1. Good catch but I’m not sure it’s related to FUSE changes. Starting the “blocking filesystem” leads to executing /usr/lib/vmware/bin/vmware-vmblock-fuse with arguments “-o subtype=vmware-vmblock,default_permissions,allow_other /var/run/vmblock-fuse”. But vmware-vmblock-fuse is actually using a wrapper named appLoader which first loads some dynamic libraries and then executes itself again with the same parameters (not sure what is the purpose of this exercise as execve() drops everything, maybe it sets some environment variables). Except the last parameter, “/var/run/vmblock-fuse” is missing for some reason, leading to error message “fuse: missing mountpoint parameter”.

   However, this “blocking filesystem” is AFAIK only used for directory sharing which is something I don’t use. I rather suspect that other strange problem are also symptoms of the same issue: appLoader losing its last argument somehow.

   Reply ↓

   • This looks promissing: with 4.18-rc1, reading /proc/self/cmdline doesn’t get trailing null byte which is there with older kernels.

     Reply ↓

     • This (kernel) patch should fix the issue with /proc/$pid/cmdline:

       https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg1715503.html

       Reply ↓

       • I’m sorry, this should be the correct version:

         https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg1715506.html

         Reply ↓

         • Thanks! I’ve applied the V2 version of the patch, and VMware behaves correctly, now..
           Hopefully this patch will pass through the LKML ‘review process’ without too much hassle..
           RG.

This site uses Akismet to reduce spam. Learn how your comment data is processed.