Bug#725417: This would not be a simple fix
On Mon, Nov 24, 2014 at 10:39:26AM -0500, Lennart Sorensen wrote: I see in some searches that some places have contacted at +mbr@chiark, while the web page there suggests neilt+www@chiark... It is entirely possible that the chiark page has been forgotten and the email no longer works of course. Well, as another person with a chiark account I can see that Neil is still using his login there and has been active in the last week. Whether or not he's reading email there, I don't know... I'll prod too. -- Steve McIntyre, Cambridge, UK.st...@einval.com We're the technical experts. We were hired so that management could ignore our recommendations and tell us how to do our jobs. -- Mike Andrews -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#725417: This would not be a simple fix
Hi! I think this bug can be fixed, at least on the standard version of mbr (not on the Y2K version, but machines needing that won't run Windows7 or later anyway). The problem is that the changes needed point to an upstream change, that's why I tried to contact Neil, the upstream author, but failing to do so I tried to do it through Santiago Vila, the previous maintainer, but he also seems to have failed. So to change this would change the data structure format in a totally non backwards compatible way (so older install-mbr versions would no longer recognize this as an install-mbr's mbr). That's what I saw and thus I said, this is something to be done by upstream. So moving the data structure is almost certainly doable. Should it be? Is there a maintainer upstream for this code anymore? Well, we haven't been able to contact him at his chiark address and don't know of any other way to contact him. If nothing else, making install-mbr recognize that the existing mbr contains non zero disk signature in bytes 440 to 443 and either or 5A5A in bytes 444 to 445 and spitting out a warning that this disk appears to be in use by a modern MBR using OS, and not just letting you overwrite it might be a good idea. I agree on this as the least we can do to fix this, but I'd rather have an upstream new binary version of the mbr, as doing it at Debian doesn't seem the right thing to do. If finaly agree that we don't have an upstream for mbr we may then do the changes by ourselves, but I'd like to try to contact him one more time (I'll do this right now). Feel free to try to contact him, I suspect that antispam may be dropping the mails, so maybe you are more lucky. If you have it clear, please send patches for this or suggest the new locations, ... Here goes one of my last mails to Neil dated 11th July: Hi! I have tried to contact you from my Debian account, then from my own domain and now from gmail, as I saw your servers have very strict email policies, let's see if this time the mail reaches you :-) I have received a bug on Debian (https://bugs.debian.org/725417) that explains that new MS Windows need what is called a Disk signature on the mbr, and if we use install-mbr to write our mbr this signature gets changed, and then windows stops booting. The bugreport talks about 4 bytes, but it is really 6 bytes that need to be moved away, 6 just before the partition table data. I started looking at this with the idea of patching this on the non Y2K version, as this one was already too tight, however even the normal version seems to be too tight, we had only 5 bytes of padding space on this version. On the other hand, on the 6 bytes that are being used for the Disk signature you had placed the pointer to the data area, so the pointer should be relocated, this is... unless we don't care about those two bytes right where we have the pointer, as the signature is right on the four previous ones and then at 1BC our pointer. This seems to work ok at least with Windows 7. The place were we have our signature should be a h word or 5A5Ah if it is copy protected, at least acording to wikipedia, but Win7 seems to work anyway. Well, what do you think about all this? Regards. -- Manty/BestiaTester - http://manty.net -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#725417: This would not be a simple fix
On Mon, Nov 24, 2014 at 09:58:42AM +0100, Santiago Garcia Mantinan wrote: I think this bug can be fixed, at least on the standard version of mbr (not on the Y2K version, but machines needing that won't run Windows7 or later anyway). The problem is that the changes needed point to an upstream change, that's why I tried to contact Neil, the upstream author, but failing to do so I tried to do it through Santiago Vila, the previous maintainer, but he also seems to have failed. It certainly can be fixed. I am pretty sure I could fix it, but it does come at the expense of breaking the existing install-mbr interface to the mbr data. I too would like to know if there is an upstream and if so what they think of it. That's what I saw and thus I said, this is something to be done by upstream. If they are still around. I think for the quik boot loader (old powerpc loader), Debian essentially is the upstream, although the last code change appears to be when I patched it to work with newer ext filesystems that needed more space 4 years ago. Well, we haven't been able to contact him at his chiark address and don't know of any other way to contact him. I agree on this as the least we can do to fix this, but I'd rather have an upstream new binary version of the mbr, as doing it at Debian doesn't seem the right thing to do. If finaly agree that we don't have an upstream for mbr we may then do the changes by ourselves, but I'd like to try to contact him one more time (I'll do this right now). Feel free to try to contact him, I suspect that antispam may be dropping the mails, so maybe you are more lucky. If you have it clear, please send patches for this or suggest the new locations, ... I have not yet attempted to fix it, but I could do that and submit a suggestion patch. Here goes one of my last mails to Neil dated 11th July: Hi! I have tried to contact you from my Debian account, then from my own domain and now from gmail, as I saw your servers have very strict email policies, let's see if this time the mail reaches you :-) I have received a bug on Debian (https://bugs.debian.org/725417) that explains that new MS Windows need what is called a Disk signature on the mbr, and if we use install-mbr to write our mbr this signature gets changed, and then windows stops booting. The bugreport talks about 4 bytes, but it is really 6 bytes that need to be moved away, 6 just before the partition table data. I started looking at this with the idea of patching this on the non Y2K version, as this one was already too tight, however even the normal version seems to be too tight, we had only 5 bytes of padding space on this version. On the other hand, on the 6 bytes that are being used for the Disk signature you had placed the pointer to the data area, so the pointer should be relocated, this is... unless we don't care about those two bytes right where we have the pointer, as the signature is right on the four previous ones and then at 1BC our pointer. This seems to work ok at least with Windows 7. The place were we have our signature should be a h word or 5A5Ah if it is copy protected, at least acording to wikipedia, but Win7 seems to work anyway. Well, what do you think about all this? Seems reasonable to me. I see in some searches that some places have contacted at +mbr@chiark, while the web page there suggests neilt+www@chiark... It is entirely possible that the chiark page has been forgotten and the email no longer works of course. -- Len Sorensen -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#725417: This would not be a simple fix
install-mbr's code currently uses bytes 442 to 445 to store the version of mbr's data structure (currently 2) and the signature. So to change this would change the data structure format in a totally non backwards compatible way (so older install-mbr versions would no longer recognize this as an install-mbr's mbr). So moving the data structure is almost certainly doable. Should it be? Is there a maintainer upstream for this code anymore? If nothing else, making install-mbr recognize that the existing mbr contains non zero disk signature in bytes 440 to 443 and either or 5A5A in bytes 444 to 445 and spitting out a warning that this disk appears to be in use by a modern MBR using OS, and not just letting you overwrite it might be a good idea. -- Len Sorensen -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
