From: Alexey Kardashevskiy <aik@ozlabs.ru>
To: Daniel Henrique Barboza <danielhb413@gmail.com>, qemu-ppc@nongnu.org
Cc: qemu-devel@nongnu.org, Fabiano Rosas <farosas@linux.ibm.com>
Subject: Re: [PATCH qemu] spapr/ddw: Reset DMA when the last non-default window is removed
Date: Wed, 22 Jun 2022 12:26:57 +1000 [thread overview]
Message-ID: <90ca3cd7-c20b-0d08-9c2b-8abcc93a0620@ozlabs.ru> (raw)
In-Reply-To: <0493683a-47e8-ce26-0cd1-671928803f80@gmail.com>
On 6/21/22 23:54, Daniel Henrique Barboza wrote:
>
>
> On 6/20/22 05:01, Alexey Kardashevskiy wrote:
>> PAPR+/LoPAPR says:
>> ===
>> The platform must restore the default DMA window for the PE on a call
>> to the ibm,remove-pe-dma-window RTAS call when all of the following
>> are true:
>> a. The call removes the last DMA window remaining for the PE.
>> b. The DMA window being removed is not the default window
>>
>> ===
>>
>> This resets DMA as PAPR mandates.
>>
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>> ---
>
> Looks good. One comment below:
>
>
>> include/hw/ppc/spapr.h | 3 ++-
>> hw/ppc/spapr_iommu.c | 8 +++++---
>> hw/ppc/spapr_pci.c | 2 +-
>> hw/ppc/spapr_rtas_ddw.c | 17 ++++++++++++++++-
>> hw/ppc/spapr_vio.c | 3 ++-
>> 5 files changed, 26 insertions(+), 7 deletions(-)
>>
>> diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
>> index 072dda2c7265..0adbe1566d40 100644
>> --- a/include/hw/ppc/spapr.h
>> +++ b/include/hw/ppc/spapr.h
>> @@ -902,6 +902,7 @@ struct SpaprTceTable {
>> bool bypass;
>> bool need_vfio;
>> bool skipping_replay;
>> + bool def_win;
>> int fd;
>> MemoryRegion root;
>> IOMMUMemoryRegion iommu;
>> @@ -928,7 +929,7 @@ void spapr_check_mmu_mode(bool guest_radix);
>> SpaprTceTable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn);
>> void spapr_tce_table_enable(SpaprTceTable *tcet,
>> uint32_t page_shift, uint64_t bus_offset,
>> - uint32_t nb_table);
>> + uint32_t nb_table, bool def_win);
>
> Do we need to add 'def_win' to spapr_tce_table_enable()? I see that
> you're using
> the 'def_win' var to assign
>
> + tcet->def_win = def_win;
>
> but the only case where def_win will be other than 'false' is:
>
> - the function is called by spapr_tce_post_load(), but in that case it's
> being
> callied like this:
>
> spapr_tce_table_enable(tcet, old_page_shift, old_bus_offset,
> tcet->mig_nb_table, tcet->def_win);
>
> which results in the function doing tcet->def_win = tcet->def_win, which is
> uneeded.
>
>
> - the function is called by spapr_phb_dma_reset(). In this case I believe
> we can just do "tcet->def_win = true" before calling
> spapr_tce_table_enable():
>
>
>> /* Register default 32bit DMA window */
>> tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[0]);
>> + tcet->def_win = true;
>> spapr_tce_table_enable(tcet, SPAPR_TCE_PAGE_SHIFT,
>> sphb->dma_win_addr,
>> sphb->dma_win_size >> SPAPR_TCE_PAGE_SHIFT);
>> }
>
> All other calls to spapr_tce_table_enable() are passing 'false' to
> def_win and
> can be left alone.
>
> Assuming that there's no way a created DMA window becomes the default
> window, or
> the default DMA window is demoted to non-default, I think we can leave
> spapr_tce_table_enable() untouched and just set tcet->def_win to true
> inside
> spapr_phb_dma_reset().
I can definitely do this. Thanks for the review, I'll repost.
>
>
> Thanks,
>
>
> Daniel
>
>
>
>> void spapr_tce_table_disable(SpaprTceTable *tcet);
>> void spapr_tce_set_need_vfio(SpaprTceTable *tcet, bool need_vfio);
>> diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
>> index 81e5a1aea3a6..f8c1627d0782 100644
>> --- a/hw/ppc/spapr_iommu.c
>> +++ b/hw/ppc/spapr_iommu.c
>> @@ -242,7 +242,7 @@ static int spapr_tce_table_post_load(void *opaque,
>> int version_id)
>> if (tcet->mig_nb_table) {
>> if (!tcet->nb_table) {
>> spapr_tce_table_enable(tcet, old_page_shift,
>> old_bus_offset,
>> - tcet->mig_nb_table);
>> + tcet->mig_nb_table, tcet->def_win);
>> }
>> memcpy(tcet->table, tcet->mig_table,
>> @@ -279,7 +279,7 @@ static const VMStateDescription
>> vmstate_spapr_tce_table_ex = {
>> static const VMStateDescription vmstate_spapr_tce_table = {
>> .name = "spapr_iommu",
>> - .version_id = 2,
>> + .version_id = 3,
>> .minimum_version_id = 2,
>> .pre_save = spapr_tce_table_pre_save,
>> .post_load = spapr_tce_table_post_load,
>> @@ -292,6 +292,7 @@ static const VMStateDescription
>> vmstate_spapr_tce_table = {
>> VMSTATE_BOOL(bypass, SpaprTceTable),
>> VMSTATE_VARRAY_UINT32_ALLOC(mig_table, SpaprTceTable,
>> mig_nb_table, 0,
>> vmstate_info_uint64, uint64_t),
>> + VMSTATE_BOOL_V(def_win, SpaprTceTable, 3),
>> VMSTATE_END_OF_LIST()
>> },
>> @@ -380,7 +381,7 @@ SpaprTceTable *spapr_tce_new_table(DeviceState
>> *owner, uint32_t liobn)
>> void spapr_tce_table_enable(SpaprTceTable *tcet,
>> uint32_t page_shift, uint64_t bus_offset,
>> - uint32_t nb_table)
>> + uint32_t nb_table, bool def_win)
>> {
>> if (tcet->nb_table) {
>> warn_report("trying to enable already enabled TCE table");
>> @@ -390,6 +391,7 @@ void spapr_tce_table_enable(SpaprTceTable *tcet,
>> tcet->bus_offset = bus_offset;
>> tcet->page_shift = page_shift;
>> tcet->nb_table = nb_table;
>> + tcet->def_win = def_win;
>> tcet->table = spapr_tce_alloc_table(tcet->liobn,
>> tcet->page_shift,
>> tcet->bus_offset,
>> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
>> index b2f5fbef0c83..e1dbccfc7547 100644
>> --- a/hw/ppc/spapr_pci.c
>> +++ b/hw/ppc/spapr_pci.c
>> @@ -2066,7 +2066,7 @@ void spapr_phb_dma_reset(SpaprPhbState *sphb)
>> /* Register default 32bit DMA window */
>> tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[0]);
>> spapr_tce_table_enable(tcet, SPAPR_TCE_PAGE_SHIFT,
>> sphb->dma_win_addr,
>> - sphb->dma_win_size >> SPAPR_TCE_PAGE_SHIFT);
>> + sphb->dma_win_size >>
>> SPAPR_TCE_PAGE_SHIFT, true);
>> }
>> static void spapr_phb_reset(DeviceState *qdev)
>> diff --git a/hw/ppc/spapr_rtas_ddw.c b/hw/ppc/spapr_rtas_ddw.c
>> index 13d339c807c1..4fe41b0c4539 100644
>> --- a/hw/ppc/spapr_rtas_ddw.c
>> +++ b/hw/ppc/spapr_rtas_ddw.c
>> @@ -182,7 +182,7 @@ static void
>> rtas_ibm_create_pe_dma_window(PowerPCCPU *cpu,
>> */
>> tcet->skipping_replay = true;
>> spapr_tce_table_enable(tcet, page_shift, win_addr,
>> - 1ULL << (window_shift - page_shift));
>> + 1ULL << (window_shift - page_shift), false);
>> tcet->skipping_replay = false;
>> if (!tcet->nb_table) {
>> goto hw_error_exit;
>> @@ -215,6 +215,7 @@ static void
>> rtas_ibm_remove_pe_dma_window(PowerPCCPU *cpu,
>> SpaprPhbState *sphb;
>> SpaprTceTable *tcet;
>> uint32_t liobn;
>> + bool def_win_removed;
>> if ((nargs != 1) || (nret != 1)) {
>> goto param_error_exit;
>> @@ -231,9 +232,23 @@ static void
>> rtas_ibm_remove_pe_dma_window(PowerPCCPU *cpu,
>> goto param_error_exit;
>> }
>> + def_win_removed = tcet->def_win;
>> spapr_tce_table_disable(tcet);
>> trace_spapr_iommu_ddw_remove(liobn);
>> + /*
>> + * PAPR+/LoPAPR says:
>> + * The platform must restore the default DMA window for the PE on
>> a call
>> + * to the ibm,remove-pe-dma-window RTAS call when all of the
>> following
>> + * are true:
>> + * a. The call removes the last DMA window remaining for the PE.
>> + * b. The DMA window being removed is not the default window
>> + */
>> + if (spapr_phb_get_active_win_num(sphb) == 0 && !def_win_removed) {
>> + spapr_phb_dma_reset(sphb);
>> + trace_spapr_iommu_ddw_reset(sphb->buid, 0);
>> + }
>> +
>> rtas_st(rets, 0, RTAS_OUT_SUCCESS);
>> return;
>> diff --git a/hw/ppc/spapr_vio.c b/hw/ppc/spapr_vio.c
>> index 9d4fec2c04d8..14506df19d62 100644
>> --- a/hw/ppc/spapr_vio.c
>> +++ b/hw/ppc/spapr_vio.c
>> @@ -533,7 +533,8 @@ static void spapr_vio_busdev_realize(DeviceState
>> *qdev, Error **errp)
>> dev->tcet = spapr_tce_new_table(qdev, liobn);
>> spapr_tce_table_enable(dev->tcet, SPAPR_TCE_PAGE_SHIFT, 0,
>> - pc->rtce_window_size >>
>> SPAPR_TCE_PAGE_SHIFT);
>> + pc->rtce_window_size >>
>> SPAPR_TCE_PAGE_SHIFT,
>> + false);
>> dev->tcet->vdev = dev;
>> memory_region_add_subregion_overlap(&dev->mrroot, 0,
>>
>> spapr_tce_get_iommu(dev->tcet), 2);
--
Alexey
prev parent reply other threads:[~2022-06-22 2:29 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-06-20 8:01 [PATCH qemu] spapr/ddw: Reset DMA when the last non-default window is removed Alexey Kardashevskiy
2022-06-21 13:54 ` Daniel Henrique Barboza
2022-06-22 2:26 ` Alexey Kardashevskiy [this message]
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=90ca3cd7-c20b-0d08-9c2b-8abcc93a0620@ozlabs.ru \
--to=aik@ozlabs.ru \
--cc=danielhb413@gmail.com \
--cc=farosas@linux.ibm.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-ppc@nongnu.org \
/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: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).