Bug 254490
| Summary: | REGRESSION: Hybrid disc1.iso fails to boot from USB on BIOS/Legacy ThinkPads and other systems | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | Base System | Reporter: | Michael Dexter <editor> | ||||||
| Component: | misc | Assignee: | freebsd-bugs (Nobody) <bugs> | ||||||
| Status: | Closed FIXED | ||||||||
| Severity: | Affects Many People | CC: | emaste, freebsd, grahamperrin, lcook, manu, marklmi26-fbsd, smithi, tsoome | ||||||
| Priority: | --- | ||||||||
| Version: | 13.0-STABLE | ||||||||
| Hardware: | amd64 | ||||||||
| OS: | Any | ||||||||
| Attachments: |
|
||||||||
|
Description
Michael Dexter
2021-03-22 23:44:57 UTC
(In reply to Michael Dexter from comment #0) This output is from cdboot boot program, which is supposed to run while booting from cd/dvd device and it does not know how to handle hard disk. It means, the isoboot boot program installation into iso image did not happen or is somehow broken. We would need to analyze iso file and its creation process. Thank you Toomas. I have manually confirmed that BETA1, 2, and 3 are working as expected, to help narrow the search. Might there be a cgit-web-thing page that shows history by directory like svnweb did? I can confirm that I can 100% reproduce this issue as well on my Acer Aspire One laptop. 13-BETA4 works exactly as expected for me, while 13-RC1, 13-RC2, and 13-RC3 all fail. This narrows down the commits to about a single week. Michael Dexter can you confirm if your issue was actually in BETA4? And if so, which exact ISO were you using? In my case, I download the ISO (linked below), and then flash it to a USB drive using Rufus on Windows 10. Multiple USB drives have been used, and multiple laptops as well. UEFI consistently works, Legacy BIOS consistently fails. The last ISO that works for me is: https://download.freebsd.org/ftp/releases/amd64/amd64/ISO-IMAGES/13.0/FreeBSD-13.0-BETA4-amd64-disc1.iso The first broken ISO for me is: https://download.freebsd.org/ftp/releases/amd64/amd64/ISO-IMAGES/13.0/FreeBSD-13.0-RC1-amd64-disc1.iso https://twitter.com/DarkainMX/status/1374582571451916290 Created attachment 223543 [details]
BETA4-RC1-oneline.log
Likewise here.
Tested amd64-{disk1,bootonly} on my ThinkPad X230 in Legacy BIOS mode with RC{1,2,3}, and anything prior works fine BETA{1,2,3,4}. I've attached commit logs between BETA4 and RC1 to hopefully bisect the commit causing this new behaviour. For anyone interested, there are 46 commits between BETA4 to RC1, so thankfully nothing *too* extreme.
As for for the commit hashes in releng/13.0:
BETA1... 638e531019fd360ec43d2662c89d12ae632f168f
BETA2... 6461715f1ae19d028fcae45768cfa91f88b31f32
BETA3... 150b4388d3b5654d02071618dcdae48d6f11161f
BETA4... e32bc25362975db9896eeffc2933e078930a5383
RC1... 60e8939aa85b07f3271dcc01730727dda04c8bf6
RC2... 13c22f7495305f5b92874128b088ab47d9512c20
RC3... 8f731a397ad4dc7b17622c0e69ac045f4a7b9d5b
Created attachment 223544 [details]
BETA4-RC1-long.log
Included long-form logs that contain full commit messages and context.
(In reply to Vincent Milum Jr from comment #3) "Michael Dexter can you confirm if your issue was actually in BETA4? And if so, which exact ISO were you using?" I have re-confirmed with BETA3 and BETA4 disc1 images from ftp.freebsd.org, which I trust are in sync. All, I suspect the issue is in one of these two components but cgit is not providing a history by date: https://cgit.freebsd.org/src/tree/release https://cgit.freebsd.org/src/tree/usr.sbin/makefs Please share how to show dates with this as one can do with svnweb.freebsd.org, or re-run your above commit summary with additional information, paths, descriptions, --stat etc. Thank you! (In reply to Michael Dexter from comment #7) You seem to possibly be working at the wrong level for: https://cgit.freebsd.org/src/tree/release https://cgit.freebsd.org/src/tree/usr.sbin/makefs For example going to: https://cgit.freebsd.org/src/tree/ and selecting "log" on the "release" line shows a dated list of the commits in release/ (one line summaries are shown). https://cgit.freebsd.org/src/tree/usr.sbin/ and selecting "log" on the "makefs" line does similarly for makefs/ . But to see what file(s) a commit changed one has to select the commit line's link and look at the commit's details. Thank you Mark, mzar and Allan Jude for pointing out the Log view! This is a new tool to all of us. These emerge as the "usual suspects" for incorrectly-built ISOs but are not guaranteed to be responsible for the regression: https://cgit.freebsd.org/src/log/usr.sbin/makefs https://cgit.freebsd.org/src/log/release If I am not mistaken, this is the date range where the regression was introduced: Update to BETA4 2021-02-26 Update to RC1 2021-03-05 Within that range these commits stand out for me as possible causes but I cannot say for sure without bisecting and building: efibootmgr: Check for efi supported after parsing args 2021-03-01 mkimg: We always want the last block of the last inserted partition 2021-03-02 mkimg: Add support for offset if the source is an image 2021-03-02 Michael I've narrowed it down to this commit. Running a couple more tests to validate it right now "mkimg: We always want the last block of the last inserted partition" https://github.com/freebsd/freebsd-src/commit/f3054d18bfbd4629cb695922f3cd1fab36b1dc5b I now think that both mkimg commits are related. I just tried to do a release build from releng/13.0@c9b4e5e9ae398f3bb8ba85507d9710d8a73a1a04 with the commit I previous mention reverted, but ended up with this build error... Its now 1am, and I'll play with this more in the morning, running a test with releng/13.0@head with both mkimg commits reverted to see if it builds and properly boots under legacy BIOS. -------------- touch disc1 sh /usr/src/release/amd64/mkisoimages.sh -b 13_0_RC3_p1_amd64_CD disc1.iso disc1 Creating `/tmp/efiboot.rQheG5' /tmp/efiboot.rQheG5: 4039 sectors in 4039 FAT12 clusters (512 bytes/cluster) BytesPerSec=512 SecPerClust=1 ResSectors=1 FATs=2 RootDirEnts=512 Sectors=4096 Media=0xf0 FATsecs=12 SecPerTrack=63 Heads=255 HiddenSecs=0 Populating `/tmp/efiboot.rQheG5' Image `/tmp/efiboot.rQheG5' complete mkimg: partition 0 overlaps partition 1 *** Error code 1 Stop. make[1]: stopped in /usr/src/release *** Error code 1 Stop. make: stopped in /usr/src/release Could you test this :
diff --git a/release/amd64/mkisoimages.sh b/release/amd64/mkisoimages.sh
index 1a1440fa198..a9e8a2c0439 100644
--- a/release/amd64/mkisoimages.sh
+++ b/release/amd64/mkisoimages.sh
@@ -89,8 +89,8 @@ if [ "$bootable" != "" ]; then
$MKIMG -s gpt \
--capacity $imgsize \
-b "$BASEBITSDIR/boot/pmbr" \
- $espparam \
-p freebsd-boot:="$BASEBITSDIR/boot/isoboot" \
+ $espparam \
-o hybrid.img
# Drop the PMBR, GPT, and boot code into the System Area of the ISO.
I think I've changed the logic that partition must now be passed in the intented order in case we specify an offset at some point.
Adding -v to mkimg in the script gives me :
Before my change
partition 1: starting block 80 ... size 2097152 bytes (4096 blocks)
location 40960 bytes (80 blocks)
partition 2: starting block 3 ... size 12288 bytes (24 blocks)
After my change
partition 1: starting block 80 ... size 2097152 bytes (4096 blocks)
location 40960 bytes (80 blocks)
partition 2: starting block 4176 ... size 12288 bytes (24 blocks)
I also think that it's logical that the partition are given in the indented order so if that fix it I'll just document this behavior.
With that patch, I'm officially booted into the installer now! :D Successfully booted into 13.0-RC3-p1 from the latest commit in releng/13.0 and this patch applied. A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/src/commit/?id=90d2f7c413f9fc4ac479fa5e91ba1de6d4ea8d45 commit 90d2f7c413f9fc4ac479fa5e91ba1de6d4ea8d45 Author: Emmanuel Vadot <manu@FreeBSD.org> AuthorDate: 2021-03-27 11:04:51 +0000 Commit: Emmanuel Vadot <manu@FreeBSD.org> CommitDate: 2021-03-27 11:04:51 +0000 release: amd64: Fix ISO/USB hybrid image Recent mkimg changes forces to have partitions given in explicit order. This is so we can have the first partition starting at a specific offset and the next ones starting after without having to specify an offset. Switch the partition in the mkisoimage.sh script so the first one created is the isoboot one. PR: 254490 Reported by: Michael Dexter <editor@callfortesting.org Tested by: Vincent Milum Jr <freebsd@darkain.com> MFC after: Right now release/amd64/mkisoimages.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) A commit in branch stable/13 references this bug: URL: https://cgit.FreeBSD.org/src/commit/?id=08639983e0384556a37d19814f55417f604964a1 commit 08639983e0384556a37d19814f55417f604964a1 Author: Emmanuel Vadot <manu@FreeBSD.org> AuthorDate: 2021-03-27 11:04:51 +0000 Commit: Emmanuel Vadot <manu@FreeBSD.org> CommitDate: 2021-03-27 11:09:22 +0000 release: amd64: Fix ISO/USB hybrid image Recent mkimg changes forces to have partitions given in explicit order. This is so we can have the first partition starting at a specific offset and the next ones starting after without having to specify an offset. Switch the partition in the mkisoimage.sh script so the first one created is the isoboot one. PR: 254490 Reported by: Michael Dexter <editor@callfortesting.org Tested by: Vincent Milum Jr <freebsd@darkain.com> MFC after: Right now (cherry picked from commit 90d2f7c413f9fc4ac479fa5e91ba1de6d4ea8d45) release/amd64/mkisoimages.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) A commit in branch releng/13.0 references this bug: URL: https://cgit.FreeBSD.org/src/commit/?id=d6616e3dbaeeb0cf336dda3eec1c5fa7faf8a5c5 commit d6616e3dbaeeb0cf336dda3eec1c5fa7faf8a5c5 Author: Emmanuel Vadot <manu@FreeBSD.org> AuthorDate: 2021-03-27 11:04:51 +0000 Commit: Emmanuel Vadot <manu@FreeBSD.org> CommitDate: 2021-03-28 18:54:28 +0000 release: amd64: Fix ISO/USB hybrid image Recent mkimg changes forces to have partitions given in explicit order. This is so we can have the first partition starting at a specific offset and the next ones starting after without having to specify an offset. Switch the partition in the mkisoimage.sh script so the first one created is the isoboot one. Approved by: re(gjb) PR: 254490 Reported by: Michael Dexter <editor@callfortesting.org Tested by: Vincent Milum Jr <freebsd@darkain.com> MFC after: Right now (cherry picked from commit 90d2f7c413f9fc4ac479fa5e91ba1de6d4ea8d45) (cherry picked from commit 08639983e0384556a37d19814f55417f604964a1) release/amd64/mkisoimages.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) Fixed in every affected branches. Closing now. Thanks for the report and sorry again for the breakage. (In reply to Emmanuel Vadot from comment #18) Not fixed in 12.x unfortunately. 12.3-R was ok but 12.4 suffered the same regression. I reported it to freebsd-questions@ and re@ in this post: Boot fail: 12.4-R-{dvd1,disc1}.iso via USB https://lists.freebsd.org/archives/freebsd-questions/2023-January/002750.html The following two messages are my replies to Xin LI who kindly pointed me to this bug, and Colin Percival as re@, saying he can't imagine a fixed "patch level release" in this case. I've not been able to find it, but I suspect that this fix likely preceeded the first 12.x version to include the ability to write the .iso to an USB stick - 12.1 going by release announcements. |
