From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43) id 1MvWv1-0007LC-Hc for qemu-devel@nongnu.org; Wed, 07 Oct 2009 09:51:03 -0400 Received: from exim by lists.gnu.org with spam-scanned (Exim 4.43) id 1MvWuw-0007D1-Oq for qemu-devel@nongnu.org; Wed, 07 Oct 2009 09:51:02 -0400 Received: from [199.232.76.173] (port=49457 helo=monty-python.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1MvWuw-0007Cj-9S for qemu-devel@nongnu.org; Wed, 07 Oct 2009 09:50:58 -0400 Received: from mx1.redhat.com ([209.132.183.28]:53018) by monty-python.gnu.org with esmtp (Exim 4.60) (envelope-from ) id 1MvWuv-0007yR-Et for qemu-devel@nongnu.org; Wed, 07 Oct 2009 09:50:58 -0400 Date: Wed, 7 Oct 2009 15:48:53 +0200 From: "Michael S. Tsirkin" Message-ID: <20091007134853.GA9769@redhat.com> References: <20091007123348.GA31537@redhat.com> <4ACC9133.9060903@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <4ACC9133.9060903@redhat.com> Subject: [Qemu-devel] Re: [PATCH] qemu/pci: optimize pci config handling List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Paolo Bonzini Cc: Isaku Yamahata , qemu-devel@nongnu.org On Wed, Oct 07, 2009 at 03:01:39PM +0200, Paolo Bonzini wrote: > On 10/07/2009 02:33 PM, Michael S. Tsirkin wrote: >> There's no need to save all of config space before each config cycle: >> just the 64 byte header is enough for our purposes. This will become >> more important as we add pci express support, which has 4K config space. > > You can even go a step further and save it only if something is actually > being changed. Untested though. I'm actively trying to get out of range-checking address. What you porpose here is certainly more code than we had. So why is this a good idea? > Not-quite-signed-off-by: Paolo Bonzini > > Paolo > diff --git a/hw/pci.c b/hw/pci.c > index 41e99a9..f9959fc 100644 > --- a/hw/pci.c > +++ b/hw/pci.c > @@ -541,19 +541,26 @@ uint32_t pci_default_read_config(PCIDevice *d, > > void pci_default_write_config(PCIDevice *d, uint32_t addr, uint32_t val, int l) > { > - uint8_t orig[PCI_CONFIG_SPACE_SIZE]; > + uint8_t orig[PCI_CONFIG_HEADER_SIZE]; > int i; > > - /* not efficient, but simple */ > - memcpy(orig, d->config, PCI_CONFIG_SPACE_SIZE); > - for(i = 0; i < l && addr < PCI_CONFIG_SPACE_SIZE; val >>= 8, ++i, ++addr) { > - uint8_t wmask = d->wmask[addr]; > - d->config[addr] = (d->config[addr] & ~wmask) | (val & wmask); > + /* not efficient, but simple. If modifying the header, save it so we > + can compare its contents later. */ > + if (addr < sizeof orig) { > + memcpy(orig, d->config, sizeof orig); > + } > + > + for(i = 0; i < l && addr+i < PCI_CONFIG_SPACE_SIZE; val >>= 8, ++i) { > + uint8_t wmask = d->wmask[addr+i]; > + d->config[addr+i] = (d->config[addr+i] & ~wmask) | (val & wmask); > + } > + > + if (addr < sizeof orig) { > + if (memcmp(orig + PCI_BASE_ADDRESS_0, d->config + PCI_BASE_ADDRESS_0, 24) > + || ((orig[PCI_COMMAND] ^ d->config[PCI_COMMAND]) > + & (PCI_COMMAND_MEMORY | PCI_COMMAND_IO))) > + pci_update_mappings(d); > } > - if (memcmp(orig + PCI_BASE_ADDRESS_0, d->config + PCI_BASE_ADDRESS_0, 24) > - || ((orig[PCI_COMMAND] ^ d->config[PCI_COMMAND]) > - & (PCI_COMMAND_MEMORY | PCI_COMMAND_IO))) > - pci_update_mappings(d); > } > > void pci_data_write(void *opaque, uint32_t addr, uint32_t val, int len)