Hello, I think that reading from CDROM is actually broken in 9.0-RELEASE/sparc64. On FreeBSD 8.2-RELEASE/sparc64, mounting a physical CDROM burned with FreeBSD-9.0-RELEASE-sparc64-disc1.iso and running sha256 on usr/freebsd-dist/*.txz yields the same results as listed in usr/freebsd-dist/MANIFEST. However, on FreeBSD 9.0-RELEASE/sparc64, mounting the same CDROM (using /dev/cd0), and running sha256 on the same usr/freebsd-dist/*.txz files yields completely different results: SHA256 (base.txz) a96e255a9379ec1a58b756c3e788e34345fd79e35234925df842d700ccb10e1a SHA256 (doc.txz) 54674bb68b111c3465b0beca8893a6b514b80e20dc22f2bbd208b992568261dc SHA256 (games.txz) b9bedddeea549640b4f41a577f4c5f09aa29d0c3d0d4c105bc4fe7d327bf9fcd SHA256 (kernel.txz) 661df36b4a2e87d2aa88ce998b50d48f2a91892fc288452a0282e3c092739046 SHA256 (ports.txz) 315cbfac6f252de5209a3513a846d14650b98e2dcf99c12779ffd1d32acaa287 SHA256 (src.txz) 6b3a4a9536b1ca2618949bb7bb44478266db551ac9f09ccfc92b5469a715f3cf This also prevents an installation of 9.0-RELEASE/sparc64 via bsdinstall from CDROM (bsdinistall says that the data sets are corrupted). I was only able to get 9.0 on that SunBlade 1500 by using a regular source upgrade from a running 8.2-RELEASE system (using usr/freebsd-dist/src.txz from that very same CDROM that checksummed correctly on 8.2-RELEASE). Just to make it clear: the CDROM media is correct and checksums correctly on 8.2-RELEASE. It doesn't checksum correctly on 9.0-RELEASE because reading from CDROM media is now (partially) broken. Just to make sure the 9.0-RELEASE sha256 program isn't at fault, I've mounted FreeBSD-9.0-RELEASE-sparc64-disc1.iso via mdconfig and run sha256 on the *.txz files again: same results as in MANIFEST. So sha256 on 9.0 itself is okay, it's only the reading from physical media that corrupts data (and prevents installation from CDROM -- funny that this passed release engineering and testing...). Thanks, -cpghost. -- Cordula's Web. http://www.cordula.ws/
On Mon, Jan 16, 2012 at 2:54 PM, C. P. Ghost <cpghost@cordula.ws> wrote:> Hello, > > I think that reading from CDROM is actually broken > in 9.0-RELEASE/sparc64. > > On FreeBSD 8.2-RELEASE/sparc64, mounting a physical > CDROM burned with FreeBSD-9.0-RELEASE-sparc64-disc1.iso > and running sha256 on usr/freebsd-dist/*.txz yields the same > results as listed in usr/freebsd-dist/MANIFEST. > > However, on FreeBSD 9.0-RELEASE/sparc64, mounting the > same CDROM (using /dev/cd0), and running sha256 on the > same usr/freebsd-dist/*.txz files yields completely different results: > > SHA256 (base.txz) > a96e255a9379ec1a58b756c3e788e34345fd79e35234925df842d700ccb10e1a > SHA256 (doc.txz) > 54674bb68b111c3465b0beca8893a6b514b80e20dc22f2bbd208b992568261dc > SHA256 (games.txz) > b9bedddeea549640b4f41a577f4c5f09aa29d0c3d0d4c105bc4fe7d327bf9fcd > SHA256 (kernel.txz) > 661df36b4a2e87d2aa88ce998b50d48f2a91892fc288452a0282e3c092739046 > SHA256 (ports.txz) > 315cbfac6f252de5209a3513a846d14650b98e2dcf99c12779ffd1d32acaa287 > SHA256 (src.txz) > 6b3a4a9536b1ca2618949bb7bb44478266db551ac9f09ccfc92b5469a715f3cf > > This also prevents an installation of 9.0-RELEASE/sparc64 > via bsdinstall from CDROM (bsdinistall says that the data sets > are corrupted). I was only able to get 9.0 on that SunBlade 1500 > by using a regular source upgrade from a running 8.2-RELEASE > system (using usr/freebsd-dist/src.txz from that very same CDROM > that checksummed correctly on 8.2-RELEASE). > > Just to make it clear: the CDROM media is correct and checksums > correctly on 8.2-RELEASE. It doesn't checksum correctly on 9.0-RELEASE > because reading from CDROM media is now (partially) broken. > > Just to make sure the 9.0-RELEASE sha256 program isn't at fault, > I've mounted FreeBSD-9.0-RELEASE-sparc64-disc1.iso via mdconfig > and run sha256 on the *.txz files again: same results as in MANIFEST. > So sha256 on 9.0 itself is okay, it's only the reading from physical > media that corrupts data (and prevents installation from CDROM -- funny > that this passed release engineering and testing...).One more thing: With a 9.0-RELEASE world installed, if I boot the old 8.2-RELEASE kernel (/boot/kernel.old/kernel), mount /dev/acd0 and checksum the files with sha256, everything is okay. If I boot again the new 9.0-RELEASE kernel (/boot/kernel/kernel), mount /dev/cd0 and checksum the files with sha256, I get the wrong results above. Since there's no /dev/acd0 on 9.0-RELEASE, I can't test mounting that on 9 and compare the results with /dev/cd0. But obviously, whatever /dev/cd0 returns to cd9660 seems to be silently corrupted on 9.0; at least on sparc64. -cpghost. -- Cordula's Web. http://www.cordula.ws/
On Mon, Jan 16, 2012 at 2:54 PM, C. P. Ghost <cpghost@cordula.ws> wrote:> Hello, > > I think that reading from CDROM is actually broken > in 9.0-RELEASE/sparc64. > > On FreeBSD 8.2-RELEASE/sparc64, mounting a physical > CDROM burned with FreeBSD-9.0-RELEASE-sparc64-disc1.iso > and running sha256 on usr/freebsd-dist/*.txz yields the same > results as listed in usr/freebsd-dist/MANIFEST. > > However, on FreeBSD 9.0-RELEASE/sparc64, mounting the > same CDROM (using /dev/cd0), and running sha256 on the > same usr/freebsd-dist/*.txz files yields completely different results: > > SHA256 (base.txz) > a96e255a9379ec1a58b756c3e788e34345fd79e35234925df842d700ccb10e1a > SHA256 (doc.txz) > 54674bb68b111c3465b0beca8893a6b514b80e20dc22f2bbd208b992568261dc > SHA256 (games.txz) > b9bedddeea549640b4f41a577f4c5f09aa29d0c3d0d4c105bc4fe7d327bf9fcd > SHA256 (kernel.txz) > 661df36b4a2e87d2aa88ce998b50d48f2a91892fc288452a0282e3c092739046 > SHA256 (ports.txz) > 315cbfac6f252de5209a3513a846d14650b98e2dcf99c12779ffd1d32acaa287 > SHA256 (src.txz) > 6b3a4a9536b1ca2618949bb7bb44478266db551ac9f09ccfc92b5469a715f3cf > > This also prevents an installation of 9.0-RELEASE/sparc64 > via bsdinstall from CDROM (bsdinistall says that the data sets > are corrupted). I was only able to get 9.0 on that SunBlade 1500 > by using a regular source upgrade from a running 8.2-RELEASE > system (using usr/freebsd-dist/src.txz from that very same CDROM > that checksummed correctly on 8.2-RELEASE). > > Just to make it clear: the CDROM media is correct and checksums > correctly on 8.2-RELEASE. It doesn't checksum correctly on 9.0-RELEASE > because reading from CDROM media is now (partially) broken. > > Just to make sure the 9.0-RELEASE sha256 program isn't at fault, > I've mounted FreeBSD-9.0-RELEASE-sparc64-disc1.iso via mdconfig > and run sha256 on the *.txz files again: same results as in MANIFEST. > So sha256 on 9.0 itself is okay, it's only the reading from physical > media that corrupts data (and prevents installation from CDROM -- funny > that this passed release engineering and testing...).Submitted as sparc64/164226 Thanks, -cpghost. -- Cordula's Web. http://www.cordula.ws/