On Wed, Aug 12, 2026 at 04:31:10PM +0000, [email protected] wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) 
> to consider:
> - [High] Incomplete tracefs gating: Caller-created tracefs files bypass the 
> newly introduced `trace_remote_tracefs_open` gating, leaving them fully 
> accessible if registration fails.
> --
> 
> --- Patch [1]: [PATCH v5 01/18] tracing/remotes: Gate tracefs files opening 
> on trace remote registration ---
> commit 63ef30c302e1341f240b4f5b82d4e65d421da4cf
> Author: Vincent Donnefort <[email protected]>
> tracing/remotes: Gate tracefs files opening on trace remote registration
> 
> This commit adds a global list of registered trace remotes and a gating
> function to prevent access to tracefs files for remotes that are being
> destroyed or failed to register. Standard trace remote file operations
> are wrapped to check this list before allowing open().
> 
> > --- a/kernel/trace/trace_remote.c
> > +++ b/kernel/trace/trace_remote.c
> 
> [ ... ]
> 
> > @@ -932,8 +999,12 @@ int trace_remote_register(const char *name, struct 
> > trace_remote_callbacks *cbs,
> >     }
> >  
> >     ret = cbs->init ? cbs->init(remote->dentry, priv) : 0;
> > -   if (ret)
> > +   if (ret) {
> >             pr_err("Init failed for trace remote '%s' (%d)\n", name, ret);
> > +   } else {
> > +           guard(mutex)(&trace_remotes_lock);
> > +           list_add(&remote->node, &trace_remotes);
> > +   }
> >  
> >     return ret;
> >  }
> 
> [Severity: High]
> Does this leave caller-created tracefs files exposed if cbs->init() fails?
> 
> If a remote's initialization fails halfway through, any custom tracefs files
> already created by cbs->init() will leak and remain accessible. Because
> trace_remote_tracefs_open() is static to trace_remote.c, the caller cannot
> wrap their custom files' open() callbacks with this new gating mechanism.
> 
> For example, if KVM's hyp_trace_init_tracefs() successfully creates
> a 'write_event' file but fails later, trace_remote_register() returns an
> error and the remote is not added to the trace_remotes list. 
> 
> Userspace could then open the leaked 'write_event' file, bypassing the
> gating entirely, and trigger kvm_call_hyp_nvhe() against the uninitialized
> hypervisor stub.
> 
> Should trace_remote_tracefs_open() be exported so that subsystems can gate
> their custom files, or should the leaked files be cleaned up when
> trace_remote_register() returns an error?
> 
> -- 
> Sashiko AI review ยท 
> https://sashiko.dev/#/patchset/[email protected]?part=1

In the following patch, the tracefs folder is entirely removed on init failure.

-- 
Vincent

Reply via email to