Re: Asynch I/O broken in 2.2.15

[Jeff V. Merkey]

"Jeff V. Merkey" wrote:
> 
>> 
> Also, your statement about the pages being locked is incorrect, these
> are block addresses on NetWare partitions, no one else should be using
> them or even touching them (and are not).  How can what you said be
> true?
> 

One point that may help clarify.  Native NetWare allows folks in the LRU
cache to write to pages of cached memory, even while the driver is
performing an active DMA on the memory.  The reason it allows this is
for speed (and it's one reason it exhibits sucn horrific speed with file
system IO operations).  The assumption here is that eventually, even if
two users write the same page, when a volume is dismounted and it's
dirty blocks finally flushed, all file system writes will coaalesce when
a volume is cleanly dismounted.  

Jeff

> Jeff
> 
> >
> > --Stephen
> 
>   ------------------------------------------------------------------------
>             Name: aio.c
>    aio.c    Type: Notepad File (application/x-unknown-content-type-cfile)
>         Encoding: base64