[opensuse-kernel] 11.3 kabi issues
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi - I'm preparing the next 11.3 kernel update and working out the details for the kABI. Most of the changes are internal and can be safely ignored. I've worked around a few others that couldn't be, but these last two remain: KABI: symbol simple_fill_super(vmlinux) changed crc from 0x49a93819 to 0x14875b1 9 KABI: symbol usb_buffer_map_sg(vmlinux) lost KABI: symbol usb_buffer_unmap_sg(vmlinux) lost I get the impression that there are no users for the usb functions. Is it safe to ignore those, Greg? The simple_fill_super change I think needs to be reverted. I seems like it was changed for correctness and doesn't actually fix a bug. - -Jeff - -- Jeff Mahoney SUSE Labs -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.15 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org/ iEYEARECAAYFAkxqv+YACgkQLPWxlyuTD7KT6gCfe5GmiVpVTtnaakSf3cbdBoit RjMAn2NBFS1EwKxGQQKCaDT1+69jfjsh =Y0HW -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On Tue, Aug 17, 2010 at 12:59:18PM -0400, Jeff Mahoney wrote:
Hi -
I'm preparing the next 11.3 kernel update and working out the details for the kABI. Most of the changes are internal and can be safely ignored. I've worked around a few others that couldn't be, but these last two remain:
KABI: symbol simple_fill_super(vmlinux) changed crc from 0x49a93819 to 0x14875b1
This does fix a potential bug, but not one that people have seen that I know of.
9 KABI: symbol usb_buffer_map_sg(vmlinux) lost KABI: symbol usb_buffer_unmap_sg(vmlinux) lost
I get the impression that there are no users for the usb functions. Is it safe to ignore those, Greg?
These functions went away? Ah, yeah, they did, there are no users of these functions in the kernel. thanks, greg k-h -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 08/17/2010 01:24 PM, Greg KH wrote:
On Tue, Aug 17, 2010 at 12:59:18PM -0400, Jeff Mahoney wrote:
Hi -
I'm preparing the next 11.3 kernel update and working out the details for the kABI. Most of the changes are internal and can be safely ignored. I've worked around a few others that couldn't be, but these last two remain:
KABI: symbol simple_fill_super(vmlinux) changed crc from 0x49a93819 to 0x14875b1
This does fix a potential bug, but not one that people have seen that I know of.
Changing the version would end up breaking any external virtual file systems, so I'm going to revert that one.
9 KABI: symbol usb_buffer_map_sg(vmlinux) lost KABI: symbol usb_buffer_unmap_sg(vmlinux) lost
I get the impression that there are no users for the usb functions. Is it safe to ignore those, Greg?
These functions went away? Ah, yeah, they did, there are no users of these functions in the kernel.
Since there was only one user in the kernel prior to this, what are the odds that an external driver is using them? - -Jeff - -- Jeff Mahoney SUSE Labs -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.15 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org/ iEYEARECAAYFAkxqx6gACgkQLPWxlyuTD7IjCwCfX59HAEL0hesyaeIibckjgYwX Da4AnAz30QoTwVRTGS+xBspCYgiPAzNK =Hbn6 -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On Tue, Aug 17, 2010 at 01:32:24PM -0400, Jeff Mahoney wrote:
9 KABI: symbol usb_buffer_map_sg(vmlinux) lost KABI: symbol usb_buffer_unmap_sg(vmlinux) lost
I get the impression that there are no users for the usb functions. Is it safe to ignore those, Greg?
These functions went away? Ah, yeah, they did, there are no users of these functions in the kernel.
Since there was only one user in the kernel prior to this, what are the odds that an external driver is using them?
Slim to none, and if one was, odds are they were using it incorrectly so it wasn't working anyway :) Anyhow, I don't know of _any_ external USB kernel modules out there these days, that is not already in the kernel tree. Does anyone else? thanks, greg k-h -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On 17.8.2010 18:59, Jeff Mahoney wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi -
I'm preparing the next 11.3 kernel update and working out the details for the kABI. Most of the changes are internal and can be safely ignored. I've worked around a few others that couldn't be, but these last two remain:
KABI: symbol simple_fill_super(vmlinux) changed crc from 0x49a93819 to 0x14875b1 9 KABI: symbol usb_buffer_map_sg(vmlinux) lost KABI: symbol usb_buffer_unmap_sg(vmlinux) lost
I get the impression that there are no users for the usb functions. Is it safe to ignore those, Greg?
The simple_fill_super change I think needs to be reverted. I seems like it was changed for correctness and doesn't actually fix a bug.
Hi, the current openSUSE-11.3 kernel is not compatible with nvidia and ati KMPs: $ rpm -ivh --test ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64.rpm nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64.rpm nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64.rpm 2>&1 | grep 'kernel(' kernel(default:drivers_pci) = adeb70553a74fa43 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64 kernel(default:arch_x86_kernel_acpi) = 59664e86a426f5a6 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64 kernel(default:drivers_pci) = adeb70553a74fa43 is needed by nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64 kernel(default:drivers_pci) = adeb70553a74fa43 is needed by nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64 But it might be a bug in the scripts and not an actual kabi change, I'm looking into it right now. Michal -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On 18.8.2010 12:03, Michal Marek wrote:
$ rpm -ivh --test ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64.rpm nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64.rpm nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64.rpm 2>&1 | grep 'kernel(' kernel(default:drivers_pci) = adeb70553a74fa43 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64 kernel(default:arch_x86_kernel_acpi) = 59664e86a426f5a6 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64 kernel(default:drivers_pci) = adeb70553a74fa43 is needed by nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64 kernel(default:drivers_pci) = adeb70553a74fa43 is needed by nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64
The drivers_pci change was my fault, but after importing the symset files as well, there is still $ rpm -ivh --test ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64.rpm nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64.rpm nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64.rpm 2>&1 | grep 'kernel(' kernel(default:arch_x86_kernel_acpi) = 59664e86a426f5a6 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64 The arch_x86_kernel_acpi symset changed because of changes in struct acpi_power_register and struct acpi_processor_cx introduced by commit 718be4a (ACPI: skip checking BM_STS if the BIOS doesn't ask for it). As these structures seem to be used only internally by the acpi implementation, so I'd vote for hiding it with the __GENKSYMS__ hack and removing the two functions from the whitelist. Can't wait for 11.4 with per-symbol dependencies and an update stact coping with them... Michal -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On 08/18/2010 02:23 PM, Michal Marek wrote:
On 18.8.2010 12:03, Michal Marek wrote:
$ rpm -ivh --test ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64.rpm nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64.rpm nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64.rpm 2>&1 | grep 'kernel(' kernel(default:drivers_pci) = adeb70553a74fa43 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64 kernel(default:arch_x86_kernel_acpi) = 59664e86a426f5a6 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64 kernel(default:drivers_pci) = adeb70553a74fa43 is needed by nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64 kernel(default:drivers_pci) = adeb70553a74fa43 is needed by nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64
The drivers_pci change was my fault, but after importing the symset files as well, there is still
$ rpm -ivh --test ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64.rpm nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64.rpm nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64.rpm 2>&1 | grep 'kernel(' kernel(default:arch_x86_kernel_acpi) = 59664e86a426f5a6 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64
The arch_x86_kernel_acpi symset changed because of changes in struct acpi_power_register and struct acpi_processor_cx introduced by commit 718be4a (ACPI: skip checking BM_STS if the BIOS doesn't ask for it). As these structures seem to be used only internally by the acpi implementation, so I'd vote for hiding it with the __GENKSYMS__ hack and removing the two functions from the whitelist.
Can't wait for 11.4 with per-symbol dependencies and an update stact coping with them...
Michal
Is this also true, if the drivers are recompiled against the new kernel & kabi ? -- Bruno Friedmann bruno@ioda-net.ch Ioda-Net Sàrl www.ioda-net.ch openSUSE Member User www.ioda.net/r/osu Blog www.ioda.net/r/blog fsfe fellowship www.fsfe.org (bruno.friedmann (at) fsfe.org ) tigerfoot on irc GPG KEY : D5C9B751C4653227 -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On 18.8.2010 15:00, Bruno Friedmann wrote:
On 08/18/2010 02:23 PM, Michal Marek wrote:
$ rpm -ivh --test ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64.rpm nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64.rpm nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64.rpm 2>&1 | grep 'kernel(' kernel(default:arch_x86_kernel_acpi) = 59664e86a426f5a6 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64
Is this also true, if the drivers are recompiled against the new kernel & kabi ?
Please only quote the relevant part of the email you are replying to, especially if you are referring to "this". I guess you are asking if the conflict persists if the KMPs are rebuilt? The modules do not use these two symbols, they just happen to use some other symbols from the same "symset". Rebuilding the KMPs would make the problem go away, but that's something I'd like to avoid (reportedly, it takes some effort to publish those KMPs, so the kernel update would not work for users of nvidia/ati cards for some time). Michal -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
Please only quote the relevant part of the email you are replying to, especially if you are referring to "this". Sorry would never happen again. I guess you are asking if the conflict persists if the KMPs are rebuilt? The modules do not use these two symbols, they just happen to use some other symbols from the same "symset". Rebuilding the KMPs would make the problem go away,
cool. but that's
something I'd like to avoid (reportedly, it takes some effort to publish those KMPs, so the kernel update would not work for users of nvidia/ati cards for some time).
It's already the case with ati, the actual repo offer just off-date driver and they just segfault. Actually each user of 11.3 with fglrx has build the drivers by himself or use rpm made by the ati-installer and they will have to launch the fglrx-kernel-build.sh script on each kernel update. I understand for the nvidia : but it's not too hard to have them rebuild, one time for the distribution. and so the nvidia.ko will be placed in the right lib/module directory Also a quite numbers of users early adopter of 11.3 have build themselves the nvidia drivers.
Michal
It's just my advise ( the poor guys who made ati-rpm for user) who say that. Just do your great job, like the habit you have. -- Bruno -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On Wed, Aug 18, 2010 at 02:23:04PM +0200, Michal Marek wrote:
The drivers_pci change was my fault, but after importing the symset files as well, there is still
$ rpm -ivh --test ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64.rpm nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64.rpm nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64.rpm 2>&1 | grep 'kernel(' kernel(default:arch_x86_kernel_acpi) = 59664e86a426f5a6 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64
The arch_x86_kernel_acpi symset changed because of changes in struct acpi_power_register and struct acpi_processor_cx introduced by commit 718be4a (ACPI: skip checking BM_STS if the BIOS doesn't ask for it). As these structures seem to be used only internally by the acpi implementation, so I'd vote for hiding it with the __GENKSYMS__ hack and removing the two functions from the whitelist.
Using __GENKSYMS__ is fine, want me to do it? But what do you mean by "remove two functions from the whitelist"? thanks, greg k-h -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On 18.8.2010 15:19, Greg KH wrote:
On Wed, Aug 18, 2010 at 02:23:04PM +0200, Michal Marek wrote:
The drivers_pci change was my fault, but after importing the symset files as well, there is still
$ rpm -ivh --test ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64.rpm nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64.rpm nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64.rpm 2>&1 | grep 'kernel(' kernel(default:arch_x86_kernel_acpi) = 59664e86a426f5a6 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64
The arch_x86_kernel_acpi symset changed because of changes in struct acpi_power_register and struct acpi_processor_cx introduced by commit 718be4a (ACPI: skip checking BM_STS if the BIOS doesn't ask for it). As these structures seem to be used only internally by the acpi implementation, so I'd vote for hiding it with the __GENKSYMS__ hack and removing the two functions from the whitelist.
Using __GENKSYMS__ is fine, want me to do it?
But what do you mean by "remove two functions from the whitelist"?
Jeff added acpi_processor_ffh_cstate_* PASS drivers/acpi/processor PASS to kabi/severities, which makes the kabi checker ignore the change. Michal -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 08/18/2010 09:28 AM, Michal Marek wrote:
On 18.8.2010 15:19, Greg KH wrote:
On Wed, Aug 18, 2010 at 02:23:04PM +0200, Michal Marek wrote:
The drivers_pci change was my fault, but after importing the symset files as well, there is still
$ rpm -ivh --test ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64.rpm nvidia-gfxG02-kmp-default-256.35_k2.6.34.0_12-14.1.x86_64.rpm nvidia-gfxG01-kmp-default-173.14.27_k2.6.34.0_12-22.1.x86_64.rpm 2>&1 | grep 'kernel(' kernel(default:arch_x86_kernel_acpi) = 59664e86a426f5a6 is needed by ati-fglrxG02-kmp-default-8.741_k2.6.34.0_12-28.1.x86_64
The arch_x86_kernel_acpi symset changed because of changes in struct acpi_power_register and struct acpi_processor_cx introduced by commit 718be4a (ACPI: skip checking BM_STS if the BIOS doesn't ask for it). As these structures seem to be used only internally by the acpi implementation, so I'd vote for hiding it with the __GENKSYMS__ hack and removing the two functions from the whitelist.
Using __GENKSYMS__ is fine, want me to do it?
But what do you mean by "remove two functions from the whitelist"?
Jeff added acpi_processor_ffh_cstate_* PASS drivers/acpi/processor PASS to kabi/severities, which makes the kabi checker ignore the change.
Ok, I have a patch for this now. - -Jeff - -- Jeff Mahoney SUSE Labs -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.15 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org/ iEYEARECAAYFAkxuwKQACgkQLPWxlyuTD7JlHACfRaUVql5yNR5PJHf38CbqQ8UW 2H0AnAi341pPU7dVu7+gNd6CXcMdqseQ =YqrK -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
participants (4)
-
Bruno Friedmann -
Greg KH -
Jeff Mahoney -
Michal Marek