From 0fb5b3282d50520f857f584126a33acbf866b774 Mon Sep 17 00:00:00 2001 From: Jeremy Allison Date: Mon, 31 Mar 2008 16:46:20 -0700 Subject: Reduce the race condition in Samba4 in RAW-RENAME test. We rename a file using trans2 setfileinfo on one connection, and then check the file name has changed on the other. In Samba we achieve this by sending a local message to the other process. This change causes us to re-scan for incoming messages after we've woken up from the select (which is cheap if there are no pending messages). This reduces the race significantly. Volker please review. Jeremy. (cherry picked from commit a7499e994aef743ea9c443f9a1618b262f6eda93) --- source/smbd/process.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'source/smbd') diff --git a/source/smbd/process.c b/source/smbd/process.c index 9c01bbaa4a8..03216a07001 100644 --- a/source/smbd/process.c +++ b/source/smbd/process.c @@ -853,6 +853,15 @@ static NTSTATUS receive_message_or_smb(TALLOC_CTX *mem_ctx, char **buffer, goto again; } + /* + * We've just woken up from a protentially long select sleep. + * Ensure we process local messages as we need to synchronously + * process any messages from other smbd's to avoid file rename race + * conditions. This call is cheap if there are no messages waiting. + * JRA. + */ + message_dispatch(smbd_messaging_context()); + /* if we get EINTR then maybe we have received an oplock signal - treat this as select returning 1. This is ugly, but is the best we can do until the oplock code knows more about -- cgit