Re: [usb-storage] Problems with USB MMC/SD card reader

> Feb 23 08:44:16 kepler kernel: SCSI device sdb: 1988608 512-byte hdwr sectors (1018 MB)
> ...
> Feb 23 08:44:16 kepler kernel: SCSI device sdb: 1988608 512-byte hdwr sectors (1018 MB)

Is stuttering the log of max Lba normal? I think I remember yes?

> but immediately after that the logs get flooded by error messages:
> ...
> Feb 23 08:44:18 kepler kernel: sdb: Current: sense key=0x3
> Feb 23 08:44:18 kepler kernel: ASC=0x11 ASCQ=0x0
> Feb 23 08:44:18 kepler kernel: Info fld=0x1e51f9
> Feb 23 08:44:18 kepler kernel: end_request: I/O error, dev sdb, sector 1987065
> Feb 23 08:44:18 kepler kernel: Buffer I/O error on device sdb1, logical block 248352
> and so on until I disconnect the device.

Ick.

x 3 11 = SK ASC from a Scsi device is a claim that Ecc is saying one or more bits of the data read are wrong, as if your flash media were wearing out. Here above we were told max Lba = 1988608 - 1 = x1E57FF.

"Info fld" is probably kmesg-ese meaning sense bytes 3:4:5:6 echoed when byte 0 mask x80 set, i.e., the Lba in error. In this snippet of log we have:

x 3 11 00 1E51F9
x 3 11 00 1E51F9
x 3 11 00 1E5271
x 3 11 00 1E5271
x 3 11 00 1E5271

This isn't a consistently repeated log - our host & device together haven't livelocked in one spot - instead they have reported the first two of a burst of two or more errors that are being tried 2 or 3 or N times.

> ... works perfecly on the same machine in Windows XP ...

Might Linux differ from Windows by failing early by design if the last blocks are unreliable?

I see we have reason to believe Windows can read data from this card thru this reader without error. Do we also have reason to believe Windows can read all blocks correctly without error?

Have we looked at the event log in Windows? I mean the one near the Disk Manager and Device Manager, that divides into App, Security, and System. x 3 11 complaints from the device often show up there. To see recent events, you can sort by timestamp or clear before experimenting.

If I were you, I'd next try plugging in without automount and then a dd, both in Linux and in the Windows patched to include dd, to see if those dd agree that all the bits read without error.

I'd expect dd reports failed reads in both Linux and Windows, until I had reason to think otherwise, like an evil root cause for the stuttered log of max Lba.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Relevant Pages

  • Re: Drive Partitioning vs. Updated Bios
    ... NOTE: If you enable 48-bit LBA ATAPI support by editing the preceding registry key, but your system does not meet the minimum requirements, you may observe the following behaviors: ... Operating systems that do not have 48-bit LBA support enabled by default (such as Microsoft Windows 98, Microsoft Windows Millennium Edition (Me), or Windows 2000) that are installed on a partition that spans beyond the 28-bit LBA boundary will experience data corruption or data loss. ... if the BIOS's interpretation of the hard disk parameters is sufficient to get the machine started and hand over control to Windows then you're safe. ...
    (microsoft.public.win2000.general)
  • Re: Corrupted hard drive
    ... 48-bit LBA running XP SP2. ... up on getting the data back, but that won't fix a corrupted MFT will it? ... don't run fdisk /mbr or chkdsk on it if you value your data. ... accesssible by Windows. ...
    (microsoft.public.windowsxp.hardware)
  • Re: Large Drive problems
    ... This article describes the Microsoft Windows XP Service Pack 1 ... 48-bit Logical Block Addressing (LBA) support for ATA Packet Interface ... disk drives that can increase the capacity of your hard disk ...
    (microsoft.public.windowsxp.device_driver.dev)
  • Re: LBA in diskperf;
    ... the byteoffset by the LBA size. ... Windows 2k/XP/2k3 Filesystem and Driver Consulting ... the byte offset in the partition. ...
    (microsoft.public.development.device.drivers)
  • Re: 4 CONSECUTIVE CORRUPT DISK DISASTERS WITH WIN2K
    ... (such as Microsoft Windows 98, Microsoft Windows Millennium Edition (Me), or ... LBA boundary will experience data corruption or data loss. ... If I perform the EnableBigLba regfix on a 200g disk with Win2K ...
    (microsoft.public.win2000.general)