Message ID | 20180601175910.28146-1-peter.maydell@linaro.org |
---|---|
State | Superseded |
Headers | show |
Series | [RFC] xen: Don't use memory_region_init_ram_nomigrate() in pci_assign_dev_load_option_rom() | expand |
On Fri, Jun 01, 2018 at 06:59:10PM +0100, Peter Maydell wrote: > The xen pci_assign_dev_load_option_rom() currently creates a RAM > memory region with memory_region_init_ram_nomigrate(), and then > manually registers it with vmstate_register_ram(). In fact for > its only callsite, the 'owner' pointer we use for the init call > and the '&dev->qdev' pointer we use for the vmstate_register_ram() > call refer to the same object. Simplify the function to only > take a pointer to the device once instead of twice, and use > memory_region_init_ram() which automatically does the vmstate > register for us. > > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > This is a fairly trivial no-behaviour-change code cleanup, but > I've marked it as RFC because I don't have a setup for doing > more than just compile-testing Xen related patches. > This was found as part of a sweep through for code using > the _nomigrate versions of functions. That patch looks fine, and seams fine after hacking my way into testing the change. Acked-by: Anthony PERARD <anthony.perard@citrix.com> Thanks, -- Anthony PERARD
On 5 June 2018 at 17:55, Anthony PERARD <anthony.perard@citrix.com> wrote: > On Fri, Jun 01, 2018 at 06:59:10PM +0100, Peter Maydell wrote: >> The xen pci_assign_dev_load_option_rom() currently creates a RAM >> memory region with memory_region_init_ram_nomigrate(), and then >> manually registers it with vmstate_register_ram(). In fact for >> its only callsite, the 'owner' pointer we use for the init call >> and the '&dev->qdev' pointer we use for the vmstate_register_ram() >> call refer to the same object. Simplify the function to only >> take a pointer to the device once instead of twice, and use >> memory_region_init_ram() which automatically does the vmstate >> register for us. >> >> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> >> --- >> This is a fairly trivial no-behaviour-change code cleanup, but >> I've marked it as RFC because I don't have a setup for doing >> more than just compile-testing Xen related patches. >> This was found as part of a sweep through for code using >> the _nomigrate versions of functions. > > That patch looks fine, and seams fine after hacking my way into testing > the change. > > Acked-by: Anthony PERARD <anthony.perard@citrix.com> Thanks for the review. Stefano, do you want to take it via the xen subtree? thanks -- PMM
On Fri, Jun 15, 2018 at 11:06:47AM +0100, Peter Maydell wrote: > On 5 June 2018 at 17:55, Anthony PERARD <anthony.perard@citrix.com> wrote: > > On Fri, Jun 01, 2018 at 06:59:10PM +0100, Peter Maydell wrote: > >> The xen pci_assign_dev_load_option_rom() currently creates a RAM > >> memory region with memory_region_init_ram_nomigrate(), and then > >> manually registers it with vmstate_register_ram(). In fact for > >> its only callsite, the 'owner' pointer we use for the init call > >> and the '&dev->qdev' pointer we use for the vmstate_register_ram() > >> call refer to the same object. Simplify the function to only > >> take a pointer to the device once instead of twice, and use > >> memory_region_init_ram() which automatically does the vmstate > >> register for us. > >> > >> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > >> --- > >> This is a fairly trivial no-behaviour-change code cleanup, but > >> I've marked it as RFC because I don't have a setup for doing > >> more than just compile-testing Xen related patches. > >> This was found as part of a sweep through for code using > >> the _nomigrate versions of functions. > > > > That patch looks fine, and seams fine after hacking my way into testing > > the change. > > > > Acked-by: Anthony PERARD <anthony.perard@citrix.com> > > Thanks for the review. Stefano, do you want to take it via the > xen subtree? Hi Peter, We don't have any other patch for xen, so feel free to commit this one. -- Anthony PERARD
On Fri, 15 Jun 2018, Anthony PERARD wrote: > On Fri, Jun 15, 2018 at 11:06:47AM +0100, Peter Maydell wrote: > > On 5 June 2018 at 17:55, Anthony PERARD <anthony.perard@citrix.com> wrote: > > > On Fri, Jun 01, 2018 at 06:59:10PM +0100, Peter Maydell wrote: > > >> The xen pci_assign_dev_load_option_rom() currently creates a RAM > > >> memory region with memory_region_init_ram_nomigrate(), and then > > >> manually registers it with vmstate_register_ram(). In fact for > > >> its only callsite, the 'owner' pointer we use for the init call > > >> and the '&dev->qdev' pointer we use for the vmstate_register_ram() > > >> call refer to the same object. Simplify the function to only > > >> take a pointer to the device once instead of twice, and use > > >> memory_region_init_ram() which automatically does the vmstate > > >> register for us. > > >> > > >> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > > >> --- > > >> This is a fairly trivial no-behaviour-change code cleanup, but > > >> I've marked it as RFC because I don't have a setup for doing > > >> more than just compile-testing Xen related patches. > > >> This was found as part of a sweep through for code using > > >> the _nomigrate versions of functions. > > > > > > That patch looks fine, and seams fine after hacking my way into testing > > > the change. > > > > > > Acked-by: Anthony PERARD <anthony.perard@citrix.com> > > > > Thanks for the review. Stefano, do you want to take it via the > > xen subtree? > > Hi Peter, > > We don't have any other patch for xen, so feel free to commit this > one. Yep, go ahead
diff --git a/hw/xen/xen_pt.h b/hw/xen/xen_pt.h index aa39a9aa5f..dbee3308fd 100644 --- a/hw/xen/xen_pt.h +++ b/hw/xen/xen_pt.h @@ -319,7 +319,7 @@ static inline bool xen_pt_has_msix_mapping(XenPCIPassthroughState *s, int bar) } extern void *pci_assign_dev_load_option_rom(PCIDevice *dev, - struct Object *owner, int *size, + int *size, unsigned int domain, unsigned int bus, unsigned int slot, unsigned int function); diff --git a/hw/xen/xen_pt_graphics.c b/hw/xen/xen_pt_graphics.c index 0f4c8d77e2..135c8df1e7 100644 --- a/hw/xen/xen_pt_graphics.c +++ b/hw/xen/xen_pt_graphics.c @@ -132,7 +132,7 @@ int xen_pt_unregister_vga_regions(XenHostPCIDevice *dev) static void *get_vgabios(XenPCIPassthroughState *s, int *size, XenHostPCIDevice *dev) { - return pci_assign_dev_load_option_rom(&s->dev, OBJECT(&s->dev), size, + return pci_assign_dev_load_option_rom(&s->dev, size, dev->domain, dev->bus, dev->dev, dev->func); } diff --git a/hw/xen/xen_pt_load_rom.c b/hw/xen/xen_pt_load_rom.c index 71063c4d79..e6a86ca818 100644 --- a/hw/xen/xen_pt_load_rom.c +++ b/hw/xen/xen_pt_load_rom.c @@ -19,7 +19,7 @@ * load the corresponding ROM data to RAM. If an error occurs while loading an * option ROM, we just ignore that option ROM and continue with the next one. */ -void *pci_assign_dev_load_option_rom(PCIDevice *dev, struct Object *owner, +void *pci_assign_dev_load_option_rom(PCIDevice *dev, int *size, unsigned int domain, unsigned int bus, unsigned int slot, unsigned int function) @@ -29,6 +29,7 @@ void *pci_assign_dev_load_option_rom(PCIDevice *dev, struct Object *owner, uint8_t val; struct stat st; void *ptr = NULL; + Object *owner = OBJECT(dev); /* If loading ROM from file, pci handles it */ if (dev->romfile || !dev->rom_bar) { @@ -59,8 +60,7 @@ void *pci_assign_dev_load_option_rom(PCIDevice *dev, struct Object *owner, fseek(fp, 0, SEEK_SET); snprintf(name, sizeof(name), "%s.rom", object_get_typename(owner)); - memory_region_init_ram_nomigrate(&dev->rom, owner, name, st.st_size, &error_abort); - vmstate_register_ram(&dev->rom, &dev->qdev); + memory_region_init_ram(&dev->rom, owner, name, st.st_size, &error_abort); ptr = memory_region_get_ram_ptr(&dev->rom); memset(ptr, 0xff, st.st_size);
The xen pci_assign_dev_load_option_rom() currently creates a RAM memory region with memory_region_init_ram_nomigrate(), and then manually registers it with vmstate_register_ram(). In fact for its only callsite, the 'owner' pointer we use for the init call and the '&dev->qdev' pointer we use for the vmstate_register_ram() call refer to the same object. Simplify the function to only take a pointer to the device once instead of twice, and use memory_region_init_ram() which automatically does the vmstate register for us. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- This is a fairly trivial no-behaviour-change code cleanup, but I've marked it as RFC because I don't have a setup for doing more than just compile-testing Xen related patches. This was found as part of a sweep through for code using the _nomigrate versions of functions. hw/xen/xen_pt.h | 2 +- hw/xen/xen_pt_graphics.c | 2 +- hw/xen/xen_pt_load_rom.c | 6 +++--- 3 files changed, 5 insertions(+), 5 deletions(-) -- 2.17.1