From: BALATON Zoltan via <qemu-devel@nongnu.org>
To: "Philippe Mathieu-Daudé" <f4bug@amsat.org>
Cc: Peter Maydell <peter.maydell@linaro.org>,
Aleksandar Rikalo <aleksandar.rikalo@syrmia.com>,
Qemu-block <qemu-block@nongnu.org>,
Huacai Chen <chenhuacai@kernel.org>,
Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>,
QEMU Developers <qemu-devel@nongnu.org>,
Wainer dos Santos Moschetta <wainersm@redhat.com>,
Artyom Tarasenko <atar4qemu@gmail.com>,
Cleber Rosa <crosa@redhat.com>, John Snow <jsnow@redhat.com>,
Aurelien Jarno <aurelien@aurel32.net>
Subject: Re: [RFC PATCH 3/5] hw/pci-host/bonito: Remap PCI "lo" regions when PCIMAP reg is modified
Date: Sat, 2 Jan 2021 12:22:24 +0100 (CET) [thread overview]
Message-ID: <293aa484-89c8-acc2-b9a3-37f17a506a2d@eik.bme.hu> (raw)
In-Reply-To: <2da14074-a4ef-e90c-ea42-74d48ca06afd@amsat.org>
[-- Attachment #1: Type: text/plain, Size: 4365 bytes --]
On Sat, 2 Jan 2021, Philippe Mathieu-Daudé wrote:
> On 1/2/21 12:19 AM, Peter Maydell wrote:
>> On Fri, 1 Jan 2021 at 23:12, Philippe Mathieu-Daudé <f4bug@amsat.org> wrote:
>>>
>>> Per the datasheet (Chapter 5.7.1. "PCI address regions"),
>>> the PCIMAP register:
>>>
>>> Map the 64Mbyte regions marked "PCI_Lo" in the CPU's memory map,
>>> each of which can be assigned to any 64 Mbyte-aligned region of
>>> PCI memory. The address appearing on the PCI bus consists of the
>>> low 26 bits of the CPU physical address, with the high 6 bits
>>> coming from the appropriate base6 field. Each of the three regions
>>> is an independent window onto PCI memory, and can be positioned on
>>> any 64Mbyte boundary in PCI space.
>>>
>>> Remap the 3 regions on reset and when PCIMAP is updated.
>>>
>>> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
>>> ---
>>> hw/pci-host/bonito.c | 49 ++++++++++++++++++++++++++++++++------------
>>> 1 file changed, 36 insertions(+), 13 deletions(-)
>>>
>>> diff --git a/hw/pci-host/bonito.c b/hw/pci-host/bonito.c
>>> index a99eced0657..c58eeaf504c 100644
>>> --- a/hw/pci-host/bonito.c
>>> +++ b/hw/pci-host/bonito.c
>>> @@ -137,6 +137,10 @@ FIELD(BONGENCFG, PCIQUEUE, 12, 1)
>>>
>>> /* 4. PCI address map control */
>>> #define BONITO_PCIMAP (0x10 >> 2) /* 0x110 */
>>> +FIELD(PCIMAP, LO0, 0, 6)
>>> +FIELD(PCIMAP, LO1, 6, 6)
>>> +FIELD(PCIMAP, LO2, 12, 6)
>>> +FIELD(PCIMAP, 2G, 18, 1)
>>> #define BONITO_PCIMEMBASECFG (0x14 >> 2) /* 0x114 */
>>> #define BONITO_PCIMAP_CFG (0x18 >> 2) /* 0x118 */
>>>
>>> @@ -237,6 +241,7 @@ struct BonitoState {
>>> qemu_irq *pic;
>>> PCIBonitoState *pci_dev;
>>> MemoryRegion pci_mem;
>>> + MemoryRegion pcimem_lo_alias[3];
>>> };
>>>
>>> #define TYPE_BONITO_PCI_HOST_BRIDGE "Bonito-pcihost"
>>> @@ -245,6 +250,31 @@ OBJECT_DECLARE_SIMPLE_TYPE(BonitoState, BONITO_PCI_HOST_BRIDGE)
>>> #define TYPE_PCI_BONITO "Bonito"
>>> OBJECT_DECLARE_SIMPLE_TYPE(PCIBonitoState, PCI_BONITO)
>>>
>>> +static void bonito_remap(PCIBonitoState *s)
>>> +{
>>> + static const char *const region_name[3] = {
>>> + "pci.lomem0", "pci.lomem1", "pci.lomem2"
>>> + };
>>> + BonitoState *bs = BONITO_PCI_HOST_BRIDGE(s->pcihost);
>>> +
>>> + for (size_t i = 0; i < 3; i++) {
>>> + uint32_t offset = extract32(s->regs[BONITO_PCIMAP], 6 * i, 6) << 26;
>>> +
>>> + if (memory_region_is_mapped(&bs->pcimem_lo_alias[i])) {
>>> + memory_region_del_subregion(get_system_memory(),
>>> + &bs->pcimem_lo_alias[i]);
>>> + object_unparent(OBJECT(&bs->pcimem_lo_alias[i]));
>>> + }
>>> +
>>> + memory_region_init_alias(&bs->pcimem_lo_alias[i], OBJECT(s),
>>> + region_name[i], &bs->pci_mem,
>>> + offset, 64 * MiB);
>>> + memory_region_add_subregion(get_system_memory(),
>>> + BONITO_PCILO_BASE + i * 64 * MiB,
>>> + &bs->pcimem_lo_alias[i]);
>>> + }
>>
>> Rather than delete-and-reinit-and-add, it's probably better to
>> just create the subregions once at device startup, and then use
>> memory_region_set_enabled() and memory_region_set_address()
>> to manipulate whether the subregion is visible and what address
>> in the system memory it is mapped at.
>
> Great! Thanks Peter :) TIL these functions.
> From what I understand from memory_region_readd_subregion (called
> from memory_region_set_address) using memory_region_set_enabled()
> directly is enough.
I have similar code in the series I've just posted where I'm mapping
regions of serial devices. I did consider using set_enabled and
set_address but ended up with removing and adding regions because I'm not
sure what happens if guest tries to move one region over another like
having one region at a default location while guest tries to map the other
one there (the pegasos2 maps serial at 0x2f8 which is normally COM2 on a
PC). This should not happen in theory but when removing disabled regions
it cannot happen so that looks safer therefore I chose to do that. Not
sure if this could be a problem here just shared my thughts about this.
Regards,
BALATON Zoltan
next prev parent reply other threads:[~2021-01-02 11:27 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-01 23:12 [RFC PATCH 0/5] hw/mips: Fix Fuloong2E to boot Linux guest again Philippe Mathieu-Daudé
2021-01-01 23:12 ` [RFC PATCH 1/5] ide: Make room for flags in PCIIDEState and add one for legacy mode Philippe Mathieu-Daudé
2021-01-01 23:12 ` [RFC PATCH 2/5] via-ide: Fix fuloong2e support Philippe Mathieu-Daudé
2021-01-03 15:14 ` Mark Cave-Ayland
2021-01-03 18:31 ` BALATON Zoltan via
2021-01-01 23:12 ` [RFC PATCH 3/5] hw/pci-host/bonito: Remap PCI "lo" regions when PCIMAP reg is modified Philippe Mathieu-Daudé
2021-01-01 23:19 ` Peter Maydell
2021-01-02 10:44 ` Philippe Mathieu-Daudé
2021-01-02 10:56 ` Philippe Mathieu-Daudé
2021-01-02 11:22 ` BALATON Zoltan via [this message]
2021-01-02 13:10 ` Peter Maydell
2021-01-02 14:12 ` BALATON Zoltan via
2021-01-01 23:12 ` [RFC PATCH 4/5] tests/acceptance: Test boot_linux_console for fuloong2e Philippe Mathieu-Daudé
2021-01-01 23:12 ` [RFC PATCH 5/5] tests/integration: Test Fuloong2E IDE drive, run userspace commands Philippe Mathieu-Daudé
2021-01-06 12:49 ` Willian Rampazzo
2021-01-01 23:56 ` [RFC PATCH 0/5] hw/mips: Fix Fuloong2E to boot Linux guest again BALATON Zoltan via
2021-01-03 14:27 ` Mark Cave-Ayland
2021-01-03 16:04 ` BALATON Zoltan via
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=293aa484-89c8-acc2-b9a3-37f17a506a2d@eik.bme.hu \
--to=qemu-devel@nongnu.org \
--cc=aleksandar.rikalo@syrmia.com \
--cc=atar4qemu@gmail.com \
--cc=aurelien@aurel32.net \
--cc=balaton@eik.bme.hu \
--cc=chenhuacai@kernel.org \
--cc=crosa@redhat.com \
--cc=f4bug@amsat.org \
--cc=jsnow@redhat.com \
--cc=mark.cave-ayland@ilande.co.uk \
--cc=peter.maydell@linaro.org \
--cc=qemu-block@nongnu.org \
--cc=wainersm@redhat.com \
/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).