OpenCloudOS-Kernel/drivers/dma/ti/cppi41.c

1258 lines
30 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
#include <linux/delay.h>
#include <linux/dmaengine.h>
#include <linux/dma-mapping.h>
#include <linux/platform_device.h>
#include <linux/module.h>
#include <linux/of.h>
#include <linux/slab.h>
#include <linux/of_dma.h>
#include <linux/of_irq.h>
#include <linux/dmapool.h>
#include <linux/interrupt.h>
#include <linux/of_address.h>
#include <linux/pm_runtime.h>
#include "../dmaengine.h"
#define DESC_TYPE 27
#define DESC_TYPE_HOST 0x10
#define DESC_TYPE_TEARD 0x13
#define TD_DESC_IS_RX (1 << 16)
#define TD_DESC_DMA_NUM 10
#define DESC_LENGTH_BITS_NUM 21
#define DESC_TYPE_USB (5 << 26)
#define DESC_PD_COMPLETE (1 << 31)
/* DMA engine */
#define DMA_TDFDQ 4
#define DMA_TXGCR(x) (0x800 + (x) * 0x20)
#define DMA_RXGCR(x) (0x808 + (x) * 0x20)
#define RXHPCRA0 4
#define GCR_CHAN_ENABLE (1 << 31)
#define GCR_TEARDOWN (1 << 30)
#define GCR_STARV_RETRY (1 << 24)
#define GCR_DESC_TYPE_HOST (1 << 14)
/* DMA scheduler */
#define DMA_SCHED_CTRL 0
#define DMA_SCHED_CTRL_EN (1 << 31)
#define DMA_SCHED_WORD(x) ((x) * 4 + 0x800)
#define SCHED_ENTRY0_CHAN(x) ((x) << 0)
#define SCHED_ENTRY0_IS_RX (1 << 7)
#define SCHED_ENTRY1_CHAN(x) ((x) << 8)
#define SCHED_ENTRY1_IS_RX (1 << 15)
#define SCHED_ENTRY2_CHAN(x) ((x) << 16)
#define SCHED_ENTRY2_IS_RX (1 << 23)
#define SCHED_ENTRY3_CHAN(x) ((x) << 24)
#define SCHED_ENTRY3_IS_RX (1 << 31)
/* Queue manager */
/* 4 KiB of memory for descriptors, 2 for each endpoint */
#define ALLOC_DECS_NUM 128
#define DESCS_AREAS 1
#define TOTAL_DESCS_NUM (ALLOC_DECS_NUM * DESCS_AREAS)
#define QMGR_SCRATCH_SIZE (TOTAL_DESCS_NUM * 4)
#define QMGR_LRAM0_BASE 0x80
#define QMGR_LRAM_SIZE 0x84
#define QMGR_LRAM1_BASE 0x88
#define QMGR_MEMBASE(x) (0x1000 + (x) * 0x10)
#define QMGR_MEMCTRL(x) (0x1004 + (x) * 0x10)
#define QMGR_MEMCTRL_IDX_SH 16
#define QMGR_MEMCTRL_DESC_SH 8
#define QMGR_PEND(x) (0x90 + (x) * 4)
#define QMGR_PENDING_SLOT_Q(x) (x / 32)
#define QMGR_PENDING_BIT_Q(x) (x % 32)
#define QMGR_QUEUE_A(n) (0x2000 + (n) * 0x10)
#define QMGR_QUEUE_B(n) (0x2004 + (n) * 0x10)
#define QMGR_QUEUE_C(n) (0x2008 + (n) * 0x10)
#define QMGR_QUEUE_D(n) (0x200c + (n) * 0x10)
/* Packet Descriptor */
#define PD2_ZERO_LENGTH (1 << 19)
struct cppi41_channel {
struct dma_chan chan;
struct dma_async_tx_descriptor txd;
struct cppi41_dd *cdd;
struct cppi41_desc *desc;
dma_addr_t desc_phys;
void __iomem *gcr_reg;
int is_tx;
u32 residue;
unsigned int q_num;
unsigned int q_comp_num;
unsigned int port_num;
unsigned td_retry;
unsigned td_queued:1;
unsigned td_seen:1;
unsigned td_desc_seen:1;
struct list_head node; /* Node for pending list */
};
struct cppi41_desc {
u32 pd0;
u32 pd1;
u32 pd2;
u32 pd3;
u32 pd4;
u32 pd5;
u32 pd6;
u32 pd7;
} __aligned(32);
struct chan_queues {
u16 submit;
u16 complete;
};
struct cppi41_dd {
struct dma_device ddev;
void *qmgr_scratch;
dma_addr_t scratch_phys;
struct cppi41_desc *cd;
dma_addr_t descs_phys;
u32 first_td_desc;
struct cppi41_channel *chan_busy[ALLOC_DECS_NUM];
void __iomem *ctrl_mem;
void __iomem *sched_mem;
void __iomem *qmgr_mem;
unsigned int irq;
const struct chan_queues *queues_rx;
const struct chan_queues *queues_tx;
struct chan_queues td_queue;
u16 first_completion_queue;
u16 qmgr_num_pend;
u32 n_chans;
u8 platform;
struct list_head pending; /* Pending queued transfers */
spinlock_t lock; /* Lock for pending list */
/* context for suspend/resume */
unsigned int dma_tdfdq;
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
bool is_suspended;
};
static struct chan_queues am335x_usb_queues_tx[] = {
/* USB0 ENDP 1 */
[ 0] = { .submit = 32, .complete = 93},
[ 1] = { .submit = 34, .complete = 94},
[ 2] = { .submit = 36, .complete = 95},
[ 3] = { .submit = 38, .complete = 96},
[ 4] = { .submit = 40, .complete = 97},
[ 5] = { .submit = 42, .complete = 98},
[ 6] = { .submit = 44, .complete = 99},
[ 7] = { .submit = 46, .complete = 100},
[ 8] = { .submit = 48, .complete = 101},
[ 9] = { .submit = 50, .complete = 102},
[10] = { .submit = 52, .complete = 103},
[11] = { .submit = 54, .complete = 104},
[12] = { .submit = 56, .complete = 105},
[13] = { .submit = 58, .complete = 106},
[14] = { .submit = 60, .complete = 107},
/* USB1 ENDP1 */
[15] = { .submit = 62, .complete = 125},
[16] = { .submit = 64, .complete = 126},
[17] = { .submit = 66, .complete = 127},
[18] = { .submit = 68, .complete = 128},
[19] = { .submit = 70, .complete = 129},
[20] = { .submit = 72, .complete = 130},
[21] = { .submit = 74, .complete = 131},
[22] = { .submit = 76, .complete = 132},
[23] = { .submit = 78, .complete = 133},
[24] = { .submit = 80, .complete = 134},
[25] = { .submit = 82, .complete = 135},
[26] = { .submit = 84, .complete = 136},
[27] = { .submit = 86, .complete = 137},
[28] = { .submit = 88, .complete = 138},
[29] = { .submit = 90, .complete = 139},
};
static const struct chan_queues am335x_usb_queues_rx[] = {
/* USB0 ENDP 1 */
[ 0] = { .submit = 1, .complete = 109},
[ 1] = { .submit = 2, .complete = 110},
[ 2] = { .submit = 3, .complete = 111},
[ 3] = { .submit = 4, .complete = 112},
[ 4] = { .submit = 5, .complete = 113},
[ 5] = { .submit = 6, .complete = 114},
[ 6] = { .submit = 7, .complete = 115},
[ 7] = { .submit = 8, .complete = 116},
[ 8] = { .submit = 9, .complete = 117},
[ 9] = { .submit = 10, .complete = 118},
[10] = { .submit = 11, .complete = 119},
[11] = { .submit = 12, .complete = 120},
[12] = { .submit = 13, .complete = 121},
[13] = { .submit = 14, .complete = 122},
[14] = { .submit = 15, .complete = 123},
/* USB1 ENDP 1 */
[15] = { .submit = 16, .complete = 141},
[16] = { .submit = 17, .complete = 142},
[17] = { .submit = 18, .complete = 143},
[18] = { .submit = 19, .complete = 144},
[19] = { .submit = 20, .complete = 145},
[20] = { .submit = 21, .complete = 146},
[21] = { .submit = 22, .complete = 147},
[22] = { .submit = 23, .complete = 148},
[23] = { .submit = 24, .complete = 149},
[24] = { .submit = 25, .complete = 150},
[25] = { .submit = 26, .complete = 151},
[26] = { .submit = 27, .complete = 152},
[27] = { .submit = 28, .complete = 153},
[28] = { .submit = 29, .complete = 154},
[29] = { .submit = 30, .complete = 155},
};
static const struct chan_queues da8xx_usb_queues_tx[] = {
[0] = { .submit = 16, .complete = 24},
[1] = { .submit = 18, .complete = 24},
[2] = { .submit = 20, .complete = 24},
[3] = { .submit = 22, .complete = 24},
};
static const struct chan_queues da8xx_usb_queues_rx[] = {
[0] = { .submit = 1, .complete = 26},
[1] = { .submit = 3, .complete = 26},
[2] = { .submit = 5, .complete = 26},
[3] = { .submit = 7, .complete = 26},
};
struct cppi_glue_infos {
const struct chan_queues *queues_rx;
const struct chan_queues *queues_tx;
struct chan_queues td_queue;
u16 first_completion_queue;
u16 qmgr_num_pend;
};
static struct cppi41_channel *to_cpp41_chan(struct dma_chan *c)
{
return container_of(c, struct cppi41_channel, chan);
}
static struct cppi41_channel *desc_to_chan(struct cppi41_dd *cdd, u32 desc)
{
struct cppi41_channel *c;
u32 descs_size;
u32 desc_num;
descs_size = sizeof(struct cppi41_desc) * ALLOC_DECS_NUM;
if (!((desc >= cdd->descs_phys) &&
(desc < (cdd->descs_phys + descs_size)))) {
return NULL;
}
desc_num = (desc - cdd->descs_phys) / sizeof(struct cppi41_desc);
BUG_ON(desc_num >= ALLOC_DECS_NUM);
c = cdd->chan_busy[desc_num];
cdd->chan_busy[desc_num] = NULL;
dmaengine: cppi41: Fix runtime PM timeouts with USB mass storage Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") added runtime PM support for cppi41, but had corner case issues. Some of the issues were fixed with commit 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when only a USB hub is connected"). That fix however caused a new regression where we can get error -115 messages with USB on BeagleBone when connecting a USB mass storage device to a hub. This is because when connecting a USB mass storage device to a hub, the initial DMA transfers can take over 200ms to complete and cppi41 autosuspend delay times out. To fix the issue, we want to implement refcounting for chan_busy array that contains the active dma transfers. Increasing the autosuspend delay won't help as that the delay could be potentially seconds, and it's best to let the USB subsystem to deal with the timeouts on errors. The earlier attempt for runtime PM was buggy as the pm_runtime_get/put() calls could get unpaired easily as they did not follow the state of the chan_busy array as described in commit 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when only a USB hub is connected". Let's fix the issue by adding pm_runtime_get() to where a new transfer is added to the chan_busy array, and calls to pm_runtime_put() where chan_busy array entry is cleared. This prevents any autosuspend timeouts from happening while dma transfers are active. Fixes: 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when only a USB hub is connected") Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:07 +08:00
/* Usecount for chan_busy[], paired with push_desc_queue() */
pm_runtime_put(cdd->ddev.dev);
return c;
}
static void cppi_writel(u32 val, void *__iomem *mem)
{
__raw_writel(val, mem);
}
static u32 cppi_readl(void *__iomem *mem)
{
return __raw_readl(mem);
}
static u32 pd_trans_len(u32 val)
{
return val & ((1 << (DESC_LENGTH_BITS_NUM + 1)) - 1);
}
static u32 cppi41_pop_desc(struct cppi41_dd *cdd, unsigned queue_num)
{
u32 desc;
desc = cppi_readl(cdd->qmgr_mem + QMGR_QUEUE_D(queue_num));
desc &= ~0x1f;
return desc;
}
static irqreturn_t cppi41_irq(int irq, void *data)
{
struct cppi41_dd *cdd = data;
u16 first_completion_queue = cdd->first_completion_queue;
u16 qmgr_num_pend = cdd->qmgr_num_pend;
struct cppi41_channel *c;
int i;
for (i = QMGR_PENDING_SLOT_Q(first_completion_queue); i < qmgr_num_pend;
i++) {
u32 val;
u32 q_num;
val = cppi_readl(cdd->qmgr_mem + QMGR_PEND(i));
if (i == QMGR_PENDING_SLOT_Q(first_completion_queue) && val) {
u32 mask;
/* set corresponding bit for completetion Q 93 */
mask = 1 << QMGR_PENDING_BIT_Q(first_completion_queue);
/* not set all bits for queues less than Q 93 */
mask--;
/* now invert and keep only Q 93+ set */
val &= ~mask;
}
if (val)
__iormb();
while (val) {
u32 desc, len;
/*
* This should never trigger, see the comments in
* push_desc_queue()
*/
WARN_ON(cdd->is_suspended);
q_num = __fls(val);
val &= ~(1 << q_num);
q_num += 32 * i;
desc = cppi41_pop_desc(cdd, q_num);
c = desc_to_chan(cdd, desc);
if (WARN_ON(!c)) {
pr_err("%s() q %d desc %08x\n", __func__,
q_num, desc);
continue;
}
if (c->desc->pd2 & PD2_ZERO_LENGTH)
len = 0;
else
len = pd_trans_len(c->desc->pd0);
c->residue = pd_trans_len(c->desc->pd6) - len;
dma_cookie_complete(&c->txd);
dmaengine_desc_get_callback_invoke(&c->txd, NULL);
}
}
return IRQ_HANDLED;
}
static dma_cookie_t cppi41_tx_submit(struct dma_async_tx_descriptor *tx)
{
dma_cookie_t cookie;
cookie = dma_cookie_assign(tx);
return cookie;
}
static int cppi41_dma_alloc_chan_resources(struct dma_chan *chan)
{
struct cppi41_channel *c = to_cpp41_chan(chan);
struct cppi41_dd *cdd = c->cdd;
int error;
error = pm_runtime_get_sync(cdd->ddev.dev);
if (error < 0) {
dev_err(cdd->ddev.dev, "%s pm runtime get: %i\n",
__func__, error);
pm_runtime_put_noidle(cdd->ddev.dev);
return error;
}
dma_cookie_init(chan);
dma_async_tx_descriptor_init(&c->txd, chan);
c->txd.tx_submit = cppi41_tx_submit;
if (!c->is_tx)
cppi_writel(c->q_num, c->gcr_reg + RXHPCRA0);
pm_runtime_mark_last_busy(cdd->ddev.dev);
pm_runtime_put_autosuspend(cdd->ddev.dev);
return 0;
}
static void cppi41_dma_free_chan_resources(struct dma_chan *chan)
{
struct cppi41_channel *c = to_cpp41_chan(chan);
struct cppi41_dd *cdd = c->cdd;
int error;
error = pm_runtime_get_sync(cdd->ddev.dev);
if (error < 0) {
pm_runtime_put_noidle(cdd->ddev.dev);
return;
}
WARN_ON(!list_empty(&cdd->pending));
pm_runtime_mark_last_busy(cdd->ddev.dev);
pm_runtime_put_autosuspend(cdd->ddev.dev);
}
static enum dma_status cppi41_dma_tx_status(struct dma_chan *chan,
dma_cookie_t cookie, struct dma_tx_state *txstate)
{
struct cppi41_channel *c = to_cpp41_chan(chan);
enum dma_status ret;
ret = dma_cookie_status(chan, cookie, txstate);
dma_set_residue(txstate, c->residue);
return ret;
}
static void push_desc_queue(struct cppi41_channel *c)
{
struct cppi41_dd *cdd = c->cdd;
u32 desc_num;
u32 desc_phys;
u32 reg;
c->residue = 0;
reg = GCR_CHAN_ENABLE;
if (!c->is_tx) {
reg |= GCR_STARV_RETRY;
reg |= GCR_DESC_TYPE_HOST;
reg |= c->q_comp_num;
}
cppi_writel(reg, c->gcr_reg);
/*
* We don't use writel() but __raw_writel() so we have to make sure
* that the DMA descriptor in coherent memory made to the main memory
* before starting the dma engine.
*/
__iowmb();
dmaengine: cppi41: Fix runtime PM timeouts with USB mass storage Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") added runtime PM support for cppi41, but had corner case issues. Some of the issues were fixed with commit 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when only a USB hub is connected"). That fix however caused a new regression where we can get error -115 messages with USB on BeagleBone when connecting a USB mass storage device to a hub. This is because when connecting a USB mass storage device to a hub, the initial DMA transfers can take over 200ms to complete and cppi41 autosuspend delay times out. To fix the issue, we want to implement refcounting for chan_busy array that contains the active dma transfers. Increasing the autosuspend delay won't help as that the delay could be potentially seconds, and it's best to let the USB subsystem to deal with the timeouts on errors. The earlier attempt for runtime PM was buggy as the pm_runtime_get/put() calls could get unpaired easily as they did not follow the state of the chan_busy array as described in commit 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when only a USB hub is connected". Let's fix the issue by adding pm_runtime_get() to where a new transfer is added to the chan_busy array, and calls to pm_runtime_put() where chan_busy array entry is cleared. This prevents any autosuspend timeouts from happening while dma transfers are active. Fixes: 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when only a USB hub is connected") Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:07 +08:00
/*
* DMA transfers can take at least 200ms to complete with USB mass
* storage connected. To prevent autosuspend timeouts, we must use
* pm_runtime_get/put() when chan_busy[] is modified. This will get
* cleared in desc_to_chan() or cppi41_stop_chan() depending on the
* outcome of the transfer.
*/
pm_runtime_get(cdd->ddev.dev);
desc_phys = lower_32_bits(c->desc_phys);
desc_num = (desc_phys - cdd->descs_phys) / sizeof(struct cppi41_desc);
WARN_ON(cdd->chan_busy[desc_num]);
cdd->chan_busy[desc_num] = c;
reg = (sizeof(struct cppi41_desc) - 24) / 4;
reg |= desc_phys;
cppi_writel(reg, cdd->qmgr_mem + QMGR_QUEUE_D(c->q_num));
}
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
/*
* Caller must hold cdd->lock to prevent push_desc_queue()
* getting called out of order. We have both cppi41_dma_issue_pending()
* and cppi41_runtime_resume() call this function.
*/
static void cppi41_run_queue(struct cppi41_dd *cdd)
{
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
struct cppi41_channel *c, *_c;
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
list_for_each_entry_safe(c, _c, &cdd->pending, node) {
push_desc_queue(c);
list_del(&c->node);
}
}
static void cppi41_dma_issue_pending(struct dma_chan *chan)
{
struct cppi41_channel *c = to_cpp41_chan(chan);
struct cppi41_dd *cdd = c->cdd;
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
unsigned long flags;
int error;
error = pm_runtime_get(cdd->ddev.dev);
if ((error != -EINPROGRESS) && error < 0) {
pm_runtime_put_noidle(cdd->ddev.dev);
dev_err(cdd->ddev.dev, "Failed to pm_runtime_get: %i\n",
error);
return;
}
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
spin_lock_irqsave(&cdd->lock, flags);
list_add_tail(&c->node, &cdd->pending);
if (!cdd->is_suspended)
cppi41_run_queue(cdd);
spin_unlock_irqrestore(&cdd->lock, flags);
pm_runtime_mark_last_busy(cdd->ddev.dev);
pm_runtime_put_autosuspend(cdd->ddev.dev);
}
static u32 get_host_pd0(u32 length)
{
u32 reg;
reg = DESC_TYPE_HOST << DESC_TYPE;
reg |= length;
return reg;
}
static u32 get_host_pd1(struct cppi41_channel *c)
{
u32 reg;
reg = 0;
return reg;
}
static u32 get_host_pd2(struct cppi41_channel *c)
{
u32 reg;
reg = DESC_TYPE_USB;
reg |= c->q_comp_num;
return reg;
}
static u32 get_host_pd3(u32 length)
{
u32 reg;
/* PD3 = packet size */
reg = length;
return reg;
}
static u32 get_host_pd6(u32 length)
{
u32 reg;
/* PD6 buffer size */
reg = DESC_PD_COMPLETE;
reg |= length;
return reg;
}
static u32 get_host_pd4_or_7(u32 addr)
{
u32 reg;
reg = addr;
return reg;
}
static u32 get_host_pd5(void)
{
u32 reg;
reg = 0;
return reg;
}
static struct dma_async_tx_descriptor *cppi41_dma_prep_slave_sg(
struct dma_chan *chan, struct scatterlist *sgl, unsigned sg_len,
enum dma_transfer_direction dir, unsigned long tx_flags, void *context)
{
struct cppi41_channel *c = to_cpp41_chan(chan);
struct dma_async_tx_descriptor *txd = NULL;
struct cppi41_dd *cdd = c->cdd;
struct cppi41_desc *d;
struct scatterlist *sg;
unsigned int i;
int error;
error = pm_runtime_get(cdd->ddev.dev);
if (error < 0) {
pm_runtime_put_noidle(cdd->ddev.dev);
return NULL;
}
if (cdd->is_suspended)
goto err_out_not_ready;
d = c->desc;
for_each_sg(sgl, sg, sg_len, i) {
u32 addr;
u32 len;
/* We need to use more than one desc once musb supports sg */
addr = lower_32_bits(sg_dma_address(sg));
len = sg_dma_len(sg);
d->pd0 = get_host_pd0(len);
d->pd1 = get_host_pd1(c);
d->pd2 = get_host_pd2(c);
d->pd3 = get_host_pd3(len);
d->pd4 = get_host_pd4_or_7(addr);
d->pd5 = get_host_pd5();
d->pd6 = get_host_pd6(len);
d->pd7 = get_host_pd4_or_7(addr);
d++;
}
txd = &c->txd;
err_out_not_ready:
pm_runtime_mark_last_busy(cdd->ddev.dev);
pm_runtime_put_autosuspend(cdd->ddev.dev);
return txd;
}
static void cppi41_compute_td_desc(struct cppi41_desc *d)
{
d->pd0 = DESC_TYPE_TEARD << DESC_TYPE;
}
static int cppi41_tear_down_chan(struct cppi41_channel *c)
{
struct dmaengine_result abort_result;
struct cppi41_dd *cdd = c->cdd;
struct cppi41_desc *td;
u32 reg;
u32 desc_phys;
u32 td_desc_phys;
td = cdd->cd;
td += cdd->first_td_desc;
td_desc_phys = cdd->descs_phys;
td_desc_phys += cdd->first_td_desc * sizeof(struct cppi41_desc);
if (!c->td_queued) {
cppi41_compute_td_desc(td);
__iowmb();
reg = (sizeof(struct cppi41_desc) - 24) / 4;
reg |= td_desc_phys;
cppi_writel(reg, cdd->qmgr_mem +
QMGR_QUEUE_D(cdd->td_queue.submit));
reg = GCR_CHAN_ENABLE;
if (!c->is_tx) {
reg |= GCR_STARV_RETRY;
reg |= GCR_DESC_TYPE_HOST;
reg |= cdd->td_queue.complete;
}
reg |= GCR_TEARDOWN;
cppi_writel(reg, c->gcr_reg);
c->td_queued = 1;
dma: cppi41: wait longer for the HW to return the descriptor For a "complete" teardown we have to wait until the teardown descriptor is returned by the hardware. The g_zero testcase "testusb -a -t 9" triggers the following warning quite reliable: |------------[ cut here ]------------ |WARNING: CPU: 0 PID: 0 at drivers/dma/cppi41.c:609 cppi41_dma_control+0x198/0x304() |[<c003f84c>] (warn_slowpath_null) from [<c02be8d8>] |[<c02be8d8>] (cppi41_dma_control) from [<bf08d25c>] |[<bf08d25c>] (cppi41_dma_channel_abort [musb_hdrc]) |[<bf08bc38>] (nuke.constprop.10 [musb_hdrc]) |[<bf08bd08>] (musb_gadget_disable [musb_hdrc]) |[<bf252524>] (disable_endpoints [usb_f_ss_lb]) |[<bf2525d8>] (disable_source_sink [usb_f_ss_lb]) |[<bf25260c>] (sourcesink_set_alt [usb_f_ss_lb]) |[<bf23ad24>] (composite_setup [libcomposite]) |[<bf08a2f4>] (musb_g_ep0_irq [musb_hdrc]) |[<bf085ec4>] (musb_interrupt [musb_hdrc]) |[<bf0aeaf4>] (dsps_interrupt [musb_dsps]) |[<c0080ea8>] (handle_irq_event_percpu) |[<c008112c>] (handle_irq_event) |[<c008348c>] (handle_level_irq) |[<c00807a8>] (generic_handle_irq) |[<c000ee80>] (handle_IRQ) |[<c00085f0>] (omap3_intc_handle_irq) and complains about a TD descriptor which is not returned. I've been looking at several things and haven't noticed anything unusual that might lead to this. The manual says "to try again" until the descriptor comes out. I limited the amount of retries to 100 retries in order to avoid an infinite number of retries and so a busy-loop. Back then testing revealed that the number of retries were around 20-30 so 100 seemed a good upper limit. This g_zero test reaches without a problem 98 retries and it jumps sometimes to 101 on am335x-evm and so the WARN_ON() triggers. Same test run on beaglebone black and the retries start at 122 and my max value so far was at 128. So lets rise the limit to 500. Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-12-03 22:09:49 +08:00
c->td_retry = 500;
}
if (!c->td_seen || !c->td_desc_seen) {
desc_phys = cppi41_pop_desc(cdd, cdd->td_queue.complete);
if (!desc_phys && c->is_tx)
desc_phys = cppi41_pop_desc(cdd, c->q_comp_num);
if (desc_phys == c->desc_phys) {
c->td_desc_seen = 1;
} else if (desc_phys == td_desc_phys) {
u32 pd0;
__iormb();
pd0 = td->pd0;
WARN_ON((pd0 >> DESC_TYPE) != DESC_TYPE_TEARD);
WARN_ON(!c->is_tx && !(pd0 & TD_DESC_IS_RX));
WARN_ON((pd0 & 0x1f) != c->port_num);
c->td_seen = 1;
} else if (desc_phys) {
WARN_ON_ONCE(1);
}
}
c->td_retry--;
/*
* If the TX descriptor / channel is in use, the caller needs to poke
* his TD bit multiple times. After that he hardware releases the
* transfer descriptor followed by TD descriptor. Waiting seems not to
* cause any difference.
* RX seems to be thrown out right away. However once the TearDown
* descriptor gets through we are done. If we have seens the transfer
* descriptor before the TD we fetch it from enqueue, it has to be
* there waiting for us.
*/
if (!c->td_seen && c->td_retry) {
udelay(1);
return -EAGAIN;
}
WARN_ON(!c->td_retry);
if (!c->td_desc_seen) {
desc_phys = cppi41_pop_desc(cdd, c->q_num);
if (!desc_phys)
desc_phys = cppi41_pop_desc(cdd, c->q_comp_num);
WARN_ON(!desc_phys);
}
c->td_queued = 0;
c->td_seen = 0;
c->td_desc_seen = 0;
cppi_writel(0, c->gcr_reg);
/* Invoke the callback to do the necessary clean-up */
abort_result.result = DMA_TRANS_ABORTED;
dma_cookie_complete(&c->txd);
dmaengine_desc_get_callback_invoke(&c->txd, &abort_result);
return 0;
}
static int cppi41_stop_chan(struct dma_chan *chan)
{
struct cppi41_channel *c = to_cpp41_chan(chan);
struct cppi41_dd *cdd = c->cdd;
u32 desc_num;
u32 desc_phys;
int ret;
desc_phys = lower_32_bits(c->desc_phys);
desc_num = (desc_phys - cdd->descs_phys) / sizeof(struct cppi41_desc);
if (!cdd->chan_busy[desc_num]) {
struct cppi41_channel *cc, *_ct;
/*
* channels might still be in the pendling list if
* cppi41_dma_issue_pending() is called after
* cppi41_runtime_suspend() is called
*/
list_for_each_entry_safe(cc, _ct, &cdd->pending, node) {
if (cc != c)
continue;
list_del(&cc->node);
break;
}
return 0;
}
ret = cppi41_tear_down_chan(c);
if (ret)
return ret;
WARN_ON(!cdd->chan_busy[desc_num]);
cdd->chan_busy[desc_num] = NULL;
dmaengine: cppi41: Fix runtime PM timeouts with USB mass storage Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") added runtime PM support for cppi41, but had corner case issues. Some of the issues were fixed with commit 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when only a USB hub is connected"). That fix however caused a new regression where we can get error -115 messages with USB on BeagleBone when connecting a USB mass storage device to a hub. This is because when connecting a USB mass storage device to a hub, the initial DMA transfers can take over 200ms to complete and cppi41 autosuspend delay times out. To fix the issue, we want to implement refcounting for chan_busy array that contains the active dma transfers. Increasing the autosuspend delay won't help as that the delay could be potentially seconds, and it's best to let the USB subsystem to deal with the timeouts on errors. The earlier attempt for runtime PM was buggy as the pm_runtime_get/put() calls could get unpaired easily as they did not follow the state of the chan_busy array as described in commit 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when only a USB hub is connected". Let's fix the issue by adding pm_runtime_get() to where a new transfer is added to the chan_busy array, and calls to pm_runtime_put() where chan_busy array entry is cleared. This prevents any autosuspend timeouts from happening while dma transfers are active. Fixes: 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when only a USB hub is connected") Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:07 +08:00
/* Usecount for chan_busy[], paired with push_desc_queue() */
pm_runtime_put(cdd->ddev.dev);
return 0;
}
static int cppi41_add_chans(struct device *dev, struct cppi41_dd *cdd)
{
struct cppi41_channel *cchan, *chans;
int i;
u32 n_chans = cdd->n_chans;
/*
* The channels can only be used as TX or as RX. So we add twice
* that much dma channels because USB can only do RX or TX.
*/
n_chans *= 2;
chans = devm_kcalloc(dev, n_chans, sizeof(*chans), GFP_KERNEL);
if (!chans)
return -ENOMEM;
for (i = 0; i < n_chans; i++) {
cchan = &chans[i];
cchan->cdd = cdd;
if (i & 1) {
cchan->gcr_reg = cdd->ctrl_mem + DMA_TXGCR(i >> 1);
cchan->is_tx = 1;
} else {
cchan->gcr_reg = cdd->ctrl_mem + DMA_RXGCR(i >> 1);
cchan->is_tx = 0;
}
cchan->port_num = i >> 1;
cchan->desc = &cdd->cd[i];
cchan->desc_phys = cdd->descs_phys;
cchan->desc_phys += i * sizeof(struct cppi41_desc);
cchan->chan.device = &cdd->ddev;
list_add_tail(&cchan->chan.device_node, &cdd->ddev.channels);
}
cdd->first_td_desc = n_chans;
return 0;
}
static void purge_descs(struct device *dev, struct cppi41_dd *cdd)
{
unsigned int mem_decs;
int i;
mem_decs = ALLOC_DECS_NUM * sizeof(struct cppi41_desc);
for (i = 0; i < DESCS_AREAS; i++) {
cppi_writel(0, cdd->qmgr_mem + QMGR_MEMBASE(i));
cppi_writel(0, cdd->qmgr_mem + QMGR_MEMCTRL(i));
dma_free_coherent(dev, mem_decs, cdd->cd,
cdd->descs_phys);
}
}
static void disable_sched(struct cppi41_dd *cdd)
{
cppi_writel(0, cdd->sched_mem + DMA_SCHED_CTRL);
}
static void deinit_cppi41(struct device *dev, struct cppi41_dd *cdd)
{
disable_sched(cdd);
purge_descs(dev, cdd);
cppi_writel(0, cdd->qmgr_mem + QMGR_LRAM0_BASE);
cppi_writel(0, cdd->qmgr_mem + QMGR_LRAM0_BASE);
dma_free_coherent(dev, QMGR_SCRATCH_SIZE, cdd->qmgr_scratch,
cdd->scratch_phys);
}
static int init_descs(struct device *dev, struct cppi41_dd *cdd)
{
unsigned int desc_size;
unsigned int mem_decs;
int i;
u32 reg;
u32 idx;
BUILD_BUG_ON(sizeof(struct cppi41_desc) &
(sizeof(struct cppi41_desc) - 1));
BUILD_BUG_ON(sizeof(struct cppi41_desc) < 32);
BUILD_BUG_ON(ALLOC_DECS_NUM < 32);
desc_size = sizeof(struct cppi41_desc);
mem_decs = ALLOC_DECS_NUM * desc_size;
idx = 0;
for (i = 0; i < DESCS_AREAS; i++) {
reg = idx << QMGR_MEMCTRL_IDX_SH;
reg |= (ilog2(desc_size) - 5) << QMGR_MEMCTRL_DESC_SH;
reg |= ilog2(ALLOC_DECS_NUM) - 5;
BUILD_BUG_ON(DESCS_AREAS != 1);
cdd->cd = dma_alloc_coherent(dev, mem_decs,
&cdd->descs_phys, GFP_KERNEL);
if (!cdd->cd)
return -ENOMEM;
cppi_writel(cdd->descs_phys, cdd->qmgr_mem + QMGR_MEMBASE(i));
cppi_writel(reg, cdd->qmgr_mem + QMGR_MEMCTRL(i));
idx += ALLOC_DECS_NUM;
}
return 0;
}
static void init_sched(struct cppi41_dd *cdd)
{
unsigned ch;
unsigned word;
u32 reg;
word = 0;
cppi_writel(0, cdd->sched_mem + DMA_SCHED_CTRL);
for (ch = 0; ch < cdd->n_chans; ch += 2) {
reg = SCHED_ENTRY0_CHAN(ch);
reg |= SCHED_ENTRY1_CHAN(ch) | SCHED_ENTRY1_IS_RX;
reg |= SCHED_ENTRY2_CHAN(ch + 1);
reg |= SCHED_ENTRY3_CHAN(ch + 1) | SCHED_ENTRY3_IS_RX;
cppi_writel(reg, cdd->sched_mem + DMA_SCHED_WORD(word));
word++;
}
reg = cdd->n_chans * 2 - 1;
reg |= DMA_SCHED_CTRL_EN;
cppi_writel(reg, cdd->sched_mem + DMA_SCHED_CTRL);
}
static int init_cppi41(struct device *dev, struct cppi41_dd *cdd)
{
int ret;
BUILD_BUG_ON(QMGR_SCRATCH_SIZE > ((1 << 14) - 1));
cdd->qmgr_scratch = dma_alloc_coherent(dev, QMGR_SCRATCH_SIZE,
&cdd->scratch_phys, GFP_KERNEL);
if (!cdd->qmgr_scratch)
return -ENOMEM;
cppi_writel(cdd->scratch_phys, cdd->qmgr_mem + QMGR_LRAM0_BASE);
cppi_writel(TOTAL_DESCS_NUM, cdd->qmgr_mem + QMGR_LRAM_SIZE);
cppi_writel(0, cdd->qmgr_mem + QMGR_LRAM1_BASE);
ret = init_descs(dev, cdd);
if (ret)
goto err_td;
cppi_writel(cdd->td_queue.submit, cdd->ctrl_mem + DMA_TDFDQ);
init_sched(cdd);
return 0;
err_td:
deinit_cppi41(dev, cdd);
return ret;
}
static struct platform_driver cpp41_dma_driver;
/*
* The param format is:
* X Y
* X: Port
* Y: 0 = RX else TX
*/
#define INFO_PORT 0
#define INFO_IS_TX 1
static bool cpp41_dma_filter_fn(struct dma_chan *chan, void *param)
{
struct cppi41_channel *cchan;
struct cppi41_dd *cdd;
const struct chan_queues *queues;
u32 *num = param;
if (chan->device->dev->driver != &cpp41_dma_driver.driver)
return false;
cchan = to_cpp41_chan(chan);
if (cchan->port_num != num[INFO_PORT])
return false;
if (cchan->is_tx && !num[INFO_IS_TX])
return false;
cdd = cchan->cdd;
if (cchan->is_tx)
queues = cdd->queues_tx;
else
queues = cdd->queues_rx;
BUILD_BUG_ON(ARRAY_SIZE(am335x_usb_queues_rx) !=
ARRAY_SIZE(am335x_usb_queues_tx));
if (WARN_ON(cchan->port_num >= ARRAY_SIZE(am335x_usb_queues_rx)))
return false;
cchan->q_num = queues[cchan->port_num].submit;
cchan->q_comp_num = queues[cchan->port_num].complete;
return true;
}
static struct of_dma_filter_info cpp41_dma_info = {
.filter_fn = cpp41_dma_filter_fn,
};
static struct dma_chan *cppi41_dma_xlate(struct of_phandle_args *dma_spec,
struct of_dma *ofdma)
{
int count = dma_spec->args_count;
struct of_dma_filter_info *info = ofdma->of_dma_data;
if (!info || !info->filter_fn)
return NULL;
if (count != 2)
return NULL;
return dma_request_channel(info->dma_cap, info->filter_fn,
&dma_spec->args[0]);
}
static const struct cppi_glue_infos am335x_usb_infos = {
.queues_rx = am335x_usb_queues_rx,
.queues_tx = am335x_usb_queues_tx,
.td_queue = { .submit = 31, .complete = 0 },
.first_completion_queue = 93,
.qmgr_num_pend = 5,
};
static const struct cppi_glue_infos da8xx_usb_infos = {
.queues_rx = da8xx_usb_queues_rx,
.queues_tx = da8xx_usb_queues_tx,
.td_queue = { .submit = 31, .complete = 0 },
.first_completion_queue = 24,
.qmgr_num_pend = 2,
};
static const struct of_device_id cppi41_dma_ids[] = {
{ .compatible = "ti,am3359-cppi41", .data = &am335x_usb_infos},
{ .compatible = "ti,da830-cppi41", .data = &da8xx_usb_infos},
{},
};
MODULE_DEVICE_TABLE(of, cppi41_dma_ids);
static const struct cppi_glue_infos *get_glue_info(struct device *dev)
{
const struct of_device_id *of_id;
of_id = of_match_node(cppi41_dma_ids, dev->of_node);
if (!of_id)
return NULL;
return of_id->data;
}
#define CPPI41_DMA_BUSWIDTHS (BIT(DMA_SLAVE_BUSWIDTH_1_BYTE) | \
BIT(DMA_SLAVE_BUSWIDTH_2_BYTES) | \
BIT(DMA_SLAVE_BUSWIDTH_3_BYTES) | \
BIT(DMA_SLAVE_BUSWIDTH_4_BYTES))
static int cppi41_dma_probe(struct platform_device *pdev)
{
struct cppi41_dd *cdd;
struct device *dev = &pdev->dev;
const struct cppi_glue_infos *glue_info;
struct resource *mem;
int index;
int irq;
int ret;
glue_info = get_glue_info(dev);
if (!glue_info)
return -EINVAL;
cdd = devm_kzalloc(&pdev->dev, sizeof(*cdd), GFP_KERNEL);
if (!cdd)
return -ENOMEM;
dma_cap_set(DMA_SLAVE, cdd->ddev.cap_mask);
cdd->ddev.device_alloc_chan_resources = cppi41_dma_alloc_chan_resources;
cdd->ddev.device_free_chan_resources = cppi41_dma_free_chan_resources;
cdd->ddev.device_tx_status = cppi41_dma_tx_status;
cdd->ddev.device_issue_pending = cppi41_dma_issue_pending;
cdd->ddev.device_prep_slave_sg = cppi41_dma_prep_slave_sg;
cdd->ddev.device_terminate_all = cppi41_stop_chan;
cdd->ddev.directions = BIT(DMA_DEV_TO_MEM) | BIT(DMA_MEM_TO_DEV);
cdd->ddev.src_addr_widths = CPPI41_DMA_BUSWIDTHS;
cdd->ddev.dst_addr_widths = CPPI41_DMA_BUSWIDTHS;
cdd->ddev.residue_granularity = DMA_RESIDUE_GRANULARITY_BURST;
cdd->ddev.dev = dev;
INIT_LIST_HEAD(&cdd->ddev.channels);
cpp41_dma_info.dma_cap = cdd->ddev.cap_mask;
index = of_property_match_string(dev->of_node,
"reg-names", "controller");
if (index < 0)
return index;
mem = platform_get_resource(pdev, IORESOURCE_MEM, index);
cdd->ctrl_mem = devm_ioremap_resource(dev, mem);
if (IS_ERR(cdd->ctrl_mem))
return PTR_ERR(cdd->ctrl_mem);
mem = platform_get_resource(pdev, IORESOURCE_MEM, index + 1);
cdd->sched_mem = devm_ioremap_resource(dev, mem);
if (IS_ERR(cdd->sched_mem))
return PTR_ERR(cdd->sched_mem);
mem = platform_get_resource(pdev, IORESOURCE_MEM, index + 2);
cdd->qmgr_mem = devm_ioremap_resource(dev, mem);
if (IS_ERR(cdd->qmgr_mem))
return PTR_ERR(cdd->qmgr_mem);
spin_lock_init(&cdd->lock);
INIT_LIST_HEAD(&cdd->pending);
platform_set_drvdata(pdev, cdd);
pm_runtime_enable(dev);
pm_runtime_set_autosuspend_delay(dev, 100);
pm_runtime_use_autosuspend(dev);
ret = pm_runtime_get_sync(dev);
if (ret < 0)
goto err_get_sync;
cdd->queues_rx = glue_info->queues_rx;
cdd->queues_tx = glue_info->queues_tx;
cdd->td_queue = glue_info->td_queue;
cdd->qmgr_num_pend = glue_info->qmgr_num_pend;
cdd->first_completion_queue = glue_info->first_completion_queue;
ret = of_property_read_u32(dev->of_node,
"#dma-channels", &cdd->n_chans);
if (ret)
goto err_get_n_chans;
ret = init_cppi41(dev, cdd);
if (ret)
goto err_init_cppi;
ret = cppi41_add_chans(dev, cdd);
if (ret)
goto err_chans;
irq = irq_of_parse_and_map(dev->of_node, 0);
if (!irq) {
ret = -EINVAL;
goto err_chans;
}
ret = devm_request_irq(&pdev->dev, irq, cppi41_irq, IRQF_SHARED,
dev_name(dev), cdd);
if (ret)
goto err_chans;
cdd->irq = irq;
ret = dma_async_device_register(&cdd->ddev);
if (ret)
goto err_chans;
ret = of_dma_controller_register(dev->of_node,
cppi41_dma_xlate, &cpp41_dma_info);
if (ret)
goto err_of;
pm_runtime_mark_last_busy(dev);
pm_runtime_put_autosuspend(dev);
return 0;
err_of:
dma_async_device_unregister(&cdd->ddev);
err_chans:
deinit_cppi41(dev, cdd);
err_init_cppi:
pm_runtime_dont_use_autosuspend(dev);
err_get_n_chans:
err_get_sync:
pm_runtime_put_sync(dev);
pm_runtime_disable(dev);
return ret;
}
static int cppi41_dma_remove(struct platform_device *pdev)
{
struct cppi41_dd *cdd = platform_get_drvdata(pdev);
int error;
error = pm_runtime_get_sync(&pdev->dev);
if (error < 0)
dev_err(&pdev->dev, "%s could not pm_runtime_get: %i\n",
__func__, error);
of_dma_controller_free(pdev->dev.of_node);
dma_async_device_unregister(&cdd->ddev);
devm_free_irq(&pdev->dev, cdd->irq, cdd);
deinit_cppi41(&pdev->dev, cdd);
pm_runtime_dont_use_autosuspend(&pdev->dev);
pm_runtime_put_sync(&pdev->dev);
pm_runtime_disable(&pdev->dev);
return 0;
}
static int __maybe_unused cppi41_suspend(struct device *dev)
{
struct cppi41_dd *cdd = dev_get_drvdata(dev);
cdd->dma_tdfdq = cppi_readl(cdd->ctrl_mem + DMA_TDFDQ);
disable_sched(cdd);
return 0;
}
static int __maybe_unused cppi41_resume(struct device *dev)
{
struct cppi41_dd *cdd = dev_get_drvdata(dev);
struct cppi41_channel *c;
int i;
for (i = 0; i < DESCS_AREAS; i++)
cppi_writel(cdd->descs_phys, cdd->qmgr_mem + QMGR_MEMBASE(i));
list_for_each_entry(c, &cdd->ddev.channels, chan.device_node)
if (!c->is_tx)
cppi_writel(c->q_num, c->gcr_reg + RXHPCRA0);
init_sched(cdd);
cppi_writel(cdd->dma_tdfdq, cdd->ctrl_mem + DMA_TDFDQ);
cppi_writel(cdd->scratch_phys, cdd->qmgr_mem + QMGR_LRAM0_BASE);
cppi_writel(QMGR_SCRATCH_SIZE, cdd->qmgr_mem + QMGR_LRAM_SIZE);
cppi_writel(0, cdd->qmgr_mem + QMGR_LRAM1_BASE);
return 0;
}
static int __maybe_unused cppi41_runtime_suspend(struct device *dev)
{
struct cppi41_dd *cdd = dev_get_drvdata(dev);
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
unsigned long flags;
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
spin_lock_irqsave(&cdd->lock, flags);
cdd->is_suspended = true;
WARN_ON(!list_empty(&cdd->pending));
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
spin_unlock_irqrestore(&cdd->lock, flags);
return 0;
}
static int __maybe_unused cppi41_runtime_resume(struct device *dev)
{
struct cppi41_dd *cdd = dev_get_drvdata(dev);
unsigned long flags;
spin_lock_irqsave(&cdd->lock, flags);
dmaengine: cppi41: Fix oops in cppi41_runtime_resume Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") together with recent MUSB changes allowed USB and DMA on BeagleBone to idle when no cable is connected. But looks like few corner case issues still remain. Looks like just by re-plugging USB cable about ten or so times on BeagleBone when configured in USB peripheral mode we can get warnings and eventually trigger an oops in cppi41 DMA: WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+ x28/0x38 [cppi41] ... WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452 push_desc_queue+0x94/0x9c [cppi41] ... Unable to handle kernel NULL pointer dereference at virtual address 00000104 pgd = c0004000 [00000104] *pgd=00000000 Internal error: Oops: 805 [#1] SMP ARM ... [<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>] (__rpm_callback+0xc0/0x214) [<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80) [<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c) [<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8) [<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808) This is because of a race with runtime PM and cppi41_dma_issue_pending() as reported by Alexandre Bailon <abailon@baylibre.com> in earlier set of patches. Based on mailing list discussions we however came to the conclusion that a different fix from Alexandre's fix is needed in order to guarantee that DMA is really active when we try to use it. To fix the issue, we need to add a driver specific flag as we otherwise can have -EINPROGRESS state set by runtime PM and can't rely on pm_runtime_active() to tell us when we can use the DMA. And we need to make sure the DMA transfers get triggered in the queued order. So let's always queue the transfers, then flush the queue from both cppi41_dma_issue_pending() and cppi41_runtime_resume() as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an earlier example patch. For reference, this is also documented in Documentation/power/runtime_pm.txt in the example at the end of the file as pointed out by Grygorii Strashko <grygorii.strashko@ti.com>. Based on earlier patches from Alexandre Bailon <abailon@baylibre.com> and Grygorii Strashko <grygorii.strashko@ti.com> modified based on testing and what was discussed on the mailing lists. Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support") Cc: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Bin Liu <b-liu@ti.com> Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Kevin Hilman <khilman@baylibre.com> Cc: Patrick Titiano <ptitiano@baylibre.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reported-by: Alexandre Bailon <abailon@baylibre.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Tested-by: Bin Liu <b-liu@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-20 00:49:08 +08:00
cdd->is_suspended = false;
cppi41_run_queue(cdd);
spin_unlock_irqrestore(&cdd->lock, flags);
return 0;
}
static const struct dev_pm_ops cppi41_pm_ops = {
SET_LATE_SYSTEM_SLEEP_PM_OPS(cppi41_suspend, cppi41_resume)
SET_RUNTIME_PM_OPS(cppi41_runtime_suspend,
cppi41_runtime_resume,
NULL)
};
static struct platform_driver cpp41_dma_driver = {
.probe = cppi41_dma_probe,
.remove = cppi41_dma_remove,
.driver = {
.name = "cppi41-dma-engine",
.pm = &cppi41_pm_ops,
.of_match_table = of_match_ptr(cppi41_dma_ids),
},
};
module_platform_driver(cpp41_dma_driver);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Sebastian Andrzej Siewior <bigeasy@linutronix.de>");