Subject : [2.6.28-rc] Sata soft reset filling log Submitter : Justin Madru <bevicm@dslextreme.com> Date : 2008-12-13 2:07 References : http://marc.info/?l=linux-kernel&m=122913412608533&w=4 This entry is being used for tracking a regression from 2.6.27. Please don't close it until the problem is fixed in the mainline.
Notify-Also : Linux IDE <linux-ide@vger.kernel.org>
On Tuesday 13 January 2009, Justin Madru wrote: > Rafael J. Wysocki wrote: > > This message has been generated automatically as a part of a report > > of regressions introduced between 2.6.27 and 2.6.28. > > > > The following bug entry is on the current list of known regressions > > introduced between 2.6.27 and 2.6.28. Please verify if it still should > > be listed and let me know (either way). > > > > > > Bug-Entry : http://bugzilla.kernel.org/show_bug.cgi?id=12263 > > Subject : Sata soft reset filling log > > Submitter : Justin Madru <bevicm@dslextreme.com> > > Date : 2008-12-13 2:07 (30 days old) > > References : http://marc.info/?l=linux-kernel&m=122913412608533&w=4 > > > > > > > > > Yes, I'm still seeing this on .28 and even on .29-rc1. > I checked again and couldn't trigger the regression on a .27 kernel (at > least with my hardware and config). > So, I'm fairly confident that it's a regression, not just a bug that's > always been there.
kudos to Justin for submitting lots of useful info. Since Intel ICH7 SATA is the only storage contoller in the machine, it's either piix or ahci driver bug. (82801GBM/GHM). The only thing I see missing from this bug report is kernel dmesg (or /var/log/messages) output relating to the SATA controller initialization. Can that please be added?
On Sunday 15 February 2009, Justin Madru wrote: > Rafael J. Wysocki wrote: > > This message has been generated automatically as a part of a report > > of regressions introduced between 2.6.27 and 2.6.28. > > > > The following bug entry is on the current list of known regressions > > introduced between 2.6.27 and 2.6.28. Please verify if it still should > > be listed and let me know (either way). > > > > > > Bug-Entry : http://bugzilla.kernel.org/show_bug.cgi?id=12263 > > Subject : Sata soft reset filling log > > Submitter : Justin Madru <bevicm@dslextreme.com> > > Date : 2008-12-13 2:07 (64 days old) > > References : http://marc.info/?l=linux-kernel&m=122913412608533&w=4 > > I'm still seeing this on .29-rc5, and I think that my bug #12263 is a > duplicate of bug #12609, > or more correctly it's a duplicate of mine because I reported first. > > It seems like the bug has been fixed in tip/master for some time now. > Below is the diff of origin and tip from when I tested. > > $ git diff origin/master..tip/master drivers/ata/ > > diff --git a/drivers/ata/ata_piix.c b/drivers/ata/ata_piix.c > index 54961c0..e004c25 100644 > --- a/drivers/ata/ata_piix.c > +++ b/drivers/ata/ata_piix.c > @@ -310,7 +310,7 @@ static struct scsi_host_template piix_sht = { > }; > > static struct ata_port_operations piix_pata_ops = { > - .inherits = &ata_bmdma32_port_ops, > + .inherits = &ata_bmdma_port_ops, > .cable_detect = ata_cable_40wire, > .set_piomode = piix_set_piomode, > .set_dmamode = piix_set_dmamode, > diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c > index 9fbf059..1ed3966 100644 > --- a/drivers/ata/libata-core.c > +++ b/drivers/ata/libata-core.c > @@ -1482,7 +1482,7 @@ static int ata_hpa_resize(struct ata_device *dev) > struct ata_eh_context *ehc = &dev->link->eh_context; > int print_info = ehc->i.flags & ATA_EHI_PRINTINFO; > u64 sectors = ata_id_n_sectors(dev->id); > - u64 native_sectors; > + u64 uninitialized_var(native_sectors); > int rc; > > /* do we need to do it? */ > diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c > index b9747fa..d65b9b2 100644 > --- a/drivers/ata/libata-scsi.c > +++ b/drivers/ata/libata-scsi.c > @@ -3247,7 +3247,7 @@ void ata_scsi_scan_host(struct ata_port *ap, int sync) > int tries = 5; > struct ata_device *last_failed_dev = NULL; > struct ata_link *link; > - struct ata_device *dev; > + struct ata_device *uninitialized_var(dev); > > if (ap->flags & ATA_FLAG_DISABLED) > return; > diff --git a/drivers/ata/libata-sff.c b/drivers/ata/libata-sff.c > index 0b299b0..416e3e2 100644 > --- a/drivers/ata/libata-sff.c > +++ b/drivers/ata/libata-sff.c > @@ -80,13 +80,6 @@ const struct ata_port_operations ata_bmdma_port_ops = { > }; > EXPORT_SYMBOL_GPL(ata_bmdma_port_ops); > > -const struct ata_port_operations ata_bmdma32_port_ops = { > - .inherits = &ata_bmdma_port_ops, > - > - .sff_data_xfer = ata_sff_data_xfer32, > -}; > -EXPORT_SYMBOL_GPL(ata_bmdma32_port_ops); > - > /** > * ata_fill_sg - Fill PCI IDE PRD table > * @qc: Metadata associated with taskfile to be transferred > @@ -743,52 +736,6 @@ unsigned int ata_sff_data_xfer(struct ata_device > *dev, unsigned char *buf, > EXPORT_SYMBOL_GPL(ata_sff_data_xfer); > > /** > - * ata_sff_data_xfer32 - Transfer data by PIO > - * @dev: device to target > - * @buf: data buffer > - * @buflen: buffer length > - * @rw: read/write > - * > - * Transfer data from/to the device data register by PIO using 32bit > - * I/O operations. > - * > - * LOCKING: > - * Inherited from caller. > - * > - * RETURNS: > - * Bytes consumed. > - */ > - > -unsigned int ata_sff_data_xfer32(struct ata_device *dev, unsigned char > *buf, > - unsigned int buflen, int rw) > -{ > - struct ata_port *ap = dev->link->ap; > - void __iomem *data_addr = ap->ioaddr.data_addr; > - unsigned int words = buflen >> 2; > - int slop = buflen & 3; > - > - /* Transfer multiple of 4 bytes */ > - if (rw == READ) > - ioread32_rep(data_addr, buf, words); > - else > - iowrite32_rep(data_addr, buf, words); > - > - if (unlikely(slop)) { > - __le32 pad; > - if (rw == READ) { > - pad = cpu_to_le32(ioread32(ap->ioaddr.data_addr)); > - memcpy(buf + buflen - slop, &pad, slop); > - } else { > - memcpy(&pad, buf + buflen - slop, slop); > - iowrite32(le32_to_cpu(pad), ap->ioaddr.data_addr); > - } > - words++; > - } > - return words << 2; > -} > -EXPORT_SYMBOL_GPL(ata_sff_data_xfer32); > - > -/** > * ata_sff_data_xfer_noirq - Transfer data by PIO > * @dev: device to target > * @buf: data buffer > diff --git a/drivers/ata/pata_ali.c b/drivers/ata/pata_ali.c > index eb99dbe..7cd48ea 100644 > --- a/drivers/ata/pata_ali.c > +++ b/drivers/ata/pata_ali.c > @@ -151,7 +151,8 @@ static void ali_fifo_control(struct ata_port *ap, > struct ata_device *adev, int o > > pci_read_config_byte(pdev, pio_fifo, &fifo); > fifo &= ~(0x0F << shift); > - fifo |= (on << shift); > + if (on) > + fifo |= (on << shift); > pci_write_config_byte(pdev, pio_fifo, fifo); > } > > @@ -369,11 +370,10 @@ static struct ata_port_operations > ali_early_port_ops = { > .inherits = &ata_sff_port_ops, > .cable_detect = ata_cable_40wire, > .set_piomode = ali_set_piomode, > - .sff_data_xfer = ata_sff_data_xfer32, > }; > > static const struct ata_port_operations ali_dma_base_ops = { > - .inherits = &ata_bmdma32_port_ops, > + .inherits = &ata_bmdma_port_ops, > .set_piomode = ali_set_piomode, > .set_dmamode = ali_set_dmamode, > }; > diff --git a/drivers/ata/pata_amd.c b/drivers/ata/pata_amd.c > index 63719ab..0ec9c7d 100644 > --- a/drivers/ata/pata_amd.c > +++ b/drivers/ata/pata_amd.c > @@ -24,7 +24,7 @@ > #include <linux/libata.h> > > #define DRV_NAME "pata_amd" > -#define DRV_VERSION "0.3.11" > +#define DRV_VERSION "0.3.10" > > /** > * timing_setup - shared timing computation and load > @@ -345,7 +345,7 @@ static struct scsi_host_template amd_sht = { > }; > > static const struct ata_port_operations amd_base_port_ops = { > - .inherits = &ata_bmdma32_port_ops, > + .inherits = &ata_bmdma_port_ops, > .prereset = amd_pre_reset, > }; > > diff --git a/drivers/ata/pata_atiixp.c b/drivers/ata/pata_atiixp.c > index 506adde..115eb00 100644 > --- a/drivers/ata/pata_atiixp.c > +++ b/drivers/ata/pata_atiixp.c > @@ -140,7 +140,7 @@ static void atiixp_set_dmamode(struct ata_port *ap, > struct ata_device *adev) > wanted_pio = 3; > else if (adev->dma_mode == XFER_MW_DMA_0) > wanted_pio = 0; > - else BUG(); > + else panic("atiixp_set_dmamode: unknown DMA mode!"); > > if (adev->pio_mode != wanted_pio) > atiixp_set_pio_timing(ap, adev, wanted_pio); > diff --git a/drivers/ata/pata_mpiix.c b/drivers/ata/pata_mpiix.c > index aa576ca..7c8faa4 100644 > --- a/drivers/ata/pata_mpiix.c > +++ b/drivers/ata/pata_mpiix.c > @@ -35,7 +35,7 @@ > #include <linux/libata.h> > > #define DRV_NAME "pata_mpiix" > -#define DRV_VERSION "0.7.7" > +#define DRV_VERSION "0.7.6" > > enum { > IDETIM = 0x6C, /* IDE control register */ > @@ -146,7 +146,6 @@ static struct ata_port_operations mpiix_port_ops = { > .cable_detect = ata_cable_40wire, > .set_piomode = mpiix_set_piomode, > .prereset = mpiix_pre_reset, > - .sff_data_xfer = ata_sff_data_xfer32, > }; > > static int mpiix_init_one(struct pci_dev *dev, const struct > pci_device_id *id) > diff --git a/drivers/ata/pata_sil680.c b/drivers/ata/pata_sil680.c > index 9e764e5..83580a5 100644 > --- a/drivers/ata/pata_sil680.c > +++ b/drivers/ata/pata_sil680.c > @@ -32,7 +32,7 @@ > #include <linux/libata.h> > > #define DRV_NAME "pata_sil680" > -#define DRV_VERSION "0.4.9" > +#define DRV_VERSION "0.4.8" > > #define SIL680_MMIO_BAR 5 > > @@ -195,7 +195,7 @@ static struct scsi_host_template sil680_sht = { > }; > > static struct ata_port_operations sil680_port_ops = { > - .inherits = &ata_bmdma32_port_ops, > + .inherits = &ata_bmdma_port_ops, > .cable_detect = sil680_cable_detect, > .set_piomode = sil680_set_piomode, > .set_dmamode = sil680_set_dmamode, > diff --git a/drivers/ata/sata_via.c b/drivers/ata/sata_via.c > index 5c62da9..f9803a2 100644 > --- a/drivers/ata/sata_via.c > +++ b/drivers/ata/sata_via.c > @@ -566,7 +566,7 @@ static int svia_init_one(struct pci_dev *pdev, const > struct pci_device_id *ent) > static int printed_version; > unsigned int i; > int rc; > - struct ata_host *host; > + struct ata_host *uninitialized_var(host); > int board_id = (int) ent->driver_data; > const unsigned *bar_sizes;
*** This bug has been marked as a duplicate of bug 12609 ***
(In reply to comment #0) > Subject : [2.6.28-rc] Sata soft reset filling log > Submitter : Justin Madru <bevicm@dslextreme.com> > Date : 2008-12-13 2:07 > References : http://marc.info/?l=linux-kernel&m=122913412608533&w=4 > This entry is being used for tracking a regression from 2.6.27. Please don't > close it until the problem is fixed in the mainline. (In reply to comment #5) > > *** This bug has been marked as a duplicate of bug 12609 *** If this is indeed post-2.6.27 regression, it can't be a duplicate of that bug. 32-bit PIO was merged only at 2.6.29-rc1 time...
On Wednesday 18 February 2009, Justin Madru wrote: > Sergei Shtylyov wrote: > > Hello. > > > > Justin Madru wrote: > > > >>>>> If 12609 is truly a post-2.6.28 regression and 12263 is post-2.6.27 > >>>>> regresssion, this just cannot be. > >>>>> > >>>> Maybe the reporter of #12609 didn't notice/test kernels 28-rc1 to > >>>> 28. Or maybe > >>>> the difference in hardware is > >>>> the issue, but the bug is still the same. Don't know. > >>>> > >>> > >>> Sorry Justin, you must be confused: as Sergei says, > >>> #12609 and #12263 can only be different. > >>> > >>> I was one of the reporters of #12609, and I do know it's a post-2.6.28 > >>> regression (and Larry said so too), and one fix (not the preferred fix) > >>> is to revert the ata_bmdma32_port_ops from 2.6.29-rc, and the preferred > >>> fix is to improve the ata_sff_data_xfer32() introduced in 2.6.29-rc1. > >>> > >>> 2.6.28 does not contain any ata_bmdma32_port_ops, nor > >>> ata_sff_data_xfer32(), > >>> not did 2.6.28-rc1 contain them. So it is impossible for the > >>> reversion of > >>> the patch that introduced them to fix any problem on 2.6.28. > >>> > >>> I'm quite prepared to believe that your #12263 manifests similarly to > >>> #12609, and that a tip tree which contains a fix for #12609 contains > >>> a fix for #12263; but please, those bugs are not the same, and they > >>> don't have the same fix. > >>> > >>> Hugh > >>> > >>> > >> Well, like I said: "[I] Don't know". I'm not a kernel developer (or > >> even any developer... yet). > >> I'm just someone that tests the -rc kernels to see if there's any > >> problems with my hardware. > >> I try to report any regressions to lkml, and hopefully help the > >> developers. > >> > >> To me, who has no knowledge of all these low level issues, the > >> following error messages > >> look strikingly similar with a quick glance. > >> > >> # bug 12609 > >> # http://marc.info/?l=linux-kernel&m=123254501314058&w=4 > >> # > >> ata2.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen > >> ata2.00: cmd a0/00:00:00:00:00/00:00:00:00:00/a0 tag 0 > >> cdb 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 > >> res 51/20:03:00:00:00/00:00:00:00:00/a0 Emask 0x5 (timeout) > >> ata2.00: status: { DRDY ERR } > >> ata2: soft resetting link > >> ata2.00: configured for UDMA/33 > >> ata2: EH complete > >> > >> # bug 12263 > >> # http://marc.info/?l=linux-kernel&m=122913412608533&w=4 > >> # > >> ata2.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen > >> ata2.00: ST_FIRST: !(DRQ|ERR|DF) > >> ata2.00: cmd a0/00:00:00:00:00/00:00:00:00:00/a0 tag 0 > >> cdb 1e 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 > >> res 50/00:01:00:00:00/00:00:00:00:00/a0 Emask 0x2 (HSM > >> violation) > > > > Note the different value of the status, error and interrupt reason > > registers: 51/20:03 vs 50/00:01. The former means (unexpected?) status > > phase interrupt with error indication and the sense key NOT READY, the > > latter means (unexpected?) command phase interrupt with no error. > > IIUC, the former happens once the 'sr' driver first sends the TEST > > UNIT READY command while probing the CD/DVD drive, the latter seems to > > be a result of some polling process (originated from userland) -- I'm > > not seeing ALLOW_MEDIUM_REMOVAL anywhere in this driver. So they only > > look similar, I think... > > And that is why I'm a tester and you're a developer ;) Thanks for the > info! Next time I'll look closer > and maybe know what I'm actually looking at. > > > > > >> So, will the patch for 12609 fix my issue also, or does there need to > >> be another patch? > > > > Most probably it'll need another patch. > So then, #12263 should be reopened and marked as not a duplicate. > Anyways, if tip/master gets merged how it is now then my bug should be > fixed.