On Fri, 2003-03-07 at 16:30, Andrew Morton wrote: > Daniel McNeil wrote: > > > > This adds i_seqcnt to inode structure and then uses i_size_read() and > > i_size_write() to provide atomic access to i_size. > > Ho hum. Everybody absolutely hates this, but I guess we should do it :( > > > +static inline void i_size_write(struct inode * inode, loff_t i_size) > > +{ > > +#if BITS_PER_LONG==32 && defined(CONFIG_SMP) > > + write_seqcntbegin(&inode->i_size_seqcnt); > > + inode->i_size = i_size; > > + write_seqcntend(&inode->i_size_seqcnt); > > +#elif BITS_PER_LONG==32 && defined(CONFIG_PREMPT) > > + prempt_disable(); > > + inode->i_size = i_size; > > + prempt_enable(); > > +#else > > + inode->i_size = i_size; > > +#endif > > +} > > You've used "PREMPT" and "prempt" throughput the patch. It is in fact > "PREEMPT" and "preempt". > > Could you please fix that up and send me fresh copies? Probably as > attachments - your mailer wordwrapped the patches. > > Thanks. Here are the updated patches as attachments. Thanks, -- Daniel McNeil