Repository navigation
Rebase edera/mainline onto master @ 4405c18c95d2 - #120
Closed
edera-rebase-app[bot] wants to merge 10000 commits into
Closed
edera-rebase-app[bot] wants to merge 10000 commits into
edera-rebase-app[bot] wants to merge 10000 commits into
Conversation
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Nightly rebase of
edera/mainline(2ece1af06645) ontomaster(4405c18c95d2), driven by Claude. Not pushed automatically because the checker found drift and a build failure and Claude's report asks for a decision. Run log.Important
Do not use the merge button: this branch replaces
edera/mainlinerather than merging into it. Approve the pull request to land it; the land workflow force-pushesedera/mainlineto the approved head, pinned to2ece1af06645, and keeps a backup.Independent checks
Builds: failure (x86_64 vmlinux on the Edera host config and arm64 vmlinux, both with the Edera features built in; per-build results in the run).
The build failed and Claude's repair round (success) left no new branch; this is the unrepaired rebase.
2ece1af06645d58f6bea5d31864c07de0c7d4405c18c95d2Commits whose hunk context moved (own +/- lines identical)
ea62ac9f1b0f iommu/xen: Add Xen PV-IOMMU driver→88e1981610c2d8221bf28f22 x86/CPU/AMD: avoid printing reset reasons on Xen domU→4d13afdf368cbd1c0dba57c3 sysctl: wire up PaX softmode setting→603bab1a0fce849ed61740ea use maximum supported mmap rnd entropy by default→d188c9b8fd785b1ff5bbc8cb 9p/xen: carry the 9P session across a restore→37e9d8605e2f85e1da02b838 9p/xen: end foreign access on the grant refs we granted, not the intf page's→4e81a994dcccCommits lost, added, or whose own +/- lines changed
f492e4d0cb26 9p/xen: reconnect to the backend on resume9e1163209bc2 9p/xen: reconnect to the backend on resume3f7ca89e3d70 xen/gntdev: don't copy to user space under priv->lock (#79)2ab25ce3a8e5 xen/gntdev: don't copy to user space under priv->lock (#79)Drift (changed lines: upstream delta vs. actual change)
Claude's report
Summary
Rebased
edera/mainline(old tip2ece1af06645) from merge base864c07de0c7d(v7.2-rc5-26646-g864c07de0c7d) onto upstream4405c18c95d2(v7.2-rc5-43498-g4405c18c95d2): 15872 upstream commits. All 130 downstream commits replayed. The new tip ofrebase-resultisd58f6bea5d31. Five commits conflicted and were resolved by hand. No commit was dropped, and no build fix or fix-up commit was needed. x86_64 and arm64vmlinuxboth build.verify-upstream-rebase.shpasses the base and linearity checks. It flags 2 commits whose own +/- lines changed and some tree drift. Both come only from the conflict resolutions below, so I expect it to report that drift rather than pass clean. The items under "Needs a decision" stop the automatic push on purpose.Needs a decision
xen/gntdev: don't copy to user space under priv->lock (#79)(downstream3f7ca89e3d70, now2ab25ce3a8e5) fixes the same map-ioctl/mmap deadlock as upstream7620a997e485 xen/gntdev: fix deadlock between map ioctl and mmap(Cc: stable), but in a different way. Both droppriv->lockbeforecopy_to_user(). On a failed copy, upstream re-takes the lock and looks the map up again byop.index/op.countwithgntdev_find_map_index(). Downstream holds an extra reference across the copy and removes the map only if that exact pointer is still onpriv->maps. What I committed: the downstream version, sogntdev_ioctl_map_grant_ref()is byte-identical to the old tip. It keeps upstream's lock ordering, the only constraint upstream's fix adds. It also can't remove a different map that was given the same index after a concurrent unmap. The other reading: take upstream's version and let the downstream commit shrink to almost nothing (or drop it), which keeps this function identical to mainline and makes future stable backports apply cleanly. A maintainer should pick one.x86/CPU/AMD: avoid printing reset reasons on Xen domU(downstreamd8221bf28f22, now4d13afdf368c) applied cleanly but is now dead code. Upstreamea51d1a4c31f x86/CPU/AMD: Don't print reset reason in VMs(Cc: stable) returns early fromprint_s5_reset_status_mmio()wheneverX86_FEATURE_HYPERVISORis set. That covers every Xen PV guest, dom0 included, so the downstreamXENPV && !xen_initial_domain()check after it can never fire. It also changes behaviour: a Xen PV dom0 used to print the S5 reset reason (downstream let dom0 through), and with upstream it no longer does. What I committed: the downstream commit, replayed unchanged (harmless and redundant). The other reading: drop the downstream commit as superseded byea51d1a4c31f. If dom0 should still print the reset reason, upstream's check would need a downstream exception. I did not add one.228e21764b38 Drivers: hv: Run VMBus kexec cleanup from syscore shutdownmoveshv_kexec_handler()(vmbus_initiate_unload()pluscpuhp_remove_state(hyperv_cpuhp_online)) from x86'shv_machine_shutdown()into a VMBus syscore.shutdownhook that runs wheneverkexec_is_in_progress(). Before, it ran only when Hyper-V was the primary hypervisor, becausehv_machine_shutdownis installed by the ms_hyperv platform code. Now it also runs in the downstream Xen-PV-dom0-nested-on-Hyper-V setup (hyperv_nested_on_xen), which registers VMBus throughhv_acpi_init()but whose machine ops are Xen's. No lines conflicted. The likely effect is good (VMBus is unloaded cleanly before kexec), but in the nested case nobody has tested this path: the SynIC/Xen-VIRQ bridge teardown happens throughhv_synic_cleanupwith Xen proxying the MSRs. What I committed: nothing changed; upstream's behaviour applies. A maintainer should confirm that kexec from a nested dom0 is expected to run this, or gate it on!hyperv_nested_on_xen.Conflicts resolved
sysctl: wire up PaX softmode setting(bd1c0dba57c3, now603bab1a0fce). File:kernel/sysctl.c. Collided with upstream443bea2f71c5 sysctl: consolidate CONFIG_SYSCTL into a single block in kernel/sysctl.c(andacd19f2c56f2), which movedsysctl_subsys_tableandsysctl_init_bases()inside the single#ifdef CONFIG_SYSCTLblock and dropped the inner#ifdef CONFIG_SYSCTL. I kept upstream's layout and added the downstreampax_table(underCONFIG_OPENPAX_SOFTMODE) right beforesysctl_subsys_table, plus itsregister_sysctl_init("kernel/pax", ...)call, in the new location. The commit's own +/- lines are identical to the original; the verifier shows it as "context moved" and as drift because the hunk now sits in a different place.use maximum supported mmap rnd entropy by default(849ed61740ea, nowd188c9b8fd78). File:arch/Kconfig. Collided with upstream8cdae75cebe0 arch/Kconfig: fix dead conditions by removing dead options, which deletedARCH_MMAP_RND_BITS_DEFAULT/ARCH_MMAP_RND_COMPAT_BITS_DEFAULTand theirdefault ... if ...lines (anintused as a condition is always false, so those lines never applied). I kept the upstream removal and applied the downstream change: the remaining fallback default goes from_MINto_MAXfor both symbols. Nothing downstream referenced the removed symbols, so behaviour is unchanged.9p/xen: reconnect to the backend on resume(f492e4d0cb26, now9e1163209bc2). File:net/9p/trans_xen.c. Collided with upstreamb3d07d64832c 9p/xen: Use array for data rings(the rings array becamerings[XEN_9PFS_NUM_RINGS]inside the priv) and7b6f2e3c29fb 9p/xen: drain response work after unbinding IRQ(Fixes, Cc: stable;cancel_work_sync()moved afterunbind_from_irqhandler()). I kept the embedded array: the downstream commit reuses the same ring slots on reconnect and never reallocates the array. That means dropping thekzalloc_objs/kfree(priv->rings)/if (priv->rings)lines fromxen_9pfs_front_init()andxen_9pfs_front_free(). In the newxen_9pfs_front_free_rings()I kept upstream's ordering: skip a ring with nointf, unbind the IRQ, thencancel_work_sync(). The downstream fieldsring_order,rings_semandconnectedand all of the reconnect logic are unchanged. The downstream disconnect path (schedule_work/flush_work, then free_rings) therefore also gets upstream's unbind-then-drain ordering.9p/xen: carry the 9P session across a restore(5b1ff5bbc8cb, now37e9d8605e2f). File:net/9p/trans_xen.c. Collided with upstream9b4cf60c6b73 9p/xen: validate the response size from the backend(Fixes, Cc: stable) inp9_xen_response(), and with the rings-array change above. I kept upstream'sh.size < sizeof(h) || h.size > queuedrejection first, then the downstream intercept of theP9_NOTAGcontrol reply on ring 0, so the control path never uses an unvalidatedh.size. A bad size nowbreaks before the intercept can completectl_done. That can't hang anything becausexen_9pfs_front_ctl()waits withwait_for_completion_timeout(). I dropped the downstream rings allocation and free again (the array is embedded) and keptINIT_LIST_HEAD(&priv->session),init_completion(&priv->ctl_done)andxen_9pfs_front_session_free().ring == priv->ringsis still valid C, because the array decays to&rings[0], so I left it as written. Own +/- lines are identical; the verifier lists it as "context moved".xen/gntdev: don't copy to user space under priv->lock (#79)(3f7ca89e3d70, now2ab25ce3a8e5). File:drivers/xen/gntdev.c. Collided with upstream7620a997e485 xen/gntdev: fix deadlock between map ioctl and mmap. Both fix the same deadlock. I kept the downstream reference-counted version (see the first UNSURE item). The upstreamgntdev_find_map_index()re-lookup lines do not appear in the final tree; this is the gntdev drift the verifier reports.Dropped downstream commits
None.
Build fixes
None. The workflow sent this branch back as a failed build. Its build logs hold only
build-arm64.log, which endskernel-build: arm64 vmlinux built(a pass), and there is no x86_64 log at all, so no build error was given to fix. Rebuildingd58f6bea5d31withkernel-build.shpassed on both arches (see Builds). I changed no commit, and the branch is stilld58f6bea5d31.Upstream commits that meet the series
Fixes (Fixes:, CVE or Cc: stable) that touch files the series touches, and what the final tree shows:
a02926f7c89b xen-blkback: Avoid UAF when reading I/O statistics(XSA-528 / CVE-2026-100082): intact.rings_lockis initialised inxen_blkif_alloc(), taken for write whenrings/nr_ringsare published and torn down, and for read in the statsshow_macro. Downstream4e54ad13ce40only addshost_nodeand node placement and doesn't touchrings/nr_rings.bfddc420afe8 xen-netback: defer xenvif_mcast_addr_list_free()(XSA-527 / CVE-2026-100081): intact. The only call is inbackend_disconnect()aftersynchronize_net()(drivers/net/xen-netback/xenbus.c:270). Downstream72d7a148f049does not bring back the call inxenvif_disconnect_data().7b6f2e3c29fb 9p/xen: drain response work after unbinding IRQ: intact inxen_9pfs_front_free_rings()(see conflicts), which also covers the downstream disconnect and reconnect-error paths.9b4cf60c6b73 9p/xen: validate the response size from the backend: intact, and placed before the downstream control-reply intercept (see conflicts).4d183e6f1532 9p/xen: fix refcount leak in p9_xen_response() on wrong tag: intact (p9_req_put()on the wrong-tag path).241d46901226 9p: reject RERROR/RLERROR with zero errcode: intact inp9_check_errors(). Downstream doesn't touch it. The downstream control path intrans_xen.calready rejectsecode == 0andecode > MAX_ERRNOwith-EPROTO.2acd800de882 9p: fix caches_show() out-of-bounds write: intact (sysfs_emit_at). Downstream only adds a separatefeaturesattribute next to it.7620a997e485 xen/gntdev: fix deadlock between map ioctl and mmap: superseded by the downstream equivalent. The lock ordering is kept; see UNSURE.ea51d1a4c31f x86/CPU/AMD: Don't print reset reason in VMs: intact. It makes downstreamd8221bf28f22redundant; see UNSURE.2b66f2cc7532 Xen/gnttab: improve error handling in gnttab_dma_{alloc,free}_pages(),4ebb3b8d129c xen: don't allow gnttab_dma_[alloc|free]_pages() in PV guest,704f812cdab8 xen/grant-table: annotate on-stack work: intact. Downstream grant-table changes (xen_mfn_to_node,gnttab_alloc_pages_node) don't touch those functions.0f4a9e682f7f xattr: always invoke cap_convert_nscap() for security.capability writes: intact. Downstream only addspax_getxattr().51e766bcd35e init/main: fix false-positive kernel panic on environment variable overwrite,b82ed0e3a1da init/main: fix off-by-one in argv_init cleanup,278b2abf660a init/bootconfig: pass init arguments in documented order: intact. Downstream only addspax_softmodeand its__setup.1c8f6308e557 iommu/dma: Catch scatterlist length overflows: intact. Downstream only changes the PCI-window reservation condition iniova_reserve_iommu_regions().13fc99f46399,c386d799f423,769417e410dd,168ed9c5d5ec,0193e8336802,b1ee87802111,aa09311c40f6,f987d9f38a46,ff82ce7f90f5,4c2dec1dce36,8146c6478224,ef694e2d2101,68cb098eadda,7135a05838e8,dc0b17144d62) against the OpenPaX commits: MM_AUDIT_PLACEHOLDEROther upstream changes that meet the series without a fix tag:
8cdae75cebe0 arch/Kconfig: fix dead conditions by removing dead options: resolved; see conflicts.443bea2f71c5,acd19f2c56f2(sysctl consolidation): resolved; see conflicts.242bac52294e,d08ab285b096,2bfbb09b505c(sysctl conversion rework): no interaction, sincepax_tableuses plainproc_dointvec.b3d07d64832c 9p/xen: Use array for data rings: resolved; see conflicts.33625427a382 net/9p: add vsock transport,8bb0279930cc,9eb9b583ea35,7fb9bd365905,5852cdf0006f: no interaction with the downstream 9p client andshare_clientchanges.6eecb67b69af xen/grant-table: stop setting PG_private on pages for grant mapping: no downstream code usesPagePrivate/PG_privatein Xen code.gnttab_alloc_pages_node()still goes throughgnttab_pages_set_private()/gnttab_pages_clear_private().228e21764b38 Drivers: hv: Run VMBus kexec cleanup from syscore shutdown: see UNSURE.dd57772e36f7 Drivers: hv: vmbus: Check ACPI companion before use: no interaction with the downstreamhyperv_nested_on_xenIRQ-bridge branches invmbus_bus_init()/vmbus_exit().118bdac10546 Drivers: hv: Add logical device ID registry for vPCI devices: the registry hashtable is initialised inhv_common_init(), which the downstreamhyperv_init_nested_on_xen()calls, sohv_pci_probe()still works when nested. No overlap with the downstream nested-MSI anduse_callschanges.2690d4156b51 iommu/hyperv: Add para-virtualized IOMMU support for Hyper-V guest:x86_init.iommu.iommu_init = hv_iommu_initis set only inhyperv_init(), not in the downstream nested-on-Xen bring-up, so it can't race the Xen PV-IOMMU.1ae140d0f272 iommu: Make iommu-dma helpers private: upstream now includes"iommu-priv.h"indma-iommu.c. Downstream1b8277179ffealso adds that include (in a different spot), so the file now includes the header twice. That's harmless because the header has an include guard, so I left it.fdd81f15fc43,641a9b9cc697,95d2fec40c48,0bf031ba8c49,42eed9bb9ca4(iommu/dma-mapping): no interaction. The downstream Xen IOMMU driver doesn't usedma_opt_mapping_size()or the privatised helpers.d0ec95a8a4e7 hv_netvsc: Advertise the SR-IOV capability for CoCo VMs,56fb075b77f8 mshv: Redesign hypervisor memory deposit logic: no interaction with the downstream nested-on-Xen netvsc DMA mapping or the asm-generic declarations.0af298e62e5c treewide: refresh kmalloc_obj() conversions,bbe5d3b3d402,c2d4ccfbba3f,f000a06da2a5,10823ced2f18,8ebd2eb87f50,5728d3523f12,084100a1b579,aabe6e1b36ce,677f83795fdf,958c01049e0a,38828d44f31e,799203afd298,e3e7a85730a4,abaa0835b4f3,264d8fd2794f,28d5df7b7274,537ec2b15fdd,681f3b68f435,cf27412f19c9,2e83e2ce34ff,69d38c792c53: touch shared files only in places away from downstream hunks; no interaction. No downstream code referencesDEBUG_WX,VM_SPECIAL,is_vm_hugetlb_page,kho_scratchorhv_setup_kexec_handler.Builds
Rebuilt in the repair run, at
d58f6bea5d31:kernel-build.sh x86_64, Edera host.config from edera-dev/linux-kernel-oci@530e23bbf906, the same config commit as the rebase run). The workflow's build left no x86_64 log, so whatever it saw there could not be reproduced or diagnosed from what this run was given.kernel-build.sh arm64, defconfig). This matches the workflow's ownbuild-arm64.log, which also passed.verify-upstream-rebase.shgives the same result as before the repair: base and linearity pass, and the 2 changed commits and the drift are exactly the conflict resolutions above. There is no new drift.