Fatal error in reading USB key

[Anchieri]

Actually I use ARCAOS 5.0 that come installed without any problem and so I have upgraded installed version of fat32 with the new:fat32-0.10-r293-os2.wpi, for testing. using:

• SanDisk Cruzer Edge 16 GB this function perfectly, I can read and Write on it

• Kingston DTSE9 G2 USB 3.0 16 GB with this the system hung immediately and I have to power off and then reboot the system (CAD don't function).

I thought that this error is due to the new widget RemMedia.dll and uninstall it, but the problem remain, this USB device is not recognized and the system hung.

Using the official version fat32_0913.wpi I can read and write on this USB key.

[Valery V. Sedletski]

Too little info. Are there any logs?

USB 3.0??? No USB 3.0. support under OS/2...

[Anchieri]

I'm have installed ARCAOS 5.0 that has installed version 11.15 of USB stack, and if you read the previous comment, using version fat32_0913.wpi i'm able to read/write over this USB 3.0 key. With your version of fat32 this isn't possible, the system hung after inserting it in a USB slot. The system stop without any information or data logged on populog.os2

G.M. Anchieri

[Valery V. Sedletski]

2Anchieri: Ok. But where are your logs? How can I determine what is happening on your machine? Please, take your f32mon.exe log.

PS: I'll need the logging output taken when you insert the USB stick. Most probaably, after reboot your fat32.log will be zero size, because your machine was hanged + a hard reboot. In this case you'll need to have a COM port, a null-modem cable and another machine with terminal emulator attached. But please first try running f32mon.exe.

[Valery V. Sedletski]

Retested my USB 3.0 hard disk case -- it works fine with USB drivers from eCS 2.2 and fat32.ifs r294. I see a fat32 partition on it fine.

PS: Did you tried to run CHKDSK under windows first? Maybe, it hangs on mount because of some FS errors. As I see my USB 3.0 harddisk working, and one user said me that USB 3.0 flash disk is working fine on his machine (with newer versions of fat32.ifs). If CHKDSK will not help, I'd suggest to format the volume first and try again. But it is desirable to see why it hangs first, of course.

[Anchieri]

I've executed the CHKDSK using my WINXP partition, the disk has no error and all data in it is visible and I can hear my song in it. Returned on ARCAOS 5.0 I installed your r295 version and retested, after inserting the USB key in a USB slot a message:

• a new USB device has been attached, please wait...

the system hang immediately without other messages. Fat32mon.log contain this two line:

• P:5a T:d6 D:0 T:18:582 FS_FSCTL returned 0
• program aborted! rc = 0

I don't now if this is of a some utility for you.

G.M. Anchieri

[Valery V. Sedletski]

It contains only that message after you rebooted? It looks that the log did not flushed and all messages are missing. In this case you need the COM port log. Do you have a COM port, null modem cable and second machine? Please add the "/monitor" switch to fat32.ifs command line and take a log. (Otherwise I can't help you).

[Anchieri]

This disk is exFat for DFSEE, and I've inserted /exfat on fat32.ifs line and tested, but the system hang as usually.

For aid you in resolving this bug I've do this step:

• REMmed the two FAT32 statement in config.sys
• rebooted the system
• inserted the USB key in a USB slot, the system attach it as drive :J
• using DFSEE I've obtained three image: dfsee-01.jpg, dfsee-02.jpg dfsee-03.jpg that I attach.

If you analyze this image it's most probable that you can resolve this ticket.

G.M. Anchieri

[Valery V. Sedletski]

It is FAT32 and not exFAT, as I can see from these screenshots. But it is useless. I need a COM port log. Do you have a COM port, null modem cable and a second machine (a laptop with a usb2serial converter could be used too)?

[Anchieri]

I'm sorry, I've a COM port but don't have a second machine to use to debug program. I can Use DFSEE to send you other information if this can help, but nothing more.

In the meanwhile I can use the old version fat32_0913.wpi that function for me. If you agree, I can send you this USB key and you can test it, when the problem has been debugged, you can return me this key.

G.M. Anchieri

[Valery V. Sedletski]

I'm sorry, I've a COM port but don't have a second machine to use to debug program.

So, you even haven't a laptop?

I can Use DFSEE to send you other information if this can help, but nothing more.

I need to know, what's happening inside fat32.ifs, so, dfsee cannot help in that area. Unfortunately, f32mon can't help too, because your system hangs, and not all debug messages gets into the log. Here, only a COM port can help. Getting a sysdump will not help too, because a dump will not contain the log info too.

If you agree, I can send you this USB key and you can test it, when the problem has been debugged, you can return me this key.

That is also possible. I can return it to you, of course, though, I didn't tried to send something by mail to other countries. But we could try. If you don't fear that your flash disk could be lost during delivery.

There would be a one more possibility. Maybe, I could add a systrace support to fat32.ifs. Then, if it hanged, you could take a sysdump and send it to me. Then I'll need to extract a trace buffer from a sysdump. But I don't know if it is possible. Also, I'll need some amount of work to add the trace support to fat32.ifs. And I'm not sure if it is possible to extract a trace buffer from a sysdump. Need to consult with people...

Do you have a SADUMP partition on your machine? Is your hard disk big? Bigger than 512 GB?

[Valery V. Sedletski]

Ok, I reread the Debugging Handbook and see a .ts command which can be used to extract a trace buffer from a system dump. So, I was right. I could try adding trace support to fat32.ifs.

[Valery V. Sedletski]

2Anchieri: I added system trace support in r296 (please see ​ftp://osfree.org/upload/fat32/). Now it is possible to use system dump for analysis. Please try creating a dump partition and take a system dump (to have the system dump smaller, you can reduce memory available to OS/2 via QS_INIT loader). Then you can load the dump in PMDF and extract a trace buffer and send it to me. (I don't need a whole dump, just a trace buffer). For that, use ".ts fat32.trc" command in PMDF. If you have questions, please ask.

PS: You just create a FAT partition with a SAMUMP label (or, use dumpfs.ifs for memory dumps bigger than 2 gb). For taking system dumps, add a TRAPDUMP=R0,D: (where d: is your SADUMP partition) to config.sys. For system trace, you need to add the "TRACE ON 254" and "TRACEBUF=512" to your config.sys and reboot.

[Anchieri]

I've created a SADUMP FAT partition, but this type of partition is only 2GB in size and my PC has 3GB of RAM, how do I can save this dump?. It's possible to reduce the size of my RAM using BIOS?

G.M. Anchieri

[Valery V. Sedletski]

It is possible, but not with BIOS. You just install QS_INIT loader and add "memlimit=512" to kernel parameters, then only 512 MB will be available. QS_INIT loader is available from there: ​ftp://212.12.30.18/public/QS/QS_LDR.ZIP.

[Valery V. Sedletski]

Also, except for FAT-based SADUMP partition, you can use dumpfs. The setup guide is available from Steven Levine's site: ​http://home.earthlink.net/~steve53/os2diags/dumpfs-user-guide.txt. The link to dumpfs.zip package is in the document. Dumpfs allows SADUMP partitions of bigger sizes, up to 4 GB. Both standard IBM os2dump and os2dump from dumpfs distribution support only standard hard disk geometry, i.e., 63 sectors per track. Hard disks bigger than 512 GB have 127 or 255 sectors per track, so it may not work. Is your hard disk > 512 GB? If so, you need a smaller hard disk. I have a 2 TB hard disk, but there is also a 80 GB hard disk available. So, I'll try install a SADUMP partition onto it.

[Anchieri]

My HD is of 160GB and I have created a 2GB empty partition on it.

• Using WinXp I've formatted in FAT mode this partition
• inserting the MEMLIMIT=1024 in OS2LDR.INI file
• os2krnl =Current OS2,LOGSIZE=4096,PRELOAD,MEMLIMIT=1024
• using PMVIEW I obtained a .JPG image of LVM information that I attach
• then I stop clear trace and start it
• inserted my USB key in a USB slot

The system stop with this messages:

• ERROR: THE TRACE PARTITION IS NOT LABELED SADUMP
• UNABLE TO DUMP TO HARDFILE, DO YOU WISH TO DUMP TO DISKETTE? (Y or F1)/N

Keying ENTER the system reboot.

G.M. Anchieri

[Anchieri]

LVM partition

[Valery V. Sedletski]

Did you labeled your SADUMP partition as SADUMP? The volume label should be "SADUMP".

then I stop clear trace and start it

You pressed Ctrl-Alt-NumLock-NumLock after it hanged? Or it automatically made dumping?

[Anchieri]

Finally after installing DUMPFS and generating a Volume of 4GB I was able to obtaining a trapdump (file name: DUMPDATA.002, THE DUMPDATA.001 has been obtained with CTRL+ALT+F10+F10 as test) of only 512MB.

But now I don't now how extract the trace data from this file. do you can give me tips on how to do that?.

G.M. Anchieri

[Valery V. Sedletski]

Good. But no reason to get a full 4 GB dump. It takes about 30 minutes to dump 4 GB of RAM. 512 MB takes not so long time. And if you need to transfer it to developer, 512 MB is easier to upload than 4 GB and it takes less space on ftp.

Also, I discovered today that if you create a 4 GB FAT partition with a 64 KB cluster, FAT version of os2dump works fine with it. I managed to format such a partition with fat32.ifs, dumped a full 4 GB of RAM, a 3.7 GB dump file was created. Files > 2 GB are supported by newer versions of fat32.ifs, so I copied it to a JFS partition and loaded into the dump formatter. But when opening such a partition in a file manager, a division by zero had occured. So, I needed to do some fixes (not finished so far). But now it appears that using dumpfs can be avoided. Standard FAT os2dump + fat32.ifs should be sufficient.

To extract the trace file from a dump, you need to issue a command:

.ts somefile.trc

in the dump formatter. Then a somefile.trc file should appear in the current directory. BTW, you'd better take a trap dump with 14.104_kernel loaded. Each kernel version needs a dump formatter of corresponding version (df_ret.exe for retail kernel and df_deb.exe for a debug kernel). You just go into \os2\pdpsi\pmdf directory, create a "somedir" subdirectory, put df_ret.exe and/or df_deb.exe into that dir, and symbol files (with *.sym extension) to the same dir. You'll need a .sym files at least, for kernel (os2krnlr.sym for a retail kernel, os2krnld.sym for a debug kernel), fat32.sym for fat32.ifs. Also, you should add a line like

somedir;14.104a_SMP;14.104

in \os2\pdpsi\pmdf\pmdfvers.lst file. So that, pmdf will search needed files for a 14.104a_SMP kernel in "somedir" subdirectory.

You start pmdf.exe and load a dump file into it. then you issue a ".ts fat32.trc" command, and you get a fat32.trc file, it should be about 512 KB in size, if you specified a "TRACEBUF=512" command in your config.sys.

[Valery V. Sedletski]

I fixed division by zero traps when accessing a FAT12/FAT16 partition with a 64 KB cluster, since r297, so now it's possible to use a 4 GB FAT partition, instead of dumpfs.

[Anchieri]

I've first uninstalled the DUMPFS, then installed your last version r297, created a new volume/partition of 4055MB labeled: SADUMP but I can't format it, do you can give me the correct command to use for format it as FAT partition as is in your previous comment?

G.M. Anchieri

[Valery V. Sedletski]

You should format it with fat32.ifs. The correct command is:

format /fs:fat16 /c:128

Here, 128 sectors per cluster (64 KB per cluster) is used, and a FAT16 file system. If you specify FAT filesystem (/fs:fat) then standard IBM's FORMAT routine will be used. My FORMAT understands FAT12, FAT16, FAT32, EXFAT as a /fs:... parameters.

[Anchieri]

Please, I'm crazy, I don't now how to extract this information from the DUMPDATA.001 file, because the program pmdf.exe give this result:

• IBM OS/2 Dump Formatter for a retail or an hstrict SMP kernel.
• Formatter is --> Internal revision 14.103a_SMP
• Dump file is --> Internal revision 14.200_SMP (system dump)

Symbol (c:\os2\pdpsi\pmdf\trap\os2krnlr.sym) linked

.tr fat32.trc

after many time of waiting, nothing succeed and I kill it.

Do you can tell me what I've missed or I'm wrong?

G.M. Anchieri

[Valery V. Sedletski]

You can try opening the dump directly from df_ret.exe, like this:

df_ret.exe l:\dumpdata.001

and enter commands in this VIO application. Also, taking a dump with 14.104 or 14.103 kernel can help. With 14.106 kernel, I got a hang on a .ts command too, if I correctly remember, so I'd suspect, the same is with 14.200. You can get 14.104 kernel from ​http://www.os2site.com. You need to take the dump again, but with 14.104 kernel, then write 14.104a_SMP in pmdfvers.lst file if you want to open it from pmdf.exe.

PS: .ts command and not .tr

PPS: You can install 14.104_SMP for testing purposes. It can be installed in QS_INIT as a separate menu item without removing your 14.200.

[Anchieri]

I've installed version 14.104a taken from my eCS 2.0ga but now we have another problem , with this kernel I don't have anymore this hang, but I can view the all content of the USB key and ear the song in it. Therefore the problem reside in kernel past version 10.104a, because using 14.106 or 14.200 we have this bug.

G.M. Anchieri

[Anchieri]

This bug is caused, for me, by a program instruction not correctly compiled by the compiler used, that don't agreed with the change introduced in new version of kernel.

G.M. Anchieri

[Valery V. Sedletski]

Hm. Can you share a 14.200_SMP kernel with me, to test it on my machine? Also, can you test it on 14.106 kernel, if it wil woork the same as 14.200? I have 14.106 and it works for me. Maybe, for you too?

Also, you could try getting an image of your flash disk with dfsee and sending it to me, if it possible. But better an image of a whole physical (not logical) disk.

[Anchieri]

Yes I can send you a 14.200_SMP kernel, and I just tested with 14.106 kernel obtaining the same result, if you will I can send you the DUMPDATA.001 file with the hang of 14.106 kernel that I have imported on ARCAOS 5.0 from my eCS 2.2b1, for the DFSEE image of the flash USB key, I have just attached 3 image of it, taken from view of DFSEE on eCS 2.2b1 that has installed the officially version 0.913 of FAT32 drivers:

• dfsee-01.jpg, dfsee-02.jpg and dfsee-03.jpg

My eMail is: giuliomario.anchieri at alice.it

if you want we can continue this debug session using the eMail.

Thank you, G.M. Anchieri

[Valery V. Sedletski]

Yes, please send me the 14.200 kernel (SMP one), my email is _valerius (at) mail (dot) ru. The dump file for 14.106/14.200 is not needed because it is problematic to extract a trace buffer from it. But it is desirable if you upload somewhere an image of your flash disk. Unfortunately, I have not more than 8 GB on my VPS, where I have an ftp, so maybe, some Google drive or something like this.

and I just tested with 14.106 kernel obtaining the same result,

So, it hangs with 14.106 too, or?

[Anchieri]

Yes, it hang also with 14106 version.

G.M. Anchieri

[Anchieri]

Dear Valerius,

I don't now if you have received my wife eMail that contained the link to www.box.com for download the image of that usb key and if you have had the time to analyze it.

Here the link to download the KINGSTON.7z file: ​https://app.box.com/s/et4vafcmalmes2ju973j3c43rnc00qpl

After my holiday I've necessity to use this usb key and so I've reformatted it using your fat32-0.10-r299-os2.wpi, with my great surprise this key function perfectly now under all smp kernel from smp10104 to smp10201.

Now for me this ticket has been resolved and you can close it.

Thank you for your work on FAT32, now this system is almost complete, apart the possibility to add the defrag option.

G.M. Anchieri

[Valery V. Sedletski]

I don't now if you have received my wife eMail that contained the link to www.box.com for download the image of that usb key and if you have had the time to analyze it.

Here the link to download the KINGSTON.7z file: ​​https://app.box.com/s/et4vafcmalmes2ju973j3c43rnc00qpl

I received that letter from you (from your wife's address), but I was unable to open the link you provided. The browser just opens a blank page. But today when I tried the latest Firefox 45, It did opened and I downloaded successfully

After my holiday I've necessity to use this usb key and so I've reformatted it using your fat32-0.10-r299-os2.wpi, with my great surprise this key function perfectly now under all smp kernel from smp10104 to smp10201.

So, it looks that your flash stick had some logical errors, that's why the machine did hanged. After reformatting it becomes clean from errors. Also I suspected that it may contain physical errors, because you said you had much "sector not found" errors from dfsee side. So, it would be some hardware errors (flash memory integrity or whatever). It is good that it now mounts ok. But the errors may show up later. So, i'd recommend to buy a new flash stick, to avoid errors in the future.

So, I glad to see that the problem is resolved.