| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <aKXQAoXZyR6SRPAA@linaro.org>
Date: Wed, 20 Aug 2025 15:39:14 +0200
From: Stephan Gerhold <stephan.gerhold@...aro.org>
To: Bjorn Andersson <andersson@...nel.org>,
Konrad Dybcio <konradybcio@...nel.org>,
Vikash Garodia <quic_vgarodia@...cinc.com>,
Dikshita Agarwal <quic_dikshita@...cinc.com>,
Mauro Carvalho Chehab <mchehab@...nel.org>,
Mathieu Poirier <mathieu.poirier@...aro.org>,
Abhinav Kumar <abhinav.kumar@...ux.dev>,
Bryan O'Donoghue <bryan.odonoghue@...aro.org>,
linux-kernel@...r.kernel.org, linux-arm-msm@...r.kernel.org,
linux-media@...r.kernel.org, linux-remoteproc@...r.kernel.org
Subject: Re: [PATCH v2 11/11] media: iris: Enable Secure PAS support with
IOMMU managed by Linux
On Wed, Aug 20, 2025 at 05:26:59PM +0530, Mukesh Ojha wrote:
> On Wed, Aug 20, 2025 at 10:46:31AM +0200, Stephan Gerhold wrote:
> > On Tue, Aug 19, 2025 at 10:24:46PM +0530, Mukesh Ojha wrote:
> > > Most Qualcomm platforms feature a proprietary hypervisor (such as Gunyah
> > > or QHEE), which typically handles IOMMU configuration. This includes
> > > mapping memory regions and device memory resources for remote processors
> > > by intercepting qcom_scm_pas_auth_and_reset() calls. These mappings are
> > > later removed during teardown. Additionally, SHM bridge setup is required
> > > to enable memory protection for both remoteproc metadata and its memory
> > > regions.
> > >
> > > When the hypervisor is absent, the operating system must perform these
> > > configurations instead.
> > >
> > > Support for handling IOMMU and SHM setup in the absence of a hypervisor
> > > is now in place. Extend the Iris driver to enable this functionality on
> > > platforms where IOMMU is managed by Linux (i.e., non-Gunyah, non-QHEE).
> > >
> > > Additionally, the Iris driver must map the firmware and its required
> > > resources to the firmware SID, which is now specified via the device tree.
> > >
> > > Co-developed-by: Vikash Garodia <quic_vgarodia@...cinc.com>
> > > Signed-off-by: Vikash Garodia <quic_vgarodia@...cinc.com>
> > > Signed-off-by: Mukesh Ojha <mukesh.ojha@....qualcomm.com>
> > > ---
> > > drivers/media/platform/qcom/iris/iris_core.c | 9 +-
> > > drivers/media/platform/qcom/iris/iris_core.h | 6 +
> > > .../media/platform/qcom/iris/iris_firmware.c | 156 ++++++++++++++++--
> > > .../media/platform/qcom/iris/iris_firmware.h | 2 +
> > > 4 files changed, 155 insertions(+), 18 deletions(-)
> > >
> > > [...]
> > > diff --git a/drivers/media/platform/qcom/iris/iris_firmware.c b/drivers/media/platform/qcom/iris/iris_firmware.c
> > > index f1b5cd56db32..e3f2fe5c9d7a 100644
> > > --- a/drivers/media/platform/qcom/iris/iris_firmware.c
> > > +++ b/drivers/media/platform/qcom/iris/iris_firmware.c
> > > @@ -3,10 +3,18 @@
> > > * Copyright (c) 2022-2024 Qualcomm Innovation Center, Inc. All rights reserved.
> > > */
> > >
> > > +#include <linux/device.h>
> > > #include <linux/firmware.h>
> > > -#include <linux/firmware/qcom/qcom_scm.h>
> > > +#include <linux/kernel.h>
> > > +#include <linux/iommu.h>
> > > +#include <linux/io.h>
> > > +#include <linux/of.h>
> > > #include <linux/of_address.h>
> > > #include <linux/of_reserved_mem.h>
> > > +#include <linux/platform_device.h>
> > > +#include <linux/of_device.h>
> > > +#include <linux/firmware/qcom/qcom_scm.h>
> > > +#include <linux/sizes.h>
> > > #include <linux/soc/qcom/mdt_loader.h>
> > >
> > > #include "iris_core.h"
> > > @@ -17,15 +25,14 @@
> > > static int iris_load_fw_to_memory(struct iris_core *core, const char *fw_name)
> > > {
> > > u32 pas_id = core->iris_platform_data->pas_id;
> > > + struct qcom_scm_pas_ctx *ctx;
> > > const struct firmware *firmware = NULL;
> > > struct device *dev = core->dev;
> > > - struct reserved_mem *rmem;
> > > - struct device_node *node;
> > > - phys_addr_t mem_phys;
> > > - size_t res_size;
> > > - ssize_t fw_size;
> > > - void *mem_virt;
> > > - int ret;
> > > + struct reserved_mem *rmem = NULL;
> > > + struct device_node *node = NULL;
> > > + ssize_t fw_size = 0;
> > > + void *mem_virt = NULL;
> > > + int ret = 0;
> > >
> > > if (strlen(fw_name) >= MAX_FIRMWARE_NAME_SIZE - 4)
> > > return -EINVAL;
> > > @@ -39,36 +46,64 @@ static int iris_load_fw_to_memory(struct iris_core *core, const char *fw_name)
> > > if (!rmem)
> > > return -EINVAL;
> > >
> > > - mem_phys = rmem->base;
> > > - res_size = rmem->size;
> > > + if (core->has_iommu)
> > > + dev = core->fw.dev;
> > >
> > > + ctx = qcom_scm_pas_ctx_init(dev, pas_id, rmem->base, rmem->size, false);
> > > + if (!ctx)
> > > + return -ENOMEM;
> > > +
> > > + ctx->has_iommu = core->has_iommu;
> > > ret = request_firmware(&firmware, fw_name, dev);
> > > if (ret)
> > > return ret;
> > >
> > > fw_size = qcom_mdt_get_size(firmware);
> > > - if (fw_size < 0 || res_size < (size_t)fw_size) {
> > > + if (fw_size < 0 || rmem->size < (size_t)fw_size) {
> > > ret = -EINVAL;
> > > goto err_release_fw;
> > > }
> > >
> > > - mem_virt = memremap(mem_phys, res_size, MEMREMAP_WC);
> > > + mem_virt = memremap(rmem->base, rmem->size, MEMREMAP_WC);
> > > if (!mem_virt) {
> > > ret = -ENOMEM;
> > > goto err_release_fw;
> > > }
> > >
> > > - ret = qcom_mdt_load(dev, firmware, fw_name,
> > > - pas_id, mem_virt, mem_phys, res_size, NULL);
> > > + ret = qcom_mdt_pas_load(ctx, firmware, fw_name, mem_virt, NULL);
> > > if (ret)
> > > goto err_mem_unmap;
> > >
> > > - ret = qcom_scm_pas_auth_and_reset(pas_id);
> > > + if (core->has_iommu) {
> > > + ret = iommu_map(core->fw.iommu_domain, 0, rmem->base, rmem->size,
> > > + IOMMU_READ | IOMMU_WRITE | IOMMU_PRIV, GFP_KERNEL);
> >
> > What is the use case for IOMMU_PRIV here? You don't have this flag for
> > the qcom_q6v5_pas change.
>
> This is there for historic regions, I may not have complete information about why
> is it required but the reference is taken from venus support for chrome.
>
Setting IOMMU_PRIV results in omitting the ARM_LPAE_PTE_AP_UNPRIV bit in
the IOMMU page tables - have you checked if QHEE sets this? Ideally we
want to do the same QHEE would normally do.
Also, please add a define for the 0 numbere here similar to
#define VENUS_FW_START_ADDR 0x0
It's quite hard to see that this is not an identity-mapping like for
qcom_q6v5_pas.
>
> > > + if (ret)
> > > + goto err_mem_unmap;
> > > +
> > > + /*
> > > + * Firmware has no support for resource table for now, so, lets
> > > + * pass NULL and zero for input resource table and input resource
> > > + * table respectively.
> > > + */
> > > + ret = qcom_mdt_pas_map_devmem_rscs(ctx, core->fw.iommu_domain, NULL, 0);
> > > + if (ret)
> > > + goto err_unmap_carveout;
> > > + }
> > > +
> > > + ret = qcom_scm_pas_prepare_and_auth_reset(ctx);
> > > if (ret)
> > > - goto err_mem_unmap;
> > > + goto err_unmap_devmem_rscs;
> > > +
> > > + core->fw.ctx = ctx;
> > >
> > > return ret;
> > >
> > > +err_unmap_devmem_rscs:
> > > + if (core->has_iommu)
> > > + qcom_mdt_pas_unmap_devmem_rscs(ctx, core->fw.iommu_domain);
> > > +err_unmap_carveout:
> > > + if (core->has_iommu)
> > > + iommu_unmap(core->fw.iommu_domain, 0, rmem->size);
> > > err_mem_unmap:
> > > memunmap(mem_virt);
> > > err_release_fw:
> > > @@ -109,10 +144,97 @@ int iris_fw_load(struct iris_core *core)
> > >
> > > int iris_fw_unload(struct iris_core *core)
> > > {
> > > - return qcom_scm_pas_shutdown(core->iris_platform_data->pas_id);
> > > + struct qcom_scm_pas_ctx *ctx;
> > > + int ret;
> > > +
> > > + ctx = core->fw.ctx;
> > > + ret = qcom_scm_pas_shutdown(ctx->peripheral);
> > > + if (core->has_iommu) {
> > > + iommu_unmap(core->fw.iommu_domain, 0, ctx->mem_size);
> > > + qcom_mdt_pas_unmap_devmem_rscs(ctx, core->fw.iommu_domain);
> > > + }
> > > +
> > > + return ret;
> > > }
> > >
> > > int iris_set_hw_state(struct iris_core *core, bool resume)
> > > {
> > > return qcom_scm_set_remote_state(resume, 0);
> > > }
> > > +
> > > +int iris_fw_init(struct iris_core *core)
> > > +{
> > > + struct platform_device_info info;
> > > + struct iommu_domain *iommu_dom;
> > > + struct platform_device *pdev;
> > > + struct device_node *np;
> > > + int ret;
> > > +
> > > + np = of_get_child_by_name(core->dev->of_node, "video-firmware");
> > > + if (!np)
> > > + return 0;
> >
> > You need a dt-bindings change for this as well. This is documented only
> > for Venus.
>
> You are right, wanted to send device tree and binding support separately.
> But if required, will add with the series in the next version.
>
You can send device tree changes separately, but dt-binding changes
always need to come before the driver changes.
Thanks,
Stephan
Powered by blists - more mailing lists