Re: [PATCH v2 3/3] pci: ROM preallocation for incoming migration

[Michael S. Tsirkin]

On Wed, Apr 26, 2023 at 11:00:46PM +0300, Vladimir Sementsov-Ogievskiy wrote:
> On 26.04.23 07:43, Michael S. Tsirkin wrote:
> > On Tue, Apr 25, 2023 at 07:14:34PM +0300, Vladimir Sementsov-Ogievskiy 
> > wrote:
> > > On incoming migration we have the following sequence to load option
> > > ROM:
> > > 
> > > 1. On device realize we do normal load ROM from the file
> > > 
> > > 2. Than, on incoming migration we rewrite ROM from the incoming RAM
> > >     block. If sizes mismatch we fail.
> > 
> > let's mention an example error message:
> >   Size mismatch: 0000:00:03.0/virtio-net-pci.rom: 0x40000 != 0x80000: 
> > Invalid argument
> > 
> > 
> > > 
> > > This is not ideal when we migrate to updated distribution: we have to
> > > keep old ROM files in new distribution and be careful around romfile
> > > property to load correct ROM file.
> > 
> > > Which is loaded actually just to
> > > allocate the ROM with correct length.
> > > Note, that romsize property doesn't really help: if we try to specify
> > > it when default romfile is larger, it fails with something like:
> > > 
> > > romfile "efi-virtio.rom" (160768 bytes) is too large for ROM size 65536
> > 
> > Something I'd like to clarify is that the comment applies to uses where
> > users/distributions supply their own ROM file.  And lots of
> > users/distributions seem to have already painted themselves into a
> > corner by supplying a mix of ROM files of unmatching sizes -
> > basically they don't understand the detail of live migration,
> > ROM size interaction with it and with memory layout, etc -
> > as a very small number of people does.
> > For example, ubuntu doubled ROM file size by padding their ROMs
> > with 0xffffffff at some point, breaking migration for all existing machine
> > types.
> > 
> > just a web search for
> >   Size mismatch: 0000:00:03.0/virtio-net-pci.rom: 0x40000 != 0x80000: 
> > Invalid argument
> > 
> > will turn up a bunch of confused distros and users.
> > 
> > 
> > > 
> > > Let's just ignore ROM file when romsize is specified and we are in
> > > incoming migration state. In other words, we need only to preallocate
> > > ROM of specified size, local ROM file is unrelated.
> > 
> > 
> > 
> > 
> > > 
> > > This way:
> > > 
> > > If romsize was specified on source, we just use same commandline as on
> > > source, and migration will work independently of local ROM files on
> > > target.
> > > 
> > > If romsize was not specified on source (and we have mismatching local
> > > ROM file on target host), we have to specify romsize on target to match
> > > source romsize. romfile parameter may be kept same as on source or may
> > > be dropped, the file is not loaded anyway.
> > > 
> > > As a bonus we avoid extra reading from ROM file on target.
> > > 
> > > Note: when we don't have romsize parameter on source command line and
> > > need it for target, it may be calculated as aligned up to power of two
> > > size of ROM file on source (if we know, which file is it) or,
> > > alternatively it may be retrieved from source QEMU by QMP qom-get
> > > command, like
> > > 
> > >    { "execute": "qom-get",
> > >      "arguments": {
> > >        "path": "/machine/peripheral/CARD_ID/virtio-net-pci.rom[0]",
> > >        "property": "size" } }
> > > 
> > > Suggested-by: Michael S. Tsirkin <mst@redhat.com>
> > > Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
> > > ---
> > >   hw/pci/pci.c | 77 ++++++++++++++++++++++++++++++----------------------
> > >   1 file changed, 45 insertions(+), 32 deletions(-)
> > > 
> > > diff --git a/hw/pci/pci.c b/hw/pci/pci.c
> > > index a442f8fce1..e2cab622e4 100644
> > > --- a/hw/pci/pci.c
> > > +++ b/hw/pci/pci.c
> > > @@ -36,6 +36,7 @@
> > >   #include "migration/vmstate.h"
> > >   #include "net/net.h"
> > >   #include "sysemu/numa.h"
> > > +#include "sysemu/runstate.h"
> > >   #include "sysemu/sysemu.h"
> > >   #include "hw/loader.h"
> > >   #include "qemu/error-report.h"
> > > @@ -2293,10 +2294,16 @@ static void pci_add_option_rom(PCIDevice *pdev, 
> > > bool is_default_rom,
> > >   {
> > >       int64_t size;
> > >       g_autofree char *path = NULL;
> > > -    void *ptr;
> > >       char name[32];
> > >       const VMStateDescription *vmsd;
> > > +    /*
> > > +     * In case of incoming migration ROM will come with migration 
> > > stream, no
> > > +     * reason to load the file.  Neither we want to fail if local ROM 
> > > file
> > > +     * mismatches with specified romsize.
> > > +     */
> > > +    bool load_file = !runstate_check(RUN_STATE_INMIGRATE);
> > > +
> > >       if (!pdev->romfile) {
> > >           return;
> > >       }
> > 
> > CC pbonzini,dgilbert,quintela,armbru : guys, is poking at runstate_check 
> > like
> > this the right way to figure out we are not going to use the
> > device locally before incoming migration will overwrite ROM contents?
> 
> RUN_STATE_INMIGRATE is set in the only one place in qemu_init() when we parse 
> cmdline option -incoming. VM is not running for sure. And starting the VM 
> comes with changing the state. So it's OK.
> 
> The possible problem, if we add netcard on target which we didn't have on 
> source. I now checked, this works.. But that doesn't seem correct to add 
> device that was not present on source - how would it work - it's not 
> guaranteed anyway.

You can add it on source too while migration is in progress, no?

> > 
> > > @@ -2329,32 +2336,35 @@ static void pci_add_option_rom(PCIDevice *pdev, 
> > > bool is_default_rom,
> > >           return;
> > >       }
> > > -    path = qemu_find_file(QEMU_FILE_TYPE_BIOS, pdev->romfile);
> > > -    if (path == NULL) {
> > > -        path = g_strdup(pdev->romfile);
> > > -    }
> > > +    if (load_file || pdev->romsize == -1) {
> > > +        path = qemu_find_file(QEMU_FILE_TYPE_BIOS, pdev->romfile);
> > > +        if (path == NULL) {
> > > +            path = g_strdup(pdev->romfile);
> > > +        }
> > > -    size = get_image_size(path);
> > > -    if (size < 0) {
> > > -        error_setg(errp, "failed to find romfile \"%s\"", pdev->romfile);
> > > -        return;
> > > -    } else if (size == 0) {
> > > -        error_setg(errp, "romfile \"%s\" is empty", pdev->romfile);
> > > -        return;
> > > -    } else if (size > 2 * GiB) {
> > > -        error_setg(errp, "romfile \"%s\" too large (size cannot exceed 2 
> > > GiB)",
> > > -                   pdev->romfile);
> > > -        return;
> > > -    }
> > > -    if (pdev->romsize != -1) {
> > > -        if (size > pdev->romsize) {
> > > -            error_setg(errp, "romfile \"%s\" (%u bytes) "
> > > -                       "is too large for ROM size %u",
> > > -                       pdev->romfile, (uint32_t)size, pdev->romsize);
> > > +        size = get_image_size(path);
> > > +        if (size < 0) {
> > > +            error_setg(errp, "failed to find romfile \"%s\"", 
> > > pdev->romfile);
> > > +            return;
> > > +        } else if (size == 0) {
> > > +            error_setg(errp, "romfile \"%s\" is empty", pdev->romfile);
> > > +            return;
> > > +        } else if (size > 2 * GiB) {
> > > +            error_setg(errp,
> > > +                       "romfile \"%s\" too large (size cannot exceed 2 
> > > GiB)",
> > > +                       pdev->romfile);
> > >               return;
> > >           }
> > > -    } else {
> > > -        pdev->romsize = pow2ceil(size);
> > > +        if (pdev->romsize != -1) {
> > > +            if (size > pdev->romsize) {
> > > +                error_setg(errp, "romfile \"%s\" (%u bytes) "
> > > +                           "is too large for ROM size %u",
> > > +                           pdev->romfile, (uint32_t)size, pdev->romsize);
> > > +                return;
> > > +            }
> > > +        } else {
> > > +            pdev->romsize = pow2ceil(size);
> > > +        }
> > >       }
> > >       vmsd = qdev_get_vmsd(DEVICE(pdev));
> > > @@ -2365,15 +2375,18 @@ static void pci_add_option_rom(PCIDevice *pdev, 
> > > bool is_default_rom,
> > >       memory_region_init_rom(&pdev->rom, OBJECT(pdev), name, 
> > > pdev->romsize,
> > >                              &error_fatal);
> > > -    ptr = memory_region_get_ram_ptr(&pdev->rom);
> > > -    if (load_image_size(path, ptr, size) < 0) {
> > > -        error_setg(errp, "failed to load romfile \"%s\"", pdev->romfile);
> > > -        return;
> > > -    }
> > > +    if (load_file) {
> > > +        void *ptr = memory_region_get_ram_ptr(&pdev->rom);
> > > -    if (is_default_rom) {
> > > -        /* Only the default rom images will be patched (if needed). */
> > > -        pci_patch_ids(pdev, ptr, size);
> > > +        if (load_image_size(path, ptr, size) < 0) {
> > > +            error_setg(errp, "failed to load romfile \"%s\"", 
> > > pdev->romfile);
> > > +            return;
> > > +        }
> > > +
> > > +        if (is_default_rom) {
> > > +            /* Only the default rom images will be patched (if needed). 
> > > */
> > > +            pci_patch_ids(pdev, ptr, size);
> > > +        }
> > >       }
> > 
> > it kind of feels weird to ignore
> 
> What do you mean we should not ignore?
> 
> 
> 
> -- 
> Best regards,
> Vladimir