From: "Michael S. Tsirkin" <mst@redhat.com> To: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro> Cc: joro@8bytes.org, blauwirbel@gmail.com, paul@codesourcery.com, avi@redhat.com, anthony@codemonkey.ws, av1474@comtv.ru, yamahata@valinux.co.jp, kvm@vger.kernel.org, qemu-devel@nongnu.org Subject: Re: [PATCH 4/7] ide: use the PCI memory access interface Date: Thu, 2 Sep 2010 12:58:13 +0300 [thread overview] Message-ID: <20100902095813.GB9085@redhat.com> (raw) In-Reply-To: <20100902091200.GB7319@localhost> On Thu, Sep 02, 2010 at 12:12:00PM +0300, Eduard - Gabriel Munteanu wrote: > On Thu, Sep 02, 2010 at 08:19:11AM +0300, Michael S. Tsirkin wrote: > > On Sat, Aug 28, 2010 at 05:54:55PM +0300, Eduard - Gabriel Munteanu wrote: > > > Emulated PCI IDE controllers now use the memory access interface. This > > > also allows an emulated IOMMU to translate and check accesses. > > > > > > Map invalidation results in cancelling DMA transfers. Since the guest OS > > > can't properly recover the DMA results in case the mapping is changed, > > > this is a fairly good approximation. > > > > > > Signed-off-by: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro> > > > --- > > [snip] > > > > +static inline void bmdma_memory_read(BMDMAState *bm, > > > + target_phys_addr_t addr, > > > + uint8_t *buf, > > > + target_phys_addr_t len) > > > +{ > > > + bm->rw(bm->opaque, addr, buf, len, 0); > > > +} > > > + > > > +static inline void bmdma_memory_write(BMDMAState *bm, > > > + target_phys_addr_t addr, > > > + uint8_t *buf, > > > + target_phys_addr_t len) > > > +{ > > > + bm->rw(bm->opaque, addr, buf, len, 1); > > > +} > > > + > > > > Here again, I am concerned about indirection and pointer chaising on data path. > > Can we have an iommu pointer in the device, and do a fast path in case > > there is no iommu? > > > > See my other reply. I don't insist on this solution, but what other way do you propose to avoid the overhead for everyone not using an iommu? I'm all for a solution that would help iommu as well, but one wasn't yet proposed. > > > static inline IDEState *idebus_active_if(IDEBus *bus) > > > { > > > return bus->ifs + bus->unit; > > > diff --git a/hw/ide/macio.c b/hw/ide/macio.c > > > index bd1c73e..962ae13 100644 > > > --- a/hw/ide/macio.c > > > +++ b/hw/ide/macio.c > > > @@ -79,7 +79,7 @@ static void pmac_ide_atapi_transfer_cb(void *opaque, int ret) > > > > > > s->io_buffer_size = io->len; > > > > > > - qemu_sglist_init(&s->sg, io->len / MACIO_PAGE_SIZE + 1); > > > + qemu_sglist_init(&s->sg, io->len / MACIO_PAGE_SIZE + 1, NULL, NULL, NULL); > > > qemu_sglist_add(&s->sg, io->addr, io->len); > > > io->addr += io->len; > > > io->len = 0; > > > @@ -141,7 +141,7 @@ static void pmac_ide_transfer_cb(void *opaque, int ret) > > > s->io_buffer_index = 0; > > > s->io_buffer_size = io->len; > > > > > > - qemu_sglist_init(&s->sg, io->len / MACIO_PAGE_SIZE + 1); > > > + qemu_sglist_init(&s->sg, io->len / MACIO_PAGE_SIZE + 1, NULL, NULL, NULL); > > > qemu_sglist_add(&s->sg, io->addr, io->len); > > > io->addr += io->len; > > > io->len = 0; > > > diff --git a/hw/ide/pci.c b/hw/ide/pci.c > > > index 4d95cc5..5879044 100644 > > > --- a/hw/ide/pci.c > > > +++ b/hw/ide/pci.c > > > @@ -183,4 +183,11 @@ void pci_ide_create_devs(PCIDevice *dev, DriveInfo **hd_table) > > > continue; > > > ide_create_drive(d->bus+bus[i], unit[i], hd_table[i]); > > > } > > > + > > > + for (i = 0; i < 2; i++) { > > > + d->bmdma[i].rw = (void *) pci_memory_rw; > > > + d->bmdma[i].map = (void *) pci_memory_map; > > > + d->bmdma[i].unmap = (void *) pci_memory_unmap; > > > + d->bmdma[i].opaque = dev; > > > + } > > > } > > > > These casts show something is wrong with the API, IMO. > > > > Hm, here's an oversight on my part: I think I should provide explicit > bmdma hooks, since pcibus_t is a uint64_t and target_phys_addr_t is a > uint{32,64}_t depending on the guest machine, so it might be buggy on > 32-bit wrt calling conventions. But that introduces yet another > non-inlined function call :-(. That would drop the (void *) cast, > though. > > > Eduard So we get away with it without casts but only because C compiler will let us silently convert the types, possibly discarding data in the process. Or we'll add a check that will try and detect this, but there's no good way to report a DMA error to user. IOW, if our code only works because target fits in pcibus, what good is the abstraction and using distinct types? This is why I think we need a generic DMA APIs using dma addresses. -- MST
WARNING: multiple messages have this Message-ID (diff)
From: "Michael S. Tsirkin" <mst@redhat.com> To: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro> Cc: kvm@vger.kernel.org, joro@8bytes.org, qemu-devel@nongnu.org, blauwirbel@gmail.com, yamahata@valinux.co.jp, paul@codesourcery.com, avi@redhat.com Subject: [Qemu-devel] Re: [PATCH 4/7] ide: use the PCI memory access interface Date: Thu, 2 Sep 2010 12:58:13 +0300 [thread overview] Message-ID: <20100902095813.GB9085@redhat.com> (raw) In-Reply-To: <20100902091200.GB7319@localhost> On Thu, Sep 02, 2010 at 12:12:00PM +0300, Eduard - Gabriel Munteanu wrote: > On Thu, Sep 02, 2010 at 08:19:11AM +0300, Michael S. Tsirkin wrote: > > On Sat, Aug 28, 2010 at 05:54:55PM +0300, Eduard - Gabriel Munteanu wrote: > > > Emulated PCI IDE controllers now use the memory access interface. This > > > also allows an emulated IOMMU to translate and check accesses. > > > > > > Map invalidation results in cancelling DMA transfers. Since the guest OS > > > can't properly recover the DMA results in case the mapping is changed, > > > this is a fairly good approximation. > > > > > > Signed-off-by: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro> > > > --- > > [snip] > > > > +static inline void bmdma_memory_read(BMDMAState *bm, > > > + target_phys_addr_t addr, > > > + uint8_t *buf, > > > + target_phys_addr_t len) > > > +{ > > > + bm->rw(bm->opaque, addr, buf, len, 0); > > > +} > > > + > > > +static inline void bmdma_memory_write(BMDMAState *bm, > > > + target_phys_addr_t addr, > > > + uint8_t *buf, > > > + target_phys_addr_t len) > > > +{ > > > + bm->rw(bm->opaque, addr, buf, len, 1); > > > +} > > > + > > > > Here again, I am concerned about indirection and pointer chaising on data path. > > Can we have an iommu pointer in the device, and do a fast path in case > > there is no iommu? > > > > See my other reply. I don't insist on this solution, but what other way do you propose to avoid the overhead for everyone not using an iommu? I'm all for a solution that would help iommu as well, but one wasn't yet proposed. > > > static inline IDEState *idebus_active_if(IDEBus *bus) > > > { > > > return bus->ifs + bus->unit; > > > diff --git a/hw/ide/macio.c b/hw/ide/macio.c > > > index bd1c73e..962ae13 100644 > > > --- a/hw/ide/macio.c > > > +++ b/hw/ide/macio.c > > > @@ -79,7 +79,7 @@ static void pmac_ide_atapi_transfer_cb(void *opaque, int ret) > > > > > > s->io_buffer_size = io->len; > > > > > > - qemu_sglist_init(&s->sg, io->len / MACIO_PAGE_SIZE + 1); > > > + qemu_sglist_init(&s->sg, io->len / MACIO_PAGE_SIZE + 1, NULL, NULL, NULL); > > > qemu_sglist_add(&s->sg, io->addr, io->len); > > > io->addr += io->len; > > > io->len = 0; > > > @@ -141,7 +141,7 @@ static void pmac_ide_transfer_cb(void *opaque, int ret) > > > s->io_buffer_index = 0; > > > s->io_buffer_size = io->len; > > > > > > - qemu_sglist_init(&s->sg, io->len / MACIO_PAGE_SIZE + 1); > > > + qemu_sglist_init(&s->sg, io->len / MACIO_PAGE_SIZE + 1, NULL, NULL, NULL); > > > qemu_sglist_add(&s->sg, io->addr, io->len); > > > io->addr += io->len; > > > io->len = 0; > > > diff --git a/hw/ide/pci.c b/hw/ide/pci.c > > > index 4d95cc5..5879044 100644 > > > --- a/hw/ide/pci.c > > > +++ b/hw/ide/pci.c > > > @@ -183,4 +183,11 @@ void pci_ide_create_devs(PCIDevice *dev, DriveInfo **hd_table) > > > continue; > > > ide_create_drive(d->bus+bus[i], unit[i], hd_table[i]); > > > } > > > + > > > + for (i = 0; i < 2; i++) { > > > + d->bmdma[i].rw = (void *) pci_memory_rw; > > > + d->bmdma[i].map = (void *) pci_memory_map; > > > + d->bmdma[i].unmap = (void *) pci_memory_unmap; > > > + d->bmdma[i].opaque = dev; > > > + } > > > } > > > > These casts show something is wrong with the API, IMO. > > > > Hm, here's an oversight on my part: I think I should provide explicit > bmdma hooks, since pcibus_t is a uint64_t and target_phys_addr_t is a > uint{32,64}_t depending on the guest machine, so it might be buggy on > 32-bit wrt calling conventions. But that introduces yet another > non-inlined function call :-(. That would drop the (void *) cast, > though. > > > Eduard So we get away with it without casts but only because C compiler will let us silently convert the types, possibly discarding data in the process. Or we'll add a check that will try and detect this, but there's no good way to report a DMA error to user. IOW, if our code only works because target fits in pcibus, what good is the abstraction and using distinct types? This is why I think we need a generic DMA APIs using dma addresses. -- MST
next prev parent reply other threads:[~2010-09-02 10:04 UTC|newest] Thread overview: 97+ messages / expand[flat|nested] mbox.gz Atom feed top 2010-08-28 14:54 [PATCH 0/7] AMD IOMMU emulation patchset v4 Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [PATCH 1/7] pci: expand tabs to spaces in pci_regs.h Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-08-31 20:29 ` Michael S. Tsirkin 2010-08-31 20:29 ` [Qemu-devel] " Michael S. Tsirkin 2010-08-31 22:58 ` Eduard - Gabriel Munteanu 2010-08-31 22:58 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-09-01 10:39 ` Michael S. Tsirkin 2010-09-01 10:39 ` [Qemu-devel] " Michael S. Tsirkin 2010-08-28 14:54 ` [PATCH 2/7] pci: memory access API and IOMMU support Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-09-02 5:28 ` Michael S. Tsirkin 2010-09-02 5:28 ` [Qemu-devel] " Michael S. Tsirkin 2010-09-02 8:40 ` Eduard - Gabriel Munteanu 2010-09-02 8:40 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-09-02 9:49 ` Michael S. Tsirkin 2010-09-02 9:49 ` [Qemu-devel] " Michael S. Tsirkin 2010-09-04 9:01 ` Blue Swirl 2010-09-04 9:01 ` [Qemu-devel] " Blue Swirl 2010-09-05 7:10 ` Michael S. Tsirkin 2010-09-05 7:10 ` [Qemu-devel] " Michael S. Tsirkin 2010-08-28 14:54 ` [PATCH 3/7] AMD IOMMU emulation Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-08-28 15:58 ` Blue Swirl 2010-08-28 15:58 ` [Qemu-devel] " Blue Swirl 2010-08-28 21:53 ` Eduard - Gabriel Munteanu 2010-08-28 21:53 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-08-29 20:37 ` Blue Swirl 2010-08-29 20:37 ` [Qemu-devel] " Blue Swirl 2010-08-30 3:07 ` [Qemu-devel] " Isaku Yamahata 2010-08-30 3:07 ` Isaku Yamahata 2010-08-30 5:54 ` Eduard - Gabriel Munteanu 2010-08-30 5:54 ` Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [PATCH 4/7] ide: use the PCI memory access interface Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-09-02 5:19 ` Michael S. Tsirkin 2010-09-02 5:19 ` [Qemu-devel] " Michael S. Tsirkin 2010-09-02 9:12 ` Eduard - Gabriel Munteanu 2010-09-02 9:12 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-09-02 9:58 ` Michael S. Tsirkin [this message] 2010-09-02 9:58 ` Michael S. Tsirkin 2010-09-02 15:01 ` Eduard - Gabriel Munteanu 2010-09-02 15:01 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-09-02 15:24 ` Avi Kivity 2010-09-02 15:24 ` [Qemu-devel] " Avi Kivity 2010-09-02 15:39 ` Michael S. Tsirkin 2010-09-02 15:39 ` [Qemu-devel] " Michael S. Tsirkin 2010-09-02 16:07 ` Avi Kivity 2010-09-02 16:07 ` [Qemu-devel] " Avi Kivity 2010-09-02 15:31 ` Michael S. Tsirkin 2010-09-02 15:31 ` [Qemu-devel] " Michael S. Tsirkin 2010-08-28 14:54 ` [PATCH 5/7] rtl8139: " Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [PATCH 6/7] eepro100: " Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [PATCH 7/7] ac97: " Eduard - Gabriel Munteanu 2010-08-28 14:54 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-08-28 16:00 ` [PATCH 0/7] AMD IOMMU emulation patchset v4 Blue Swirl 2010-08-28 16:00 ` [Qemu-devel] " Blue Swirl 2010-08-29 9:55 ` Joerg Roedel 2010-08-29 9:55 ` [Qemu-devel] " Joerg Roedel 2010-08-29 20:44 ` Blue Swirl 2010-08-29 20:44 ` [Qemu-devel] " Blue Swirl 2010-08-29 22:08 ` [PATCH 2/7] pci: memory access API and IOMMU support Eduard - Gabriel Munteanu 2010-08-29 22:08 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-08-29 22:11 ` Eduard - Gabriel Munteanu 2010-08-29 22:11 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-09-01 20:10 ` [Qemu-devel] " Stefan Weil 2010-09-01 20:10 ` Stefan Weil 2010-09-02 6:00 ` Michael S. Tsirkin 2010-09-02 6:00 ` Michael S. Tsirkin 2010-09-02 9:08 ` Eduard - Gabriel Munteanu 2010-09-02 9:08 ` Eduard - Gabriel Munteanu 2010-09-02 13:24 ` Anthony Liguori 2010-09-02 13:24 ` Anthony Liguori 2010-09-02 8:51 ` Eduard - Gabriel Munteanu 2010-09-02 8:51 ` Eduard - Gabriel Munteanu 2010-09-02 16:05 ` Stefan Weil 2010-09-02 16:05 ` Stefan Weil 2010-09-02 16:14 ` Eduard - Gabriel Munteanu 2010-09-02 16:14 ` Eduard - Gabriel Munteanu 2010-09-13 20:01 ` [PATCH RFC] dma_rw.h (was Re: [PATCH 0/7] AMD IOMMU emulation patchset v4) Michael S. Tsirkin 2010-09-13 20:01 ` [Qemu-devel] " Michael S. Tsirkin 2010-09-13 20:45 ` Anthony Liguori 2010-09-13 20:45 ` Anthony Liguori 2010-09-16 7:12 ` Eduard - Gabriel Munteanu 2010-09-16 7:12 ` Eduard - Gabriel Munteanu 2010-09-16 9:35 ` Michael S. Tsirkin 2010-09-16 9:35 ` Michael S. Tsirkin 2010-09-16 7:06 ` Eduard - Gabriel Munteanu 2010-09-16 7:06 ` [Qemu-devel] " Eduard - Gabriel Munteanu 2010-09-16 9:20 ` Michael S. Tsirkin 2010-09-16 9:20 ` [Qemu-devel] " Michael S. Tsirkin 2010-09-16 11:15 ` Eduard - Gabriel Munteanu 2010-09-16 11:15 ` [Qemu-devel] " Eduard - Gabriel Munteanu -- strict thread matches above, loose matches on Subject: below -- 2010-08-15 19:27 [PATCH 0/7] AMD IOMMU emulation patches v3 Eduard - Gabriel Munteanu 2010-08-15 19:27 ` [PATCH 4/7] ide: use the PCI memory access interface Eduard - Gabriel Munteanu
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20100902095813.GB9085@redhat.com \ --to=mst@redhat.com \ --cc=anthony@codemonkey.ws \ --cc=av1474@comtv.ru \ --cc=avi@redhat.com \ --cc=blauwirbel@gmail.com \ --cc=eduard.munteanu@linux360.ro \ --cc=joro@8bytes.org \ --cc=kvm@vger.kernel.org \ --cc=paul@codesourcery.com \ --cc=qemu-devel@nongnu.org \ --cc=yamahata@valinux.co.jp \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.