aboutsummaryrefslogtreecommitdiff
path: root/Documentation/scsi/qlogicfas.txt
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 15:20:36 -0700
committerLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 15:20:36 -0700
commit1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch)
tree0bba044c4ce775e45a88a51686b5d9f90697ea9d /Documentation/scsi/qlogicfas.txt
Linux-2.6.12-rc2
Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
Diffstat (limited to 'Documentation/scsi/qlogicfas.txt')
-rw-r--r--Documentation/scsi/qlogicfas.txt79
1 files changed, 79 insertions, 0 deletions
diff --git a/Documentation/scsi/qlogicfas.txt b/Documentation/scsi/qlogicfas.txt
new file mode 100644
index 00000000000..398f9916807
--- /dev/null
+++ b/Documentation/scsi/qlogicfas.txt
@@ -0,0 +1,79 @@
+
+This driver supports the Qlogic FASXXX family of chips. This driver
+only works with the ISA, VLB, and PCMCIA versions of the Qlogic
+FastSCSI! cards as well as any other card based on the FASXX chip
+(including the Control Concepts SCSI/IDE/SIO/PIO/FDC cards).
+
+This driver does NOT support the PCI version. Support for these PCI
+Qlogic boards:
+
+ * IQ-PCI
+ * IQ-PCI-10
+ * IQ-PCI-D
+
+is provided by the qlogicisp.c driver. Check README.qlogicisp for
+details.
+
+Nor does it support the PCI-Basic, which is supported by the
+'am53c974' driver.
+
+PCMCIA SUPPORT
+
+This currently only works if the card is enabled first from DOS. This
+means you will have to load your socket and card services, and
+QL41DOS.SYS and QL40ENBL.SYS. These are a minimum, but loading the
+rest of the modules won't interfere with the operation. The next
+thing to do is load the kernel without resetting the hardware, which
+can be a simple ctrl-alt-delete with a boot floppy, or by using
+loadlin with the kernel image accessible from DOS. If you are using
+the Linux PCMCIA driver, you will have to adjust it or otherwise stop
+it from configuring the card.
+
+I am working with the PCMCIA group to make it more flexible, but that
+may take a while.
+
+ALL CARDS
+
+The top of the qlogic.c file has a number of defines that controls
+configuration. As shipped, it provides a balance between speed and
+function. If there are any problems, try setting SLOW_CABLE to 1, and
+then try changing USE_IRQ and TURBO_PDMA to zero. If you are familiar
+with SCSI, there are other settings which can tune the bus.
+
+It may be a good idea to enable RESET_AT_START, especially if the
+devices may not have been just powered up, or if you are restarting
+after a crash, since they may be busy trying to complete the last
+command or something. It comes up faster if this is set to zero, and
+if you have reliable hardware and connections it may be more useful to
+not reset things.
+
+SOME TROUBLESHOOTING TIPS
+
+Make sure it works properly under DOS. You should also do an initial FDISK
+on a new drive if you want partitions.
+
+Don't enable all the speedups first. If anything is wrong, they will make
+any problem worse.
+
+IMPORTANT
+
+The best way to test if your cables, termination, etc. are good is to
+copy a very big file (e.g. a doublespace container file, or a very
+large executable or archive). It should be at least 5 megabytes, but
+you can do multiple tests on smaller files. Then do a COMP to verify
+that the file copied properly. (Turn off all caching when doing these
+tests, otherwise you will test your RAM and not the files). Then do
+10 COMPs, comparing the same file on the SCSI hard drive, i.e. "COMP
+realbig.doc realbig.doc". Then do it after the computer gets warm.
+
+I noticed my system which seems to work 100% would fail this test if
+the computer was left on for a few hours. It was worse with longer
+cables, and more devices on the SCSI bus. What seems to happen is
+that it gets a false ACK causing an extra byte to be inserted into the
+stream (and this is not detected). This can be caused by bad
+termination (the ACK can be reflected), or by noise when the chips
+work less well because of the heat, or when cables get too long for
+the speed.
+
+Remember, if it doesn't work under DOS, it probably won't work under
+Linux.