Date: Sat, 23 Mar 2013 18:45:02 -0500 From: Alan Cox <alan.l.cox@gmail.com> To: Konstantin Belousov <kostikbel@gmail.com> Cc: arch@freebsd.org, David Wolfskill <david@catwhisker.org> Subject: Re: VM_BCACHE_SIZE_MAX on i386 Message-ID: <CAJUyCcN6N_t6qRb5Hzf6HEU-8RU8-xPqz-0NHDOkGHWqAoUOKA@mail.gmail.com> In-Reply-To: <20130323221039.GO3794@kib.kiev.ua> References: <20130323211001.GN3794@kib.kiev.ua> <CAJUyCcMXysO95CqxtWnbkzU5nJx757ktijQi6D%2BUEU5BE=z91g@mail.gmail.com> <20130323221039.GO3794@kib.kiev.ua>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Mar 23, 2013 at 5:10 PM, Konstantin Belousov <kostikbel@gmail.com>wrote: > On Sat, Mar 23, 2013 at 04:49:06PM -0500, Alan Cox wrote: > > The additional KVA that we had to reserve for the vm_page radix tree > nodes > > already got me thinking about VM_BCACHE_SIZE_MAX a couple weeks ago. > With > > the extra KVA pressure that is inherent to PAE, e.g., a larger vm_page > > struct, we really can't afford to allow the buffer map KVA allocation to > > grow much beyond what it would be for a 4GB machine anyway. Moreover, > your > > work makes the size of the buffer map less important, because it will see > > decreasing use as drivers are converted to allow unmapped I/O. So, I > would > > encourage you to simply use the same cap based on a 4 GB machine for both > > PAE and !PAE. > > I did not checked it, but isn't default PAE config splits user/kernel > on the 2GB boundary ? This makes the KVA pressure on PAE less severe, > but user mode should be not very happy. > > No, the PAE split between user/kernel is the same as !PAE. The PAE config is deceptive because it takes twice as many page table pages to get the same 1 GB of KVA. > Currently, the number of the regular buffer headers allocated is equal > to the size of the buffer map / BKVASIZE still. This could be changed > now, I believe that Peter' testing fixed most of the bugs in the > handling of the maxbufspace. I was too coward to make this change > together with the rest of the work. But it could indeed be useful, since > buffer map is used only by the metadata buffers for UFS. > > I did changed VM_BCACHE_SIZE_MAX as you suggested, thanks. > > http://people.freebsd.org/~kib/misc/i386_maxbcache.2.patch >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJUyCcN6N_t6qRb5Hzf6HEU-8RU8-xPqz-0NHDOkGHWqAoUOKA>