Re: [oi-dev] Obstacle to GPT label boot?
In fact the text installer ISO *will* install 2020.10 to an existing s0 slice in an EFI label. Because of the white on pale turquoise text I did not see the "Select existing pool" option which is located where F3_back is typically. Of course, the text install disk doesn't provide X. So some additional work required. A "pkg install lightdm" has been spinning for some time because of the delays imposed by flashing the entire line of progress text. I'm hoping that will drag in X, but we shall see. The frustration and general misery that the "artistic" choice of white on pale turquoise text has caused me is incredible. Days of time. Long very miserable days. I've had 2 retinal detachments, 3 cataract surgeries and a YAG capsulectomy. To add to the fun I have open angle glaucoma and wet and dry macular degeneration. I'm in a horse race between going blind and dying of cancer and very much hoping dying comes first. I have paid out of pocket in excess of $40,000 to be able to see at all. I consider that the best money I ever spent because if I had been born much earlier I'd have been totally blind at 55 or 56. I routinely tell ophthalmologists, "If you can guarantee I can see, you can have *all* my money." Unfortunately, no one can do that. The GPT label boot works as Toomas has asserted. At present I don't see any need for work on that. I shall clone the OI repository and take a look at the installers. Reg ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] Obstacle to GPT label boot?
OK, now we are getting somewhere. I don't have an EFI BIOS so far as I know. In fact, the erroneous disk sizes reported in the SETUP menus confirm it does not read EFI labels. However, after some fiddle I've located a more recent BIOS for Linux which I shall try out. Perhaps it does handle EFI booting now. I can put an EFI label on a disk with format(1m), but that does not allow the 2020.10 installers (any of the 3) to let me select the 100 GB s0 slices I create for the root pool. It doesn't even appear to see the EFI label. I can easily use >2 TB disks on u8; I just can't boot from them. I recently put a zfs pool on a 4 and a 12 TB USB drive for data transfers and backups using u8. So the EFI labels it creates are just fine. In my case OI *is* my daily driver unless I'm doing serious programming and data analysis in which case I use my Sol 10 u8 system. Debian and Windows are just for binary only software and live on a dual boot machine. As EFI has an MBR it should be possible to use that sector to load the EFI label and boot code on an old system in such a fashion that the boot process is agnostic about the BIOS and will work with either. I should note that Debian doesn't have an issue using the current BIOS and a >2 TB disk. So a BIOS agnostic loader is not technically a problem. The obvious first step is to clone the repositories for Illumos, OI and FreeBSD on the 4 TB USB disk so I can move them around. A search tells me there are 33 Illumos repositories on github. Should I use this one? https://github.com/illumos For OI I located this as the obvious choice: https://openindiana.github.io/ And for FreeBSD, this: https://cgit.freebsd.org/ For Linux this appears to be what I should use: https://github.com/torvalds/linux I'd appreciate confirmation that those are the correct, canonical repositories for each project. I presume that there is a "boot" directory somewhere in each tree, but if it has a different name please let me know. Have Fun! Reg ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] Obstacle to GPT label boot?
> On 19. Mar 2021, at 09:53, Marcel Telka wrote: > > On Fri, Mar 19, 2021 at 09:34:01AM +0200, Toomas Soome wrote: > > [...snip...] > > Thanks for all those commands and references! > > >> To install on slice, you would need to select GPT partitioning and >> *not* select whole disk setup. However, I have not used OI installer >> for a long time, therefore I can not recall what exactly are on those >> screens:D Also it is very likely, some improvements can be made there >> (like for >2TB disks we should only offer GPT). > > IIRC, this is not possible with Hipster 2020.10 installer. hm. then it needs to be fixed. > >> Yes, this BIOS does see all 4 disks, therefore I can boot from this >> rpool. We have been able to do this since loader was integrated.. hm, >> log is telling… > > Then it must be just my stupid BIOS that does not allow me to boot from > NVMe when I try legacy BIOS boot :-(. > this is often the case, same with vmware fusion VM; also, if the device is having 4kn (4k logical/physical) sector size, then it is quite probable, the BIOS will not allow booting from it, and our pmbr.s code does not support 4kn either - without actual test system, I see no point:) from practical point of view, NVMe boot is possible with UEFI. rgds, toomas ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] Obstacle to GPT label boot?
On Fri, Mar 19, 2021 at 09:34:01AM +0200, Toomas Soome wrote: [...snip...] Thanks for all those commands and references! > To install on slice, you would need to select GPT partitioning and > *not* select whole disk setup. However, I have not used OI installer > for a long time, therefore I can not recall what exactly are on those > screens:D Also it is very likely, some improvements can be made there > (like for >2TB disks we should only offer GPT). IIRC, this is not possible with Hipster 2020.10 installer. > Yes, this BIOS does see all 4 disks, therefore I can boot from this > rpool. We have been able to do this since loader was integrated.. hm, > log is telling… Then it must be just my stupid BIOS that does not allow me to boot from NVMe when I try legacy BIOS boot :-(. Thanks. -- +---+ | Marcel Telka e-mail: mar...@telka.sk | |homepage: http://telka.sk/ | +---+ ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] Obstacle to GPT label boot?
> On 19. Mar 2021, at 09:04, Marcel Telka wrote: > > On Fri, Mar 19, 2021 at 08:17:44AM +0200, Toomas Soome via oi-dev wrote: >> root@beastie:/code/illumos-gate# zpool status rpool >> pool: rpool >> state: ONLINE >> scan: resilvered 1,68T in 0 days 10:10:07 with 0 errors on Fri Oct 25 >> 05:05:34 2019 >> config: >> >>NAMESTATE READ WRITE CKSUM >>rpool ONLINE 0 0 0 >> raidz1-0 ONLINE 0 0 0 >>c3t0d0 ONLINE 0 0 0 >>c3t1d0 ONLINE 0 0 0 >>c3t3d0 ONLINE 0 0 0 >>c3t4d0 ONLINE 0 0 0 >> >> errors: No known data errors > > What is the slice layout at those disks? Is it one small EFI boot slice > (pcfs) + one big slice covering the rest of the disk and used for rpool? > If I'm not mistaken this is the default layout created by the OI > installer when installed on GPT/EFI disk (without MBR). > > Is there a way to install OI on a GPT disk in a slice not covering the > whole disk? > Part TagFlag First Sector Size Last Sector 0 systemwm 256 256.00MB 524543 1usrwm524544 3.64TB 7814020750 2 unassignedwm 000 3 unassignedwm 000 4 unassignedwm 000 5 unassignedwm 000 6 unassignedwm 000 8 reservedwm7814020751 8.00MB 7814037134 format> This is the default layout generated with zpool create -B rpool raidz c3t0d0 c3t1d0 c3t3d0 c3t4d0 Technically *every* GPT partitioned disk does have MBR, but it is called protective MBR (PMBR). This can be checked with fdisk, but here is alternate way: root@beastie:/home/tsoome# mdb /dev/rdsk/c3t0d0 > ::load disk_label > ::mbr Format: loader (illumos) Signature: 0xaa55 (valid) UniqueMBRDiskSignature: 0 Loader STAGE1_STAGE2_LBA: 524544 Loader STAGE1_STAGE2_SIZE: 1 Loader STAGE1_STAGE2_UUID: 7f0a04f2-6a11-e6b8-b0ae-dc14a75a96fc STAGE1 in VBR: Format: loader (illumos) Signature: 0xaa55 (valid) UniqueMBRDiskSignature: 0 Loader STAGE1_STAGE2_LBA: 525568 Loader STAGE1_STAGE2_SIZE: 328 Loader STAGE1_STAGE2_UUID: 7f0a04f2-6a11-e6b8-b0ae-dc14a75a96fc PART TYPE ACTIVE STARTCHSENDCHS SECTOR NUMSECT 0EFI_PMBR:0xee 0 1023/255/63 1023/255/63 1 4294967295 1UNUSED:0 2UNUSED:0 3UNUSED:0 > ::gpt Signature: EFI PART (valid) Revision: 1.0 HeaderSize: 512 bytes HeaderCRC32: 0x375ff497 (should be 0x375ff497) Reserved1: 0 (should be 0x0) MyLBA: 1 (should be 1) AlternateLBA: 7814037167 FirstUsableLBA: 34 LastUsableLBA: 7814037134 DiskGUID: c350e9d4-baee-63db-f81f-f702e840eb5f PartitionEntryLBA: 2 NumberOfPartitionEntries: 9 SizeOfPartitionEntry: 0x80 bytes PartitionEntryArrayCRC32: 0xf2fd883d (should be 0xf2fd883d) PART TYPESTARTLBA ENDLBAATTR NAME 0EFI_SYSTEM 256 5245430loader 1EFI_USR 52454478140207500zfs 2EFI_UNUSED 3EFI_UNUSED 4EFI_UNUSED 5EFI_UNUSED 6EFI_UNUSED 7EFI_UNUSED 8EFI_RESERVED781402075178140371340 > ::quit root@beastie:/home/tsoome# To install on slice, you would need to select GPT partitioning and *not* select whole disk setup. However, I have not used OI installer for a long time, therefore I can not recall what exactly are on those screens:D Also it is very likely, some improvements can be made there (like for >2TB disks we should only offer GPT). >> I can tell, this system does boot just fine with both UEFI and BIOS firmware. > > Legacy BIOS? Is your system really able to boot in legacy BIOS mode > with the rpool as outlined above? > Yes, this BIOS does see all 4 disks, therefore I can boot from this rpool. We have been able to do this since loader was integrated.. hm, log is telling… Author: Toomas Soome Date: Sun Oct 25 00:06:51 2015 +0300 5061 freebsd boot loader integration (loader project) Reviewed by: Richard Lowe Reviewed by: Cody Mello Approved by: Robert Mustacchi 5 years ago.. time flies... rgds, toomas ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] Obstacle to GPT label boot?
On Fri, Mar 19, 2021 at 08:17:44AM +0200, Toomas Soome via oi-dev wrote: > root@beastie:/code/illumos-gate# zpool status rpool > pool: rpool > state: ONLINE > scan: resilvered 1,68T in 0 days 10:10:07 with 0 errors on Fri Oct 25 > 05:05:34 2019 > config: > > NAMESTATE READ WRITE CKSUM > rpool ONLINE 0 0 0 > raidz1-0 ONLINE 0 0 0 > c3t0d0 ONLINE 0 0 0 > c3t1d0 ONLINE 0 0 0 > c3t3d0 ONLINE 0 0 0 > c3t4d0 ONLINE 0 0 0 > > errors: No known data errors What is the slice layout at those disks? Is it one small EFI boot slice (pcfs) + one big slice covering the rest of the disk and used for rpool? If I'm not mistaken this is the default layout created by the OI installer when installed on GPT/EFI disk (without MBR). Is there a way to install OI on a GPT disk in a slice not covering the whole disk? > I can tell, this system does boot just fine with both UEFI and BIOS firmware. Legacy BIOS? Is your system really able to boot in legacy BIOS mode with the rpool as outlined above? Thanks. -- +---+ | Marcel Telka e-mail: mar...@telka.sk | |homepage: http://telka.sk/ | +---+ ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] Obstacle to GPT label boot?
> On 19. Mar 2021, at 02:31, Reginald Beardsley via oi-dev > wrote: > > Would someone please direct me to an explanation of why we can't boot from > disk >2 TB in 2021? root@beastie:/code/illumos-gate# format Searching for disks...done AVAILABLE DISK SELECTIONS: 0. c3t0d0 /pci@0,0/pci15d9,805@1f,2/disk@0,0 1. c3t1d0 /pci@0,0/pci15d9,805@1f,2/disk@1,0 2. c3t3d0 /pci@0,0/pci15d9,805@1f,2/disk@3,0 3. c3t4d0 /pci@0,0/pci15d9,805@1f,2/disk@4,0 Specify disk (enter its number): ^D root@beastie:/code/illumos-gate# zpool status rpool pool: rpool state: ONLINE scan: resilvered 1,68T in 0 days 10:10:07 with 0 errors on Fri Oct 25 05:05:34 2019 config: NAMESTATE READ WRITE CKSUM rpool ONLINE 0 0 0 raidz1-0 ONLINE 0 0 0 c3t0d0 ONLINE 0 0 0 c3t1d0 ONLINE 0 0 0 c3t3d0 ONLINE 0 0 0 c3t4d0 ONLINE 0 0 0 errors: No known data errors root@beastie:/code/illumos-gate# I can tell, this system does boot just fine with both UEFI and BIOS firmware. rgds, toomas > > It's very hard for me to see a significant obstacle to reading an x86 MBR > that loads code that will then boot from GPT label disks. > > I'm rebuilding my Sol 10 u8 system with my spare 2 TB disk. I'd like to > replace the 3x 2 TB disk RAIDZ1 setup with a 4-5x disk RAIDZ2 using 4 TB > disks using my s0 & s1 configuration. > > I must assume that the Illumos market is dominated by customers who don't > care about having a few small SMI labeled boot disks in a farm of large GPT > labeled disks. But for a 7 SATA port system, that's not very viable. > > Linux, MS and BSD can do it, so we should be able to do it also. It's become > enough of a personal nuisance that I'm willing to fix it if I get a modest > level of cooperation. > > So please, point me at any known issues. For workstations the s0 root pool > and s1 export pool works really well. > > Reg > > BTW After a DIMM shuffle, "format -e" and scrubs no longer dump core on u8. > Now X goes off into la-la land :-( Clearly I need new DIMMS. > ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
[oi-dev] Obstacle to GPT label boot?
Would someone please direct me to an explanation of why we can't boot from disk >2 TB in 2021? It's very hard for me to see a significant obstacle to reading an x86 MBR that loads code that will then boot from GPT label disks. I'm rebuilding my Sol 10 u8 system with my spare 2 TB disk. I'd like to replace the 3x 2 TB disk RAIDZ1 setup with a 4-5x disk RAIDZ2 using 4 TB disks using my s0 & s1 configuration. I must assume that the Illumos market is dominated by customers who don't care about having a few small SMI labeled boot disks in a farm of large GPT labeled disks. But for a 7 SATA port system, that's not very viable. Linux, MS and BSD can do it, so we should be able to do it also. It's become enough of a personal nuisance that I'm willing to fix it if I get a modest level of cooperation. So please, point me at any known issues. For workstations the s0 root pool and s1 export pool works really well. Reg BTW After a DIMM shuffle, "format -e" and scrubs no longer dump core on u8. Now X goes off into la-la land :-( Clearly I need new DIMMS. ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
