CVE-2024-42256

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> cifs: Fix server re-repick on subrequest retry<br /> <br /> When a subrequest is marked for needing retry, netfs will call<br /> cifs_prepare_write() which will make cifs repick the server for the op<br /> before renegotiating credits; it then calls cifs_issue_write() which<br /> invokes smb2_async_writev() - which re-repicks the server.<br /> <br /> If a different server is then selected, this causes the increment of<br /> server-&gt;in_flight to happen against one record and the decrement to happen<br /> against another, leading to misaccounting.<br /> <br /> Fix this by just removing the repick code in smb2_async_writev(). As this<br /> is only called from netfslib-driven code, cifs_prepare_write() should<br /> always have been called first, and so server should never be NULL and the<br /> preparatory step is repeated in the event that we do a retry.<br /> <br /> The problem manifests as a warning looking something like:<br /> <br /> WARNING: CPU: 4 PID: 72896 at fs/smb/client/smb2ops.c:97 smb2_add_credits+0x3f0/0x9e0 [cifs]<br /> ...<br /> RIP: 0010:smb2_add_credits+0x3f0/0x9e0 [cifs]<br /> ...<br /> smb2_writev_callback+0x334/0x560 [cifs]<br /> cifs_demultiplex_thread+0x77a/0x11b0 [cifs]<br /> kthread+0x187/0x1d0<br /> ret_from_fork+0x34/0x60<br /> ret_from_fork_asm+0x1a/0x30<br /> <br /> Which may be triggered by a number of different xfstests running against an<br /> Azure server in multichannel mode. generic/249 seems the most repeatable,<br /> but generic/215, generic/249 and generic/308 may also show it.