Announcement

Collapse
No announcement yet.

The New Microsoft exFAT File-System Driver Is Set To Land With Linux 5.7

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

  • #41
    Originally posted by chithanh View Post
    What exactly do you mean?
    See my post here for detailed information and citations.

    Summary:
    • Until 2013, UDF on Linux would incorrectly claim the filesystem was full for storage media >64GB
    • POSIX permissions is a mute point for UDF in regards to portable storage as both users need mounting of it configured appropriately, the default automount support tends to ignore it otherwise.
    • Block size logic for UDF has varied with previous kernels, but now expects logical sector size, typically this will be 512 blocks, for modern Windows and Linux(not sure about macOS), you should be able to use 4k blocks if the device is 4kn(which is only certain HDDs afaik, SSDs still offer 512e?). Therefore max capacity for UDF in a portable cross-OS storage media, is 2TiB(or 16TiB on 4kn), exFAT can achieve up to 128PiB by comparison.
    • Only 127 unicode or 254 ASCII characters for max file name length, FAT32/exFAT/NTFS can all handle 255 unicode.
    • Apple apparently made changes in 2018 and the UDF support got worse with High Sierra or just doesn't exist anymore? A user was able to do it via a third-party UDF driver from Sony.

    From others in the discussion:
    • > Which is not UDFs fault, but how it's implemented. It's not supposed to be partitioned at all and Linux can handle that fine. I think Windows just can't handle HDDs without a partition table though… maybe because of how it assigns drive letters or something. - Source
    • > All platforms' UDF implementations are lacking in one way or another. None of them has a complete, proper implementation. Linux has no way to repair a broken UDF disk for example, but Windows also expects a specific setup and their repair tool sometimes even makes it worse. - Source(Same as prior)
    • > The best option should be a single UDF 1.50 partition spanning the whole removable flash drive and including an MBR record in the first - Source
    • > The broader problem with UDF on non-optical media is expectations about partitioning. One of Windows and macOS wants UDF inside a partition; the other wants it on the raw device itself. Linux doesn't care either way - Source

    The useful format-udf tool points out further differences to keep in mind when choosing UDF as a cross-platform FS for portable storage. Worth noting:
    • Windows expects a partition table for HDDs(unclear about SSDs), but Flash Drives do not.
    • macOS expects a full disk to be UDF, not just a partition.
    Note the compatibility table for block sizes, I have two SSDs on this system, one is 512e, and another 512n, the 512e is the better SSD with much higher capacity(1TB Crucial MX500 vs 128GB Colorful SL300).. Apparently the Crucial 512e would not work with UDF on Windows 10?

    Not that exFAT is perfect, but far less hassle as you don't need to be aware of all those gotchas with UDF. Have had exFAT corrupt itself though, so it's only really useful for temporary cross-platform portable data, not reliable external storage. Useful for portable storage as it likely sees better performance when it comes to transferring data to/from the device vs UDF, due to better block size support.
    Last edited by polarathene; 09 March 2020, 08:46 PM.

    Comment


    • #42
      UDF partition working fine here, after ditching NTFS partition for archiving WIndows' related software. The heck with exFat, not resizeable unless you buy a China made PartitionGuru software partition tool, although Wikipedia states exfat is not resizeable.

      Comment


      • #43
        Originally posted by calc View Post
        Are you sure there is PC hardware newer than 2011 that a company actually wrote an old style BIOS for instead of just using a CSM layer on top of it? I really doubt AMI/AWARD/etc wasted the effort and wrote a full old style BIOS for newer platforms that only obscure companies bother to use. But yea if you get really ancient hardware from China it would still have a BIOS.
        Big thing they don't need to write a new old school bios from complete scratch.
        https://en.wikipedia.org/wiki/SeaBIOS

        You have SeaBIOS so they only need to write the bios drivers for the hardware and some cases not even that because all the required drivers are mainline. Yes there are lot of cases of copyright infringement in some of the boards because they ignore the SeaBIOS LGPL..

        Same reason why they are not paying for AMI/AWARD.. other paid for vendors for a closed source firmware/bios and using old school bios its cost cutting. SeaBIOS in hardware requires a small flashchip(cheaper flashchip) than a UEFI implementation.

        These are in your dirt cheap class of motherboards where the maker in a lot of cases does not put company name, model number or serial number on the board. Yes please note this class of cheap the bios does not always work out the box either so you may be required to fix bios yourself before shipping systems on to customer. The board have bios only I would say are in the too cheap class for most customers. Think party doing a 10000+ unit nas build around x86 hardware cutting corners class boards we are talking here but these are still standard form factor motherboards and will at some point turn up in the second hard market in some amount of volume. This is why you cannot say 2011 and newer is fine because some of the 2011 and newer stuff is these horrible dirt cheap stuff.

        While there is a cost reason and customers are buying these boards they will keep on being made. I have seen a odd make ITX AM4 motherboard without UEFI and only seabios and that takes some serous screwing around and did result in smaller flash required.

        Bios is not dead and gone yet. Its getting close. Hopefully in the next 5 years they stop being made they are currently getting fairly rare hit but they are still being made new at this stage.

        Comment


        • #44
          Originally posted by chithanh View Post
          So? You can use UDF 2.01 if you want interoperability. What does UDF 2.60 give you that exFAT/NTFS has but UDF 2.01 not?
          UDF 2.50 is the big change.
          https://en.wikipedia.org/wiki/Univer...rmat#Revisions
          Added the Metadata Partition facilitating metadata clustering, easier crash recovery and optional duplication of file system information: All metadata like nodes and directory contents are written on a separate partition which can optionally be mirrored.
          This is something fat32, NTFS and ext4 has but exfat does not with backups of the file allocation table/MFT... If everywhere properly support UDF 2.50 things would be good.

          The extra in 2.60 we could live without but could be useful with the Shingled magnetic recording (SMR) harddrives because a lot of their behavour is close to the bluray media behaviour..

          We are coming up to the point of needing a new UDF revision as well now that we are getting drives the max size UDF allows if it taken as this long to get 2.50/2.60 now long until we could get another revision deployed for newer disc sizes.
          Last edited by oiaohm; 10 March 2020, 01:19 AM.

          Comment


          • #45
            Originally posted by Alex Zu View Post
            A statement «it’s OK to integrate an exFAT driver» is totally false as of today. The exFAT modules must be included into the ION definition of Linux Kernel. They’re not yet in and this is yet to happen in future.
            No you need to speak to your legal and get them to explain the following page.
            https://www.openinventionnetwork.com.../linux-system/

            Originally posted by Alex Zu View Post
            Also retrofit of the code to earlier version of Linux is not covered by OIN use.
            This is in fact wrong OIN particular covers doing exactly that.
            [QUOTE]“Predecessor Release” shall mean as to any Linux Environment Component, a previous release of such Linux Environment Component the overall functionality of which is the same as or a subset of such Linux Environment Component, but to which such Linux Environment Component may have added bug fixes, modifications for compatibility, performance enhancements (e.g. increasing execution speed, code maintainability, or bug resistance), or functional enhancements or new functionality.[/QUOTE]

            OIN terms are written that a prior version with backported feature from new version is covered by the Predecessor Release clause yes this clause covers the Linux kernel,

            Originally posted by Alex Zu View Post
            Thus, the code still need to be accepted into the Linux Kernel, the definition of Linux shall be changed by OIN,
            Read the Predecessor and Successor Release clauses closely. The definition of Linux kernel by ION is very super broad in ways that to covers past and future versions. Yes you are right the code has to be accepted by upstream as in mainline to be sure of patent coverage once the code is there it automatically comes part of the OIN define of what the Linux kernel due to the way ION written that define automatically allows forwards and back porting.

            But the new functionality clause in the Predecessor Release does not automatically mean that if the code is not accepted into the Linux Kernel that its not covered by ION it is a grey kind of area so code working way to mainline that party has not raised a patent issue about then attempted todo a patent claim could be on the wrong side of the new functionality clause.

            Originally posted by Alex Zu View Post
            the person/company using the exFAT implementation must be an OIN member..
            This is also wrong. A person/company using items protected by OIN either need to be a OIN member or a customer of a OIN member to be patented covered.

            https://www.openinventionnetwork.com...nse-agreement/
            “Customer”, as to a Person, shall mean an end-user or other customer, direct or indirect, of such Person.
            Customer is another form of super broad define due to the word indirect here. So I build kernel from kernel.org source as is I am customer of Linux foundation by the OIN agreement I give that kernel to another party they are still a customer of Linux foundation this is due to the word indirect there. Yes Linux foundation is still OIN member and everyone in this line is a Linux foundation customer by the OIN agreement.

            Those who are up-streaming all the drivers to mainline on kernel.org are technically a customer of Linux foundation and their customers are also automatically protected as long as their customers are only getting kernels based on the source from kernel.org. This of course leaves a lot of embedded developers out in the cold with there custom kernel patches and binary third party drivers but not all.

            Originally posted by Alex Zu View Post
            Once any of the 3 is false - the use will infringe the patents.
            Not one of the three points you raised is in fact 100% true. The first one is straight up false. Second one ignores the effect of the Predecessor and Successor clauses in the ION define of the Linux platform particularly that new functionality bit in the predecessor so no the code does not need to always be accepted upstream so is way too narrow of a define compared to reality. Finally you don't need to be a OIN member to have OIN protection just have to be a customer and the wording of customer is insanely broad.

            Like Paragon who you worked for came a OIN member all your customers and your customers own customers are covered by that basically this is turtles all the way down. Really if I was Paragon right now I would be talking to Microsoft asking why am I paying for patents right now because I am thinking about joining the OIN group.

            When IBM legal wrote the ION define and ION agreement they were legally creative buggers as it designed to scorch earth profiting from patents against the Linux platform by writing stuff with as broad of coverage as possible. The way ION is written it counter to most patent pool licenses that are written as narrow as possible. Yes there will be a lot parties paying for exfat licenses that are in fact covered by the OIN agreement because they are truly a customer of a OIN member.

            Customer of a OIN member you don't have to share you patent pool this is a loop hole around the OIN requirement to share you patent pool by correctly spitting your company.

            Comment


            • #46
              Originally posted by chithanh View Post
              I have witnessed it myself and you can try it yourself. Install Windows 10 RTM through 1809 on an empty disk, it will create 4 partitions, in Windows Setup they are called (in that order) "Recovery", "System", "MSR (Reserved)", and "Primary". If the "Recovery" partition is 500 MB or less (I think these versions defaulted to 450 MB) then it will be removed during upgrade, and a new 870 MB "Recovery" partition, moved to the tail of the "Primary" partition, will be created.
              'System' is the EFI System partition that all GPT drives need to have. The other 3 are for Windows.

              My RTM fresh install shows the partition as 450M of which 147M is free, my 1910 fresh install shows the partition as 529M of which 96M is free. Upgrading RTM to current (1910) resized the Windows (Primary) partition and added a 534MB partition in the space freed up, but didn't delete the original recovery partition or any others. However, given Microsoft's general lack of QA perhaps an older version of the upgrade process actually deleted partitions at some point, I wouldn't be particularly surprised, but its not actually possible to verify at this point. Never upgrade Windows until several months after the major update drops, they have made this much easier in settings after their continual bungling every new release.

              Comment


              • #47
                Originally posted by [email protected] View Post
                Also, it do not accept to be installed on a partition that is second or third on the drive (in the case you install Linux first), it must be first, or else. And finally, it takes over the boot sector and do not recognize anything that is not Windows.
                you just don't understand anything about Windows. It never erases the Linux partitions as you claimed. It does clear the boot sector on MBR drives and that's the only pain beside the lack of good Linux file system drivers. And it CAN be installed into any drives, any partitions on the system. How can Windows be in the first partition when you can in fact install any numbers of Windows on a single hard drive? In the past it was actually common to have multiple Windows versions for different purposes

                NTLDR on XP is bad and can only be installed onto primary partitions but it still allow Windows to be in any position. BOOTMGR on Vista and up can boot from any drives including logical drives. In fact Windows will typically be needed to be in the second partition in order to leave room for the first one to contain the boot manager, esp. in case of booting from a dynamic disk or encrypted drive

                Comment


                • #48
                  Originally posted by phuclv View Post

                  you just don't understand anything about Windows. It never erases the Linux partitions as you claimed. It does clear the boot sector on MBR drives and that's the only pain beside the lack of good Linux file system drivers. And it CAN be installed into any drives, any partitions on the system. How can Windows be in the first partition when you can in fact install any numbers of Windows on a single hard drive? In the past it was actually common to have multiple Windows versions for different purposes

                  NTLDR on XP is bad and can only be installed onto primary partitions but it still allow Windows to be in any position. BOOTMGR on Vista and up can boot from any drives including logical drives. In fact Windows will typically be needed to be in the second partition in order to leave room for the first one to contain the boot manager, esp. in case of booting from a dynamic disk or encrypted drive
                  Have you considered that what I wrote was based in my experiences with Windows? Have you considered it, or just rushed here to tell I'm wrong? Did you took the time to read AND understand what I wrote, in the context I was referring to? Because to me it didn't appear you did.

                  Comment


                  • #49
                    Originally posted by phuclv View Post
                    you just don't understand anything about Windows. It never erases the Linux partitions as you claimed.
                    https://forum.level1techs.com/t/solv...n-table/110271

                    I am sorry but particular windows updates have delete Linux partition from the partition tables before. Microsoft has made updates bad enough to delete windows own partition from the partition tables. Please note that 2016 update did not just only delete Linux partition table entries but random-ally deleted files out of ntfs partitions.

                    Normally if a windows update is bad enough to be harming a Linux install more than messing up the boot loader you have other worse problems with your windows install to be thinking about.

                    When Microsoft screws up updates they do a really good job of it at times.

                    Normally Microsoft when they screw up normally don't mess you over if you have windows on one harddrive and Linux on another.

                    I would class a Linux install badly broken by a Windows update as a major smoke signal of major problems incoming from what ever windows update triggered it.

                    Comment


                    • #50
                      Originally posted by oiaohm View Post

                      No you need to speak to your legal and get them to explain the following page.
                      https://www.openinventionnetwork.com.../linux-system/
                      We may endlessly exercise in how we all understand OIN and legal situation around exFAT. This will all however be our assumptions and theories, unless they’re taken into a court and judge says what the reality is.

                      The statements Paragon is making on a link that I have provided here (https://www.paragon-software.com/exfat-license/) may be questioned and treated in various ways and from different perspectives.

                      Commercial exFAT deals are usually strictly confidential and cannot be referred to publicly, except for rare exceptions. These deals are the result of careful legal analysis by a lot more professional lawyers that we can find in this forum (at least a lot more professional than I am)

                      I can quote one very recent deal announced publicly by Sagemcom.

                      Without falling into a detailed analysis of your statements I can certainly assert, that big commercial players (dozens of them) do not tend to agree with your vision of the situation. They did choose, they still choose and will be choosing to pay exFAT fee for as long as exFAT patents are valid. This is the reality.
                      Last edited by Alex Zu; 26 March 2020, 10:44 AM.

                      Comment

                      Working...
                      X