1 File Locking Release Notes 2 3 Andy Walker <andy@lysaker.kvaerner.no> 4 5 12 May 1997 6 7 81. What's New? 9-------------- 10 111.1 Broken Flock Emulation 12-------------------------- 13 14The old flock(2) emulation in the kernel was swapped for proper BSD 15compatible flock(2) support in the 1.3.x series of kernels. With the 16release of the 2.1.x kernel series, support for the old emulation has 17been totally removed, so that we don't need to carry this baggage 18forever. 19 20This should not cause problems for anybody, since everybody using a 212.1.x kernel should have updated their C library to a suitable version 22anyway (see the file "Documentation/Changes".) 23 241.2 Allow Mixed Locks Again 25--------------------------- 26 271.2.1 Typical Problems - Sendmail 28--------------------------------- 29Because sendmail was unable to use the old flock() emulation, many sendmail 30installations use fcntl() instead of flock(). This is true of Slackware 3.0 31for example. This gave rise to some other subtle problems if sendmail was 32configured to rebuild the alias file. Sendmail tried to lock the aliases.dir 33file with fcntl() at the same time as the GDBM routines tried to lock this 34file with flock(). With pre 1.3.96 kernels this could result in deadlocks that, 35over time, or under a very heavy mail load, would eventually cause the kernel 36to lock solid with deadlocked processes. 37 38 391.2.2 The Solution 40------------------ 41The solution I have chosen, after much experimentation and discussion, 42is to make flock() and fcntl() locks oblivious to each other. Both can 43exists, and neither will have any effect on the other. 44 45I wanted the two lock styles to be cooperative, but there were so many 46race and deadlock conditions that the current solution was the only 47practical one. It puts us in the same position as, for example, SunOS 484.1.x and several other commercial Unices. The only OS's that support 49cooperative flock()/fcntl() are those that emulate flock() using 50fcntl(), with all the problems that implies. 51 52 531.3 Mandatory Locking As A Mount Option 54--------------------------------------- 55 56Mandatory locking, as described in 'Documentation/mandatory.txt' was prior 57to this release a general configuration option that was valid for all 58mounted filesystems. This had a number of inherent dangers, not the least 59of which was the ability to freeze an NFS server by asking it to read a 60file for which a mandatory lock existed. 61 62From this release of the kernel, mandatory locking can be turned on and off 63on a per-filesystem basis, using the mount options 'mand' and 'nomand'. 64The default is to disallow mandatory locking. The intention is that 65mandatory locking only be enabled on a local filesystem as the specific need 66arises. 67 68Until an updated version of mount(8) becomes available you may have to apply 69this patch to the mount sources (based on the version distributed with Rick 70Faith's util-linux-2.5 package): 71 72*** mount.c.orig Sat Jun 8 09:14:31 1996 73--- mount.c Sat Jun 8 09:13:02 1996 74*************** 75*** 100,105 **** 76--- 100,107 ---- 77 { "noauto", 0, MS_NOAUTO }, /* Can only be mounted explicitly */ 78 { "user", 0, MS_USER }, /* Allow ordinary user to mount */ 79 { "nouser", 1, MS_USER }, /* Forbid ordinary user to mount */ 80+ { "mand", 0, MS_MANDLOCK }, /* Allow mandatory locks on this FS */ 81+ { "nomand", 1, MS_MANDLOCK }, /* Forbid mandatory locks on this FS */ 82 /* add new options here */ 83 #ifdef MS_NOSUB 84 { "sub", 1, MS_NOSUB }, /* allow submounts */ 85