On Wed, Nov 18, 2009 at 3:11 PM, Ben Scott <[email protected]> wrote:
> On Wed, Nov 18, 2009 at 1:07 PM, Alan Johnson <[email protected]> wrote: > > The only reason I've ever had to install a driver > > for a RAID controller is for online management. As far as drive access, > all > > the controllers I've come across just look like any other SATA or SCSI > > controller from a device exposure. > > There is no standard software/hardware interface for SATA or SCSI > controllers.[1] > I think this is a very bold statement, but perhaps we are only separated by systematics here. If we replace the word "standard" with the word "generic", does that work? Would you agree that there are plenty of SATA and SCSI drivers that work with most or all correctly implemented devices? Still, I would argue that SATA and SCSI are definitively standards, and that they include both physical and electronic hardware specifications, as well as software specifications, and probably a reference driver that works with any devices that properly implements the other bits. If all of these things are missing, then what isn't SATA/SCSI? I don't expect you would disagree with that, so perhaps I am missing your point. > Some of Intel's ATA chipsets come close -- they're widely emulated > and might qualify as a de facto standard -- but even Intel changes > their stuff now and again. That's why Win XP needs drivers for many > SATA chipsets -- they don't look like the venerable PIIX IDE > controller anymore. > My understanding is that WinXP needs SATA drivers because the latest installer predates the widespread adoption of SATA devices. Looking at a CentOS 5.x box I have near to hand (kernel package > 2.6.18-164.6.1.el5), the kernel has roughly 55 different SATA/SCSI > drivers available. There are at least three for Adaptec chipsets > alone (aic7xxx, aic7xxx_old, aic94xx). > Can you be sure that there is not one generic or reference driver in there that does not cover nearly any standard SATA device? Another for SCSI? Likely the rest are for accessing additional features or performance enhancements that are implemented by the relative controllers but are not part of the standard. Management interfaces would be included here. A more extreme example in another device category would be the VGA standard. You can buy the latest, cutting-edge video card that uses data swapping algorithms your OS has never heard of, put it in your computer, and if it properly implements VGA like every other modern video card, the OS will boot and display something useful on it. If I want to access all the icons on my lower panel, well, I better get the drivers specific to the card and introduce them to the OS. I'm not sure exactly what you mean when you say you've never needed > to install a driver, but your Linux computer is certainly loading a > driver for your disk controller, regardless of whether you've got a > > RAID controller or just a plain old disk controller. If you mean your preferred Linux distro came with the driver > built-in, well, that's good, but there's still a driver being loaded, > and if that driver is buggy you'll still be having a bad time[2]. > I think you got my gist. I only mean to say that the only RAID controller specific drivers I have had to install were for accessing the management features on line. It is entirely possible that there were drivers specific for every RAID controller I have ever installed an OS on built into the installer I was using. Like you, I have had to provider access drivers to older OS installers (SATA on XP, SCSI on NT and 2K), but because I used the driver disk that came with the device does not mean that is the only driver that would get the installer access to the controller or drives on it. That said, I concede that my point is not conclusive either. I've just been lead me to believe. > If you mean your preferred RAID controller emulates some other > interface, that's also valid, but Linux will still needs a driver for > that other interface. That's exactly what I'm suggesting. I'm suggesting they have implemented IDE, SATA, SCSI, or SAS interfaces for basic data access to the drive (or in this case, RAID volume). > For example, I've seen gizmos with three IDE > connectors. You connect a disk each to two of those connectors. You > connect the third to the motherboard's IDE connector. The card makes > the two IDE disks look like one to the motherboard. So Linux does not > need to know about that RAID card -- but it still needs a driver for > the motherboard's IDE controller. > Again, said driver for the motherboard's IDE controller is not necessarily specific to that controller, but is just as possibly a reference implementation. Also, there was nothing specific to your IDE RAID gizmo. It just made itself look like a drive, which is a layer lower, but is the same idea. > [1] Well, unless you count BIOS INT13, but Linux can't use the > real-mode BIOS, and even MS-DOS systems generally loaded supplementary > drivers because INT13 sucks so much. > This is in the realm of what I am talking about. > [2] Myself, I've never had a problem the "megaraid" driver that's been > part of the standard Linux kernel since circa 2001. Obviously, > experiences vary. > I'm not familiar with megaraid, but this also sounds a lot like a generic driver that works with devices that adhere to some standard or less formal set of common rules. > -- Ben I think we are mostly on the same page, but are my ramblings above more in line with your understanding?
_______________________________________________ gnhlug-discuss mailing list [email protected] http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
