{"id":"AZL-115031","summary":"CVE-2026-99528 affecting package kernel 6.6.161.1-1","details":"In the Linux kernel, the following vulnerability has been resolved:\n\nSUNRPC: fix netns use-after-free in write_gssp()\n\nWhile fuzzing with a customized syzkaller, I hit a refcount warning in\nxprt_alloc(), reached from a write to /proc/net/rpc/use-gss-proxy:\n\n  refcount_t: addition on 0; use-after-free.\n  WARNING: lib/refcount.c:25 at refcount_warn_saturate+0xf8/0x120\n  CPU: 0 PID: 10404 Comm: syz.0.17 Not tainted 7.3.0-rc1-00096-gcfebfd3db73d\n  Call Trace:\n   \u003cTASK\u003e\n   xprt_alloc+0x83f/0x9d0\n   xs_setup_xprt+0xaf/0x3c0\n   xs_setup_local+0x47/0x7f0\n   xprt_create_transport+0x16c/0x730\n   rpc_create+0x38e/0x7f0\n   gssp_rpc_create+0xe2/0x180\n   set_gssp_clnt+0xba/0x1b0\n   write_gssp+0x200/0x310\n   proc_reg_write+0x240/0x330\n   vfs_write+0x2aa/0x1050\n   ksys_write+0x12a/0x250\n   do_syscall_64+0x117/0x750\n   entry_SYSCALL_64_after_hwframe+0x77/0x7f\n   \u003c/TASK\u003e\n\ncreate_use_gss_proxy_proc_entry() stores the struct net pointer as the\nproc entry's private data without taking a reference on it, and an open\nfile descriptor does not pin the namespace either, as procfs only pins\nthe proc_dir_entry. write_gssp() reads that pointer back with\npde_data() and hands it to set_gssp_clnt(), which reaches xprt_init():\n\n\txprt-\u003exprt_net = get_net_track(net, &xprt-\u003ens_tracker, GFP_KERNEL);\n\nget_net_track() -\u003e get_net() -\u003e refcount_inc() is unconditional, so\nrpc_create() assumes its caller holds a reference on args.net.\nwrite_gssp() does not.\n\nDropping the last ns.count reference does not make the file go away.\n__put_net() only queues cleanup_net() on a workqueue, and the\nremove_proc_entry() that fences off further writes runs from the\nrpcsec_gss pernet .exit method, that is, from inside cleanup_net(). A\nwrite landing between those two points increments a refcount that is\nalready zero. cleanup_net() does not re-read ns.count, so the namespace\nis freed anyway and the rpc_xprt -- along with the AF_LOCAL socket\nopened for it -- is left pointing at freed memory. The resurrected\ncount is visible to the rest of that teardown too: with\nCONFIG_IPV6_MROUTE the same cleanup_net() worker then trips\n!mr_can_free_table() in ip6mr_free_table(), which tests check_net() on\nthe namespace it is freeing.\n\nIt reduces to opening /proc/net/rpc/use-gss-proxy inside a new network\nnamespace, calling setns() back to the original one to drop the last\nreference, and then writing \"1\" to the still-open descriptor.\n\nTake the reference in write_gssp() itself and refuse the write when the\nnamespace is already gone. This is what procfs does for every other\n/proc/net file: get_proc_net() is maybe_get_net(PDE_NET(PDE(inode)))\nand seq_open_net() returns -ENXIO when it fails. PDE_NET() cannot be\nreused here because the parent directory /proc/net/rpc carries no\nnamespace pointer. Adding a .pre_exit to rpcsec_gss_net_ops so the\nentry is removed earlier would only narrow the window, since pre_exit\nalso runs from cleanup_net().\n\nLive namespaces are unaffected: a write still returns the error from\nset_gssp_clnt(), a write of \"2\" still returns -EINVAL, and reads are\nuntouched.","modified":"2026-11-20T14:17:30.847911760Z","published":"2026-11-19T15:18:08Z","upstream":["CVE-2026-99528"],"references":[{"type":"WEB","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-99528"}],"affected":[{"package":{"name":"kernel","ecosystem":"Azure Linux:3","purl":"pkg:rpm/azure-linux/kernel"},"ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"last_affected":"6.6.161.1-1"}]}],"database_specific":{"source":"https://github.com/microsoft/AzureLinuxVulnerabilityData/blob/main/osv/AZL-115031.json"}}],"schema_version":"1.9.2"}