ide/ide-cd.c 2006-11-24 13:58:06.000000000 +0000 +7
-0@@ -687,8 +687,15 @@ static int cdrom_decode_status(ide_drive_t *drive, int good_stat, int *stat_ret) { struct request *rq = HWGROUP(drive)->rq; + ide_hwif_t *hwif = HWIF(drive); int stat, err, sense_key; + /* We may have bogus DMA interrupts in PIO state here */ + if (HWIF(drive)->dma_status && hwif->atapi_irq_bogon) { + stat = hwif->INB(hwif->dma_status); + /* Should we force the bit as well ? */ + hwif->OUTB(stat, hwif->dma_status); + } /* Check for errors. */ stat = HWIF(drive)->INB(IDE_STATUS_REG); if (stat_ret)diff -u --new-file --recursive --exclude-from
/usr/src/excludelinux.vanilla-2.6.19-rc6-mm1/drivers/ide/pci/piix.clinux-2.6.19-rc6-mm1/drivers/ide/pci/piix.cdrivers/ide/pci/piix.c 2006-12-01 19:20:46.000000000 +0000 renamed from drivers/ide/pci/piix.c 2006-11-24 13:58:29.000000000 +0000 +4
-0@@ -473,6 +473,10 @@ /* This is a painful system best to let it self tune for now */ return; } + /* ESB2 appears to generate spurious DMA interrupts in PIO mode + when in native mode */ + if (hwif->pci_dev->device == PCI_DEVICE_ID_INTEL_ESB2_18) + hwif->atapi_irq_bogon = 1; hwif->autodma = 0; hwif->tuneproc = &piix_tune_drive;diff -u --new-file --recursive --exclude-from
/usr/src/excludelinux.vanilla-2.6.19-rc6-mm1/include/linux/ide.hlinux-2.6.19-rc6-mm1/include/linux/ide.hinclude/linux/ide.h 2006-12-01 19:16:27.000000000 +0000 renamed from include/linux/ide.h 2006-11-24 13:58:12.000000000 +0000 +1
-0@@ -796,6 +796,7 @@ unsigned sg_mapped : 1; /* sg_table and sg_nents are ready */ unsigned no_io_32bit : 1; /* 1 = can not do 32-bit IO ops */ unsigned err_stops_fifo : 1; /* 1=data FIFO is cleared by an error */ + unsigned atapi_irq_bogon : 1; /* Generates spurious DMA interrupts in PIO mode */ struct device gendev; struct completion gendev_rel_comp; /* To deal with device release() */
Quoted 6 lines · On Wed, 6 Dec 2006 00:23:03 +0100 "Bartlomiej Zolnierkiewicz"…
Looks good but aren't this trying to fix the same ICH issue that is fixed in libata by
ap->ops->irq_clear(ap)?[ please see Tejun's mail: http://lkml.org/lkml/2006/11/15/94 ]
If so shouldn't we apply this fix for all ICH5/6/7/8 chipsets?
Possibly. Thats one reason I made it a quirk bit. I'd certainly expect it to be a group of related chipsets.
Quoted 4 lines
Also shouldn't the fix be in IRQ handler? Currently the fix is limited to ide-cd driver which seems to be the wrong place as the problem is supposed to happen also when using other IDE device drivers or/and other ATA/ATAPI devices?
The problem has only be observed with CD devices doing PIO ATAPI commands. I am not aware of an Intel errata document that characterises this errata so anything else so cannot guess further. Perhaps Intel can advise ?
Alan
Quoted 18 lines · Alan
Quoted 6 lines · On Wed, 6 Dec 2006 00:23:03 +0100 "Bartlomiej Zolnierkiewicz"…
Looks good but aren't this trying to fix the same ICH issue that is fixed in libata by
ap->ops->irq_clear(ap)?[ please see Tejun's mail: http://lkml.org/lkml/2006/11/15/94 ]
If so shouldn't we apply this fix for all ICH5/6/7/8 chipsets?
Possibly. Thats one reason I made it a quirk bit. I'd certainly expect it to be a group of related chipsets.
Quoted 4 lines
Also shouldn't the fix be in IRQ handler? Currently the fix is limited to ide-cd driver which seems to be the wrong place as the problem is supposed to happen also when using other IDE device drivers or/and other ATA/ATAPI devices?
Quoted 4 lines
The problem has only be observed with CD devices doing PIO ATAPI commands. I am not aware of an Intel errata document that characterises this errata so anything else so cannot guess further. Perhaps Intel can advise ?
FWIW, on my ICH7, the IRQ storm does occur on ATA devices.
Signature · 1 line
tejun