From: Nathaniel C. <na...@mi...> - 2006-04-12 14:14:41
|
Because when I added the comments when I added the WRITE_VERIFY* initially, when there was no verification. It's vestigial and incorrect. - nate Vladislav Bolkhovitin wrote: > Commited, thanks. > > (I didn't understand why you think that WRITE_VERIFY* don't actually > verify. I fixed fsync'ing, if you mean that. I know that > fileio_exec_verify() has at least one bug with global virt_dev->bufflen) > > Vlad > > Nathaniel Clark wrote: >> I've updated the kernel patch for the 2.6.16 kernel, I'll have an >> updated qla2x00tgt/qla2xxx driver soon also. >> Changes in scst patch: >> - move scst_set_cmd_error_sense declaration into scsi_tgt.h >> - change EXPORT_SYMBOL_GPL's that describe the target and dev_handler >> interfaces into EXPORT_SYMBOL's >> - struct scst_cmd: make bufflen a size_t instead of int (and fix >> necessary TRACE statements) >> - remove EXTRACHECKS code in alloc_space from the code path that is >> exercised when allocations are already completed in the device handler >> (it handled that incorrectly). >> - replace all "TRACE(DEBUG,..." statements with "TRACE(TRACE_DEBUG,..." >> - print S/G buffers with TRACE_RECV_TOP and TRACE_SEND_BOT are enabled >> - add check for existence of on_free_cmd before calling on dev_handler >> > > |