Announcement

Collapse
No announcement yet.

Legacy IDE Driver Now Deprecated, To Be Removed From Linux In 2021

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    If IDE drivers are going away and replaced with libata PATA drivers, that would be bad for the old Powerbook G4 Titanium that I have around.

    There is some unholy interaction between AGP and libata PATA drivers, that the system will become unstable and hard lockup after a while. No problem with IDE drivers, or if I force AGP to off.

    Comment


    • #22
      Originally posted by monraaf

      Why do you think you speak for everyone?
      Just my opinion, but like I said, we all have choices. More frustrations from work where we're fighting to get folks to move their code to 64bit that has been a 10+ year battle, and they're stalling moving to RHEL7 & 8 to keep using some old RHEL5 and 6 systems in 32bit.

      Comment


      • #23
        Originally posted by L_A_G View Post
        Good to see that they were able to kick out the old PATA code like this, but it would also be nice if they could kick out the old SCSI code in a similar way. However in the early 2000s device makers decided not to kill SCSI but to create a serial and electrically SATA-compatible version called SAS. Now SATA drivers and controllers are now so heavily intertwined with SCSI that you can't get rid of the drivers for them without getting rid of SATA drivers at the same time.

        Found this out the hard way when I was setting up a minimal kernel for an embedded application with very high uptime and IP security requirements when my efforts to remove unnecessary cruft (less to go wrong and be to attacked) ended up killing the SATA port in the process.
        ​​
        Should be pretty logical, if you have ever used SAS drives/adapters. Switching cables behind SAS adapter/RAID controller suffices for running SATA drives. SATA is connector-wise and electrically subset of SAS but not vice versa. You cannot run SAS drives off SATA controller, voltage requirements differ.
        Imagine you'd have system with SAS controller, you'd benefit there. Upon drive dying you have option of using SATA or SAS. You are not locked into one form factor. Not all bad thus.

        Comment


        • #24
          Originally posted by aht0 View Post
          ...
          My issues with this is that SAS itself is kind of pointless. It's really is just a superset of SATA made to piggyback on it and keep a practically ancient standard (SCSI) alive.

          It's just a whole bunch of unnecessary complexity for something I can't see having any tangible benefit. SATA came first and then SAS was made as a superset of it to piggyback off it's success.

          Comment


          • #25
            Ehm.. It can be faster (example: I have some Dell 12Gbps 2,5" inch SAS drived here), cables could be much longer, SAS has command queue offloading taking work from CPU and SAS has better failure prediction. For mech drives, SAS has better seek times and better average read/write speeds, because drives can spin 10k rpm or even 15krpm as opposed to SATA's 5400rpm or 7200rpm.

            Yeah, Nvme or m2 have theoretically faster interfaces but amount you can use is limited by motherboard. Bunch of 12Gbps SAS ssds in a backplane behind single SAS adapter could beat what you can install on a your PC or server.

            SATA servers are more sluggish in average response times, compared to SAS servers, even if you want straight comparison. Before you mention SATA SSD's, SAS SSD's exist too and are superior. Enterprise SATA actually exists. Fact that SAS still prospers, should be hint that SATA ain't so fully awesome as you imagine.
            ​​​​​​
            Not all SATA controllers even support hotswapping.
            Last edited by aht0; 14 May 2019, 02:27 AM.

            Comment


            • #26
              Originally posted by starshipeleven View Post
              excellent, I'm waiting for people to not post this kind of bullshit.
              x86 is going to be deprecated so I don't understand your reluctance.

              Comment


              • #27
                Originally posted by Azrael5 View Post
                x86 is going to be deprecated so I don't understand your reluctance.
                x86 isn't going to be deprecated in Linux kernel for at least another decade.

                Desktop distros can deprecate 32bit images, but that does not mean there is no more x86 hardware around used for other tasks with other specialist or embedded device distros like OpenWrt or Yocto or whatever.

                Even OPNSense for example (which is a router/firewall distro based on FreeBSD) is not planning to drop support for 32bit hardware. Nor is IPFire (Linux-based firewall distro).

                Comment


                • #28
                  Originally posted by starshipeleven View Post
                  x86 isn't going to be deprecated in Linux kernel for at least another decade.

                  Desktop distros can deprecate 32bit images, but that does not mean there is no more x86 hardware around used for other tasks with other specialist or embedded device distros like OpenWrt or Yocto or whatever.

                  Even OPNSense for example (which is a router/firewall distro based on FreeBSD) is not planning to drop support for 32bit hardware. Nor is IPFire (Linux-based firewall distro).
                  just 2 years and x86 will be a souvenir of inefficiency.

                  Comment


                  • #29
                    Originally posted by Azrael5 View Post
                    just 2 years and x86 will be a souvenir of inefficiency.
                    Low power parts back then are still low power parts now. AMD Geodes and Atom boards won't suddenly start consuming more just because you think they are obsolete.

                    Comment


                    • #30
                      Originally posted by starshipeleven View Post
                      Low power parts back then are still low power parts now. AMD Geodes and Atom boards won't suddenly start consuming more just because you think they are obsolete.
                      Assuming more Intel-related security mitigation in future, even performance of these ancient boards ain't going to be that bad relatively speaking, lol.

                      Comment

                      Working...
                      X