On Thu, Aug 11, 2022 at 01:04:35PM +0200, Alexander Grund wrote: > On 11.08.22 12:21, Greg KH wrote: > > On Thu, Aug 11, 2022 at 11:46:09AM +0200, Alexander Grund wrote: > >> I mean this patch removes a superflous pointer of the superblock struct > >> making the kernel use less memory. > > > > Also, how much measurable memory does this save? You did not quantify > > it. > > It saves one pointer, i.e. usually 8 Byte, per superblock when using SELinux. > > I don't know how many of those superblocks will be allocated in total on usual systems > as I'm not familiar with the details of the filesystems. > However following one callchain leads to > > > /* > > * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that > > * will never be mountable) > > */ > > struct dentry *mount_pseudo_xattr(struct file_system_type *fs_type, char *name, > > So it seems one superblock even for each pseudo-fs of which there can be many. > > A quick experiment [1] on my phone shows about 300 superblocks allocated which means > that the patch saves a bit over 2kB of memory. > So not that much on usual systems but could be much for some embedded systems. Again, we are not adding "slim the kernel down by an infinitesimal %" patches to older stable kernels, especially ones that only have a few more months to live. Let's stick with real bugfixes please, that's what matters here. thanks, greg k-h
