On 12/04/2013 02:58 PM, Gene Cumm wrote:> On Wed, Dec 4, 2013 at 12:02 PM, Adam Sawyer <acsawyer at gmail.com> wrote: >> All, >> >> I have scoured the mailing list and the web and I can't seem to find any >> help on this issue. I am attempting to have EFI machines boot over PXE and >> then load various Linux Live distributions. I have a similar setup over >> standard BIOS machines and all is well on that end. First I have tried >> SysLinux 6.0, 6.1, and 6.2 for various EFI enabled machines and thus far >> 6.1 gets me to the furthest point. >> >> My clients successfully boot over PXE and grab the relevant files over TFTP >> to get me to a simple selection menu. I have watched the interaction in the >> TFTP logs and over tcpdump. All is well until I attempt to select a kernel >> (Ubuntu 13.10 vmlinuz.efi). Regardless of where the kernel is located, >> nested in a sub-directory of /tftpboot or not, the client always responds >> with "File not found." It is odd since all other support files >> (syslinux64.efi, pxelinux.cfg/default, vesamenu.c32, and so forth) have no >> issue. > > What exact kernel? This almost sounds familiar to some errors I've > seen in the last two months. What do you observe in the packet > capture? >I believe this is exactly the TFTP port number bug. -hpa
On Wed, Dec 4, 2013 at 6:18 PM, H. Peter Anvin <hpa at zytor.com> wrote:> On 12/04/2013 02:58 PM, Gene Cumm wrote: >> On Wed, Dec 4, 2013 at 12:02 PM, Adam Sawyer <acsawyer at gmail.com> wrote: >>> All, >>> >>> I have scoured the mailing list and the web and I can't seem to find any >>> help on this issue. I am attempting to have EFI machines boot over PXE and >>> then load various Linux Live distributions. I have a similar setup over >>> standard BIOS machines and all is well on that end. First I have tried >>> SysLinux 6.0, 6.1, and 6.2 for various EFI enabled machines and thus far >>> 6.1 gets me to the furthest point. >>> >>> My clients successfully boot over PXE and grab the relevant files over TFTP >>> to get me to a simple selection menu. I have watched the interaction in the >>> TFTP logs and over tcpdump. All is well until I attempt to select a kernel >>> (Ubuntu 13.10 vmlinuz.efi). Regardless of where the kernel is located, >>> nested in a sub-directory of /tftpboot or not, the client always responds >>> with "File not found." It is odd since all other support files >>> (syslinux64.efi, pxelinux.cfg/default, vesamenu.c32, and so forth) have no >>> issue. >> >> What exact kernel? This almost sounds familiar to some errors I've >> seen in the last two months. What do you observe in the packet >> capture? >> > > I believe this is exactly the TFTP port number bug. > > -hpa >http://bugzilla.syslinux.org/show_bug.cgi?id=35 If so, I think this is resolved but needing more testing. -- -Gene
2013/12/5, H. Peter Anvin <hpa at zytor.com>:> On 12/04/2013 02:58 PM, Gene Cumm wrote: >> On Wed, Dec 4, 2013 at 12:02 PM, Adam Sawyer <acsawyer at gmail.com> wrote: >>> All, >>> >>> I have scoured the mailing list and the web and I can't seem to find any >>> help on this issue. I am attempting to have EFI machines boot over PXE >>> and >>> then load various Linux Live distributions. I have a similar setup over >>> standard BIOS machines and all is well on that end. First I have tried >>> SysLinux 6.0, 6.1, and 6.2 for various EFI enabled machines and thus far >>> 6.1 gets me to the furthest point. >>> >>> My clients successfully boot over PXE and grab the relevant files over >>> TFTP >>> to get me to a simple selection menu. I have watched the interaction in >>> the >>> TFTP logs and over tcpdump. All is well until I attempt to select a >>> kernel >>> (Ubuntu 13.10 vmlinuz.efi). Regardless of where the kernel is located, >>> nested in a sub-directory of /tftpboot or not, the client always >>> responds >>> with "File not found." It is odd since all other support files >>> (syslinux64.efi, pxelinux.cfg/default, vesamenu.c32, and so forth) have >>> no >>> issue. >> >> What exact kernel? This almost sounds familiar to some errors I've >> seen in the last two months. What do you observe in the packet >> capture? >> > > I believe this is exactly the TFTP port number bug.Actually, this is exactly bug #26 <http://bugzilla.syslinux.org/show_bug.cgi?id=26> the "inode size" bug. The port number bug has been introduced between 6.02-pre16 and 6.02 and is only a layer of bug in front of the "inode size" bug. It would be nice if my patches for bug #26 (inode size) would be merged into syslinux. And I think Gene Cumm have a better fix for bug #35 (port number). And why not catch up on the patches and pull requests sent to the mailing list? :) -- Celelibi
It looks like it is more like the issue in bug #26. In the tcpdump I can see the client attempt to RRQ the kernel but the client itself reports "No file found." I am having similiar issues as reported in both bug reports with getting Syslinux 6.02 to work as in syslinux64.efi gets pulled but then the client gets stuck on ldlinux.e64. Can I use http/nfs instead after syslinux64.efi gets pulled? Seems to be a TFTP issue. Thanks for your help. Adam On Wed, Dec 4, 2013 at 7:09 PM, Celelibi <celelibi at gmail.com> wrote:> 2013/12/5, H. Peter Anvin <hpa at zytor.com>: > > On 12/04/2013 02:58 PM, Gene Cumm wrote: > >> On Wed, Dec 4, 2013 at 12:02 PM, Adam Sawyer <acsawyer at gmail.com> > wrote: > >>> All, > >>> > >>> I have scoured the mailing list and the web and I can't seem to find > any > >>> help on this issue. I am attempting to have EFI machines boot over PXE > >>> and > >>> then load various Linux Live distributions. I have a similar setup over > >>> standard BIOS machines and all is well on that end. First I have tried > >>> SysLinux 6.0, 6.1, and 6.2 for various EFI enabled machines and thus > far > >>> 6.1 gets me to the furthest point. > >>> > >>> My clients successfully boot over PXE and grab the relevant files over > >>> TFTP > >>> to get me to a simple selection menu. I have watched the interaction in > >>> the > >>> TFTP logs and over tcpdump. All is well until I attempt to select a > >>> kernel > >>> (Ubuntu 13.10 vmlinuz.efi). Regardless of where the kernel is located, > >>> nested in a sub-directory of /tftpboot or not, the client always > >>> responds > >>> with "File not found." It is odd since all other support files > >>> (syslinux64.efi, pxelinux.cfg/default, vesamenu.c32, and so forth) have > >>> no > >>> issue. > >> > >> What exact kernel? This almost sounds familiar to some errors I've > >> seen in the last two months. What do you observe in the packet > >> capture? > >> > > > > I believe this is exactly the TFTP port number bug. > > > Actually, this is exactly bug #26 > <http://bugzilla.syslinux.org/show_bug.cgi?id=26> the "inode size" > bug. > The port number bug has been introduced between 6.02-pre16 and 6.02 > and is only a layer of bug in front of the "inode size" bug. > > It would be nice if my patches for bug #26 (inode size) would be > merged into syslinux. And I think Gene Cumm have a better fix for bug > #35 (port number). > > And why not catch up on the patches and pull requests sent to the > mailing list? :) > > -- > Celelibi >-- Adam "Arock" Sawyer 301-204-7421 acsawyer at gmail.com