[PATCH v4 2/2] soc: mediatek: add mtk-devapc driver
Chun-Kuang Hu
chunkuang.hu at kernel.org
Fri Jul 31 11:03:14 EDT 2020
Hi, Neal:
Neal Liu <neal.liu at mediatek.com> 於 2020年7月31日 週五 上午10:44寫道：
>
> Hi Chun-Kuang,
>
>
> On Thu, 2020-07-30 at 00:38 +0800, Chun-Kuang Hu wrote:
> > Hi, Neal:
> >
> > Neal Liu <neal.liu at mediatek.com> 於 2020年7月29日 週三 下午4:29寫道：
> > >
> > > MediaTek bus fabric provides TrustZone security support and data
> > > protection to prevent slaves from being accessed by unexpected
> > > masters.
> > > The security violation is logged and sent to the processor for
> > > further analysis or countermeasures.
> > >
> > > Any occurrence of security violation would raise an interrupt, and
> > > it will be handled by mtk-devapc driver. The violation
> > > information is printed in order to find the murderer.
> > >
> > > Signed-off-by: Neal Liu <neal.liu at mediatek.com>
> > > ---
> >
> > [snip]
> >
> > > +
> > > +static int get_shift_group(struct mtk_devapc_context *ctx)
> > > +{
> > > + u32 vio_shift_sta;
> > > + void __iomem *reg;
> > > +
> > > + reg = ctx->devapc_pd_base + ctx->offset->vio_shift_sta;
> > > + vio_shift_sta = readl(reg);
> > > +
> > > + if (vio_shift_sta)
> > > + return __ffs(vio_shift_sta);
> > > +
> > > + return -EIO;
> > > +}
> >
> > get_shift_group() is a small function, I would like to merge this
> > function into sync_vio_dbg() to make code more simple.
>
> This function have a specific functionality. And it would make this
> driver more readability. I would like to keep it as a function. Is that
> okay for you?
After merge, the function would be:
static bool sync_min_shift_group_vio_dbg(struct mtk_devapc_context *ctx)
{
int min_shift_group;
int ret;
u32 val;
/* find the minimum shift group which has violation */
val = readl(ctx->devapc_pd_base + ctx->offset->vio_shift_sta);
if (!val)
return false;
min_shift_group = __ffs(val);
/* Assign the group to sync */
writel(0x1 << min_shift_group, ctx->devapc_pd_base +
ctx->offset->vio_shift_sel);
/* Start syncing */
writel(0x1, ctx->devapc_pd_base + ctx->offset->vio_shift_con);
ret = readl_poll_timeout(pd_vio_shift_con_reg, val, val == 0x3, 0,
PHY_DEVAPC_TIMEOUT);
if (ret) {
dev_err(ctx->dev, "%s: Shift violation info failed\n", __func__);
return false;
}
/* Stop syncing */
writel(0x0, ctx->devapc_pd_base + ctx->offset->vio_shift_con);
/* ? */
writel(0x1 << min_shift_group, ctx->devapc_pd_base +
ctx->offset->vio_shift_sta);
return true;
}
The whole function is to sync min_shift_group violation info, I don't
know why separate any part to an independent function? Any function
call would cause penalty on CPU performance, so I does not like to
break this function. After good comment, I think every body could
understand the function of each register.
After the merge, the code would be so simple as:
while(sync_min_shift_group_vio_dbg(ctx))
devapc_extract_vio_dbg(ctx);
>
> >
> > > +
> >
> > [snip]
> >
> > > +
> > > +#define PHY_DEVAPC_TIMEOUT 0x10000
> > > +
> > > +/*
> > > + * sync_vio_dbg - do "shift" mechansim" to get full violation information.
> > > + * shift mechanism is depends on devapc hardware design.
> > > + * Mediatek devapc set multiple slaves as a group. When violation
> > > + * is triggered, violation info is kept inside devapc hardware.
> > > + * Driver should do shift mechansim to "shift" full violation
> > > + * info to VIO_DBGs registers.
> > > + *
> > > + */
> > > +static int sync_vio_dbg(struct mtk_devapc_context *ctx, u32 shift_bit)
> > > +{
> > > + void __iomem *pd_vio_shift_sta_reg;
> > > + void __iomem *pd_vio_shift_sel_reg;
> > > + void __iomem *pd_vio_shift_con_reg;
> > > + int ret;
> > > + u32 val;
> > > +
> > > + pd_vio_shift_sta_reg = ctx->devapc_pd_base + ctx->offset->vio_shift_sta;
> > > + pd_vio_shift_sel_reg = ctx->devapc_pd_base + ctx->offset->vio_shift_sel;
> > > + pd_vio_shift_con_reg = ctx->devapc_pd_base + ctx->offset->vio_shift_con;
> > > +
> > > + /* Enable shift mechansim */
> > > + writel(0x1 << shift_bit, pd_vio_shift_sel_reg);
> > > + writel(0x1, pd_vio_shift_con_reg);
> > > +
> > > + ret = readl_poll_timeout(pd_vio_shift_con_reg, val, val == 0x3, 0,
> > > + PHY_DEVAPC_TIMEOUT);
> > > + if (ret)
> > > + dev_err(ctx->dev, "%s: Shift violation info failed\n", __func__);
> > > +
> > > + /* Disable shift mechanism */
> > > + writel(0x0, pd_vio_shift_con_reg);
> > > + writel(0x0, pd_vio_shift_sel_reg);
> > > + writel(0x1 << shift_bit, pd_vio_shift_sta_reg);
> > > +
> > > + return ret;
> > > +}
> > > +
> >
> > [snip]
> >
> > > +
> > > +/*
> > > + * devapc_extract_vio_dbg - extract full violation information after doing
> > > + * shift mechanism.
> > > + */
> > > +static void devapc_extract_vio_dbg(struct mtk_devapc_context *ctx)
> > > +{
> > > + const struct mtk_devapc_vio_dbgs *vio_dbgs;
> > > + struct mtk_devapc_vio_info *vio_info;
> > > + void __iomem *vio_dbg0_reg;
> > > + void __iomem *vio_dbg1_reg;
> > > + u32 dbg0;
> > > +
> > > + vio_dbg0_reg = ctx->devapc_pd_base + ctx->offset->vio_dbg0;
> > > + vio_dbg1_reg = ctx->devapc_pd_base + ctx->offset->vio_dbg1;
> > > +
> > > + vio_dbgs = ctx->vio_dbgs;
> > > + vio_info = ctx->vio_info;
> > > +
> > > + /* Starts to extract violation information */
> > > + dbg0 = readl(vio_dbg0_reg);
> > > + vio_info->vio_addr = readl(vio_dbg1_reg);
> > > +
> > > + vio_info->master_id = (dbg0 & vio_dbgs->mstid.mask) >>
> > > + vio_dbgs->mstid.start;
> >
> > What is master_id? How could we use it to debug? For example, if we
> > get a master_id = 1, what should we do for this?
> >
> > > + vio_info->domain_id = (dbg0 & vio_dbgs->dmnid.mask) >>
> > > + vio_dbgs->dmnid.start;
> >
> > What is domain_id? How could we use it to debug? For example, if we
> > get a domain_id = 2, what should we do for this?
> >
>
> master_id and domain_id belongs our bus side-band signal info. It can
> help us to find the violation master.
Does 'violation master' means the hardware could access the protected
register? (ex. CPU, GCE, ...) If so, I think it's better to add
comment to explain how to map (master_id, domain_id) to a hardware
(maybe the device in device tree) because every body does not know
what the number means. Don't try to translate the number to a string
because this would cost much time to do this. Just print a number and
we could find out the master by the comment.
>
> > > + vio_info->write = ((dbg0 & vio_dbgs->vio_w.mask) >>
> > > + vio_dbgs->vio_w.start) == 1;
> > > + vio_info->read = ((dbg0 & vio_dbgs->vio_r.mask) >>
> > > + vio_dbgs->vio_r.start) == 1;
> > > + vio_info->vio_addr_high = (dbg0 & vio_dbgs->addr_h.mask) >>
> > > + vio_dbgs->addr_h.start;
> >
> > What is vio_addr_high? As I know all register address are 32 bits, is
> > vio_addr_high the address above 32 bits?
>
> Yes, you are right. In MT6779, all register base are 32 bits. We can
> ignore this info for current driver. I'll update on next patch.
> Thanks !
Such a strange hardware, all register is 32 bits but it has a
vio_addr_high in its register. OK, just drop this.
>
> >
> > > +
> > > + devapc_vio_info_print(ctx);
> > > +}
> > > +
> >
> > [snip]
> >
> > > +
> > > +/*
> > > + * devapc_violation_irq - the devapc Interrupt Service Routine (ISR) will dump
> > > + * violation information including which master violates
> > > + * access slave.
> > > + */
> > > +static irqreturn_t devapc_violation_irq(int irq_number,
> > > + struct mtk_devapc_context *ctx)
> > > +{
> > > + u32 vio_idx;
> > > +
> > > + /*
> > > + * Mask slave's irq before clearing vio status.
> > > + * Must do it to avoid nested interrupt and prevent
> > > + * unexpected behavior.
> > > + */
> > > + for (vio_idx = 0; vio_idx < ctx->vio_idx_num; vio_idx++)
> > > + mask_module_irq(ctx, vio_idx, true);
> >
> > I would like to rewrite this for-loop as below to prevent too many
> > function call in irq handler.
> >
> > for (i = 0; i < VIO_MOD_TO_REG_IND(ctx->vio_idx_num); i++)
> > writel(0xffffffff, ctx->devapc_pd_base + ctx->offset->vio_mask + 4 * i);
> >
>
> This idea is okay for me. Is there any macro to replace 0xffffffff?
GENMASK(31, 0);
>
> > reg = readl(ctx->devapc_pd_base + ctx->offset->vio_mask + 4 * i);
> > reg |= 1 << (ctx->vio_idx_num - 32 * i + 1) - 1;
> > writel(reg, ctx->devapc_pd_base + ctx->offset->vio_mask + 4 * i);
>
> Are you trying to clear the bits which over vio_idx_num?
> If yes, I think the second line should be:
> reg &= 1 << (ctx->vio_idx_num - 32 * i) - 1;
>
> For example, if vio_idx_num is 40:
> after for loop:
> vio_mask0 = 0xffffffff;
> vio_mask1 = 0xffffffff;
when vio_idx_num is 40, VIO_MOD_TO_REG_IND(ctx->vio_idx_num) is 1, so
after for-loop:
vio_mask0 = 0xffffffff;
And the code after for-loop just to do this:
vio_mask1 = 0xff;
>
> reg = readl(vio_mask1);
> reg &= (1 << 8) - 1; (which is 0x000000ff)
> reg will be 0xff
> writel(reg, vio_mask1);
>
> Does it make sense?
>
> Actually, it is okay to overwrite the unused register bits.
> So it's no matter to do this step.
OK, the code would be
for (i = 0; i <= VIO_MOD_TO_REG_IND(ctx->vio_idx_num - 1); i++)
writel(GENMASK(31, 0), ctx->devapc_pd_base + ctx->offset->vio_mask + 4 * i);
Regards,
Chun-Kuang.
>
> >
> > > +
> > > + devapc_dump_vio_dbg(ctx);
> > > +
> > > + /*
> > > + * Ensure that violation info are written
> > > + * before further operations
> > > + */
> > > + smp_mb();
> > > +
> > > + for (vio_idx = 0; vio_idx < ctx->vio_idx_num; vio_idx++) {
> > > + clear_vio_status(ctx, vio_idx);
> > > + mask_module_irq(ctx, vio_idx, false);
> > > + }
> >
> > Ditto for this for-loop.
>
> Ditto.
>
> >
> > > +
> > > + return IRQ_HANDLED;
> > > +}
> > > +
> >
> > [snip]
> >
> > > +
> > > +static int mtk_devapc_probe(struct platform_device *pdev)
> > > +{
> > > + struct device_node *node = pdev->dev.of_node;
> > > + struct mtk_devapc_context *ctx;
> > > + struct clk *devapc_infra_clk;
> > > + u32 devapc_irq;
> > > + int ret;
> > > +
> > > + if (IS_ERR(node))
> > > + return -ENODEV;
> > > +
> > > + ctx = devm_kzalloc(&pdev->dev, sizeof(*ctx), GFP_KERNEL);
> > > + if (!ctx)
> > > + return -ENOMEM;
> > > +
> > > + ctx = (struct mtk_devapc_context *)of_device_get_match_data(&pdev->dev);
> > > + ctx->dev = &pdev->dev;
> > > +
> > > + ctx->vio_info = devm_kzalloc(&pdev->dev,
> > > + sizeof(struct mtk_devapc_vio_info),
> > > + GFP_KERNEL);
> > > + if (!ctx->vio_info)
> > > + return -ENOMEM;
> > > +
> > > + ctx->devapc_pd_base = of_iomap(node, 0);
> > > + if (!ctx->devapc_pd_base)
> > > + return -EINVAL;
> > > +
> > > + devapc_irq = irq_of_parse_and_map(node, 0);
> > > + if (!devapc_irq)
> > > + return -EINVAL;
> > > +
> > > + devapc_infra_clk = devm_clk_get(&pdev->dev, "devapc-infra-clock");
> > > + if (IS_ERR(devapc_infra_clk))
> > > + return -EINVAL;
> > > +
> > > + if (clk_prepare_enable(devapc_infra_clk))
> > > + return -EINVAL;
> > > +
> > > + ret = devm_request_irq(&pdev->dev, devapc_irq,
> > > + (irq_handler_t)devapc_violation_irq,
> > > + IRQF_TRIGGER_NONE, "devapc", ctx);
> > > + if (ret)
> >
> > You should clk_disable_unprepare(devapc_infra_clk);
>
> Yes, I miss this part. Thanks for your remind.
> I'll update it on next patch.
>
> >
> > > + return ret;
> > > +
> > > + start_devapc(ctx);
> > > +
> > > + return 0;
> > > +}
> > > +
> > > +static int mtk_devapc_remove(struct platform_device *dev)
> > > +{
> >
> > Ditto.
>
> Ditto.
>
> >
> > Regards,
> > Chun-Kuang.
> >
> > > + return 0;
> > > +}
> > > +
> > > +static struct platform_driver mtk_devapc_driver = {
> > > + .probe = mtk_devapc_probe,
> > > + .remove = mtk_devapc_remove,
> > > + .driver = {
> > > + .name = KBUILD_MODNAME,
> > > + .of_match_table = mtk_devapc_dt_match,
> > > + },
> > > +};
> > > +
> > > +module_platform_driver(mtk_devapc_driver);
> > > +
>
More information about the Linux-mediatek
mailing list