linux_dsm_epyc7002/drivers/dma/pl330.c

3053 lines
68 KiB
C
Raw Normal View History

/*
* Copyright (c) 2012 Samsung Electronics Co., Ltd.
* http://www.samsung.com
*
* Copyright (C) 2010 Samsung Electronics Co. Ltd.
* Jaswinder Singh <jassi.brar@samsung.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*/
#include <linux/kernel.h>
#include <linux/io.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/string.h>
#include <linux/delay.h>
#include <linux/interrupt.h>
#include <linux/dma-mapping.h>
#include <linux/dmaengine.h>
#include <linux/amba/bus.h>
#include <linux/scatterlist.h>
#include <linux/of.h>
#include <linux/of_dma.h>
#include <linux/err.h>
#include <linux/pm_runtime.h>
#include "dmaengine.h"
#define PL330_MAX_CHAN 8
#define PL330_MAX_IRQS 32
#define PL330_MAX_PERI 32
#define PL330_MAX_BURST 16
#define PL330_QUIRK_BROKEN_NO_FLUSHP BIT(0)
enum pl330_cachectrl {
CCTRL0, /* Noncacheable and nonbufferable */
CCTRL1, /* Bufferable only */
CCTRL2, /* Cacheable, but do not allocate */
CCTRL3, /* Cacheable and bufferable, but do not allocate */
INVALID1, /* AWCACHE = 0x1000 */
INVALID2,
CCTRL6, /* Cacheable write-through, allocate on writes only */
CCTRL7, /* Cacheable write-back, allocate on writes only */
};
enum pl330_byteswap {
SWAP_NO,
SWAP_2,
SWAP_4,
SWAP_8,
SWAP_16,
};
/* Register and Bit field Definitions */
#define DS 0x0
#define DS_ST_STOP 0x0
#define DS_ST_EXEC 0x1
#define DS_ST_CMISS 0x2
#define DS_ST_UPDTPC 0x3
#define DS_ST_WFE 0x4
#define DS_ST_ATBRR 0x5
#define DS_ST_QBUSY 0x6
#define DS_ST_WFP 0x7
#define DS_ST_KILL 0x8
#define DS_ST_CMPLT 0x9
#define DS_ST_FLTCMP 0xe
#define DS_ST_FAULT 0xf
#define DPC 0x4
#define INTEN 0x20
#define ES 0x24
#define INTSTATUS 0x28
#define INTCLR 0x2c
#define FSM 0x30
#define FSC 0x34
#define FTM 0x38
#define _FTC 0x40
#define FTC(n) (_FTC + (n)*0x4)
#define _CS 0x100
#define CS(n) (_CS + (n)*0x8)
#define CS_CNS (1 << 21)
#define _CPC 0x104
#define CPC(n) (_CPC + (n)*0x8)
#define _SA 0x400
#define SA(n) (_SA + (n)*0x20)
#define _DA 0x404
#define DA(n) (_DA + (n)*0x20)
#define _CC 0x408
#define CC(n) (_CC + (n)*0x20)
#define CC_SRCINC (1 << 0)
#define CC_DSTINC (1 << 14)
#define CC_SRCPRI (1 << 8)
#define CC_DSTPRI (1 << 22)
#define CC_SRCNS (1 << 9)
#define CC_DSTNS (1 << 23)
#define CC_SRCIA (1 << 10)
#define CC_DSTIA (1 << 24)
#define CC_SRCBRSTLEN_SHFT 4
#define CC_DSTBRSTLEN_SHFT 18
#define CC_SRCBRSTSIZE_SHFT 1
#define CC_DSTBRSTSIZE_SHFT 15
#define CC_SRCCCTRL_SHFT 11
#define CC_SRCCCTRL_MASK 0x7
#define CC_DSTCCTRL_SHFT 25
#define CC_DRCCCTRL_MASK 0x7
#define CC_SWAP_SHFT 28
#define _LC0 0x40c
#define LC0(n) (_LC0 + (n)*0x20)
#define _LC1 0x410
#define LC1(n) (_LC1 + (n)*0x20)
#define DBGSTATUS 0xd00
#define DBG_BUSY (1 << 0)
#define DBGCMD 0xd04
#define DBGINST0 0xd08
#define DBGINST1 0xd0c
#define CR0 0xe00
#define CR1 0xe04
#define CR2 0xe08
#define CR3 0xe0c
#define CR4 0xe10
#define CRD 0xe14
#define PERIPH_ID 0xfe0
#define PERIPH_REV_SHIFT 20
#define PERIPH_REV_MASK 0xf
#define PERIPH_REV_R0P0 0
#define PERIPH_REV_R1P0 1
#define PERIPH_REV_R1P1 2
#define CR0_PERIPH_REQ_SET (1 << 0)
#define CR0_BOOT_EN_SET (1 << 1)
#define CR0_BOOT_MAN_NS (1 << 2)
#define CR0_NUM_CHANS_SHIFT 4
#define CR0_NUM_CHANS_MASK 0x7
#define CR0_NUM_PERIPH_SHIFT 12
#define CR0_NUM_PERIPH_MASK 0x1f
#define CR0_NUM_EVENTS_SHIFT 17
#define CR0_NUM_EVENTS_MASK 0x1f
#define CR1_ICACHE_LEN_SHIFT 0
#define CR1_ICACHE_LEN_MASK 0x7
#define CR1_NUM_ICACHELINES_SHIFT 4
#define CR1_NUM_ICACHELINES_MASK 0xf
#define CRD_DATA_WIDTH_SHIFT 0
#define CRD_DATA_WIDTH_MASK 0x7
#define CRD_WR_CAP_SHIFT 4
#define CRD_WR_CAP_MASK 0x7
#define CRD_WR_Q_DEP_SHIFT 8
#define CRD_WR_Q_DEP_MASK 0xf
#define CRD_RD_CAP_SHIFT 12
#define CRD_RD_CAP_MASK 0x7
#define CRD_RD_Q_DEP_SHIFT 16
#define CRD_RD_Q_DEP_MASK 0xf
#define CRD_DATA_BUFF_SHIFT 20
#define CRD_DATA_BUFF_MASK 0x3ff
#define PART 0x330
#define DESIGNER 0x41
#define REVISION 0x0
#define INTEG_CFG 0x0
#define PERIPH_ID_VAL ((PART << 0) | (DESIGNER << 12))
#define PL330_STATE_STOPPED (1 << 0)
#define PL330_STATE_EXECUTING (1 << 1)
#define PL330_STATE_WFE (1 << 2)
#define PL330_STATE_FAULTING (1 << 3)
#define PL330_STATE_COMPLETING (1 << 4)
#define PL330_STATE_WFP (1 << 5)
#define PL330_STATE_KILLING (1 << 6)
#define PL330_STATE_FAULT_COMPLETING (1 << 7)
#define PL330_STATE_CACHEMISS (1 << 8)
#define PL330_STATE_UPDTPC (1 << 9)
#define PL330_STATE_ATBARRIER (1 << 10)
#define PL330_STATE_QUEUEBUSY (1 << 11)
#define PL330_STATE_INVALID (1 << 15)
#define PL330_STABLE_STATES (PL330_STATE_STOPPED | PL330_STATE_EXECUTING \
| PL330_STATE_WFE | PL330_STATE_FAULTING)
#define CMD_DMAADDH 0x54
#define CMD_DMAEND 0x00
#define CMD_DMAFLUSHP 0x35
#define CMD_DMAGO 0xa0
#define CMD_DMALD 0x04
#define CMD_DMALDP 0x25
#define CMD_DMALP 0x20
#define CMD_DMALPEND 0x28
#define CMD_DMAKILL 0x01
#define CMD_DMAMOV 0xbc
#define CMD_DMANOP 0x18
#define CMD_DMARMB 0x12
#define CMD_DMASEV 0x34
#define CMD_DMAST 0x08
#define CMD_DMASTP 0x29
#define CMD_DMASTZ 0x0c
#define CMD_DMAWFE 0x36
#define CMD_DMAWFP 0x30
#define CMD_DMAWMB 0x13
#define SZ_DMAADDH 3
#define SZ_DMAEND 1
#define SZ_DMAFLUSHP 2
#define SZ_DMALD 1
#define SZ_DMALDP 2
#define SZ_DMALP 2
#define SZ_DMALPEND 2
#define SZ_DMAKILL 1
#define SZ_DMAMOV 6
#define SZ_DMANOP 1
#define SZ_DMARMB 1
#define SZ_DMASEV 2
#define SZ_DMAST 1
#define SZ_DMASTP 2
#define SZ_DMASTZ 1
#define SZ_DMAWFE 2
#define SZ_DMAWFP 2
#define SZ_DMAWMB 1
#define SZ_DMAGO 6
#define BRST_LEN(ccr) ((((ccr) >> CC_SRCBRSTLEN_SHFT) & 0xf) + 1)
#define BRST_SIZE(ccr) (1 << (((ccr) >> CC_SRCBRSTSIZE_SHFT) & 0x7))
#define BYTE_TO_BURST(b, ccr) ((b) / BRST_SIZE(ccr) / BRST_LEN(ccr))
#define BURST_TO_BYTE(c, ccr) ((c) * BRST_SIZE(ccr) * BRST_LEN(ccr))
/*
* With 256 bytes, we can do more than 2.5MB and 5MB xfers per req
* at 1byte/burst for P<->M and M<->M respectively.
* For typical scenario, at 1word/burst, 10MB and 20MB xfers per req
* should be enough for P<->M and M<->M respectively.
*/
#define MCODE_BUFF_PER_REQ 256
/* Use this _only_ to wait on transient states */
#define UNTIL(t, s) while (!(_state(t) & (s))) cpu_relax();
#ifdef PL330_DEBUG_MCGEN
static unsigned cmd_line;
#define PL330_DBGCMD_DUMP(off, x...) do { \
printk("%x:", cmd_line); \
printk(x); \
cmd_line += off; \
} while (0)
#define PL330_DBGMC_START(addr) (cmd_line = addr)
#else
#define PL330_DBGCMD_DUMP(off, x...) do {} while (0)
#define PL330_DBGMC_START(addr) do {} while (0)
#endif
/* The number of default descriptors */
#define NR_DEFAULT_DESC 16
/* Delay for runtime PM autosuspend, ms */
#define PL330_AUTOSUSPEND_DELAY 20
/* Populated by the PL330 core driver for DMA API driver's info */
struct pl330_config {
u32 periph_id;
#define DMAC_MODE_NS (1 << 0)
unsigned int mode;
unsigned int data_bus_width:10; /* In number of bits */
unsigned int data_buf_dep:11;
unsigned int num_chan:4;
unsigned int num_peri:6;
u32 peri_ns;
unsigned int num_events:6;
u32 irq_ns;
};
/**
* Request Configuration.
* The PL330 core does not modify this and uses the last
* working configuration if the request doesn't provide any.
*
* The Client may want to provide this info only for the
* first request and a request with new settings.
*/
struct pl330_reqcfg {
/* Address Incrementing */
unsigned dst_inc:1;
unsigned src_inc:1;
/*
* For now, the SRC & DST protection levels
* and burst size/length are assumed same.
*/
bool nonsecure;
bool privileged;
bool insnaccess;
unsigned brst_len:5;
unsigned brst_size:3; /* in power of 2 */
enum pl330_cachectrl dcctl;
enum pl330_cachectrl scctl;
enum pl330_byteswap swap;
struct pl330_config *pcfg;
};
/*
* One cycle of DMAC operation.
* There may be more than one xfer in a request.
*/
struct pl330_xfer {
u32 src_addr;
u32 dst_addr;
/* Size to xfer */
u32 bytes;
};
/* The xfer callbacks are made with one of these arguments. */
enum pl330_op_err {
/* The all xfers in the request were success. */
PL330_ERR_NONE,
/* If req aborted due to global error. */
PL330_ERR_ABORT,
/* If req failed due to problem with Channel. */
PL330_ERR_FAIL,
};
enum dmamov_dst {
SAR = 0,
CCR,
DAR,
};
enum pl330_dst {
SRC = 0,
DST,
};
enum pl330_cond {
SINGLE,
BURST,
ALWAYS,
};
struct dma_pl330_desc;
struct _pl330_req {
u32 mc_bus;
void *mc_cpu;
struct dma_pl330_desc *desc;
};
/* ToBeDone for tasklet */
struct _pl330_tbd {
bool reset_dmac;
bool reset_mngr;
u8 reset_chan;
};
/* A DMAC Thread */
struct pl330_thread {
u8 id;
int ev;
/* If the channel is not yet acquired by any client */
bool free;
/* Parent DMAC */
struct pl330_dmac *dmac;
/* Only two at a time */
struct _pl330_req req[2];
/* Index of the last enqueued request */
unsigned lstenq;
/* Index of the last submitted request or -1 if the DMA is stopped */
int req_running;
};
enum pl330_dmac_state {
UNINIT,
INIT,
DYING,
};
enum desc_status {
/* In the DMAC pool */
FREE,
/*
* Allocated to some channel during prep_xxx
* Also may be sitting on the work_list.
*/
PREP,
/*
* Sitting on the work_list and already submitted
* to the PL330 core. Not more than two descriptors
* of a channel can be BUSY at any time.
*/
BUSY,
/*
* Sitting on the channel work_list but xfer done
* by PL330 core
*/
DONE,
};
struct dma_pl330_chan {
/* Schedule desc completion */
struct tasklet_struct task;
/* DMA-Engine Channel */
struct dma_chan chan;
/* List of submitted descriptors */
struct list_head submitted_list;
/* List of issued descriptors */
struct list_head work_list;
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
/* List of completed descriptors */
struct list_head completed_list;
/* Pointer to the DMAC that manages this channel,
* NULL if the channel is available to be acquired.
* As the parent, this DMAC also provides descriptors
* to the channel.
*/
struct pl330_dmac *dmac;
/* To protect channel manipulation */
spinlock_t lock;
/*
* Hardware channel thread of PL330 DMAC. NULL if the channel is
* available.
*/
struct pl330_thread *thread;
/* For D-to-M and M-to-D channels */
int burst_sz; /* the peripheral fifo width */
int burst_len; /* the number of burst */
phys_addr_t fifo_addr;
/* DMA-mapped view of the FIFO; may differ if an IOMMU is present */
dma_addr_t fifo_dma;
enum dma_data_direction dir;
/* for cyclic capability */
bool cyclic;
/* for runtime pm tracking */
bool active;
};
struct pl330_dmac {
/* DMA-Engine Device */
struct dma_device ddma;
/* Holds info about sg limitations */
struct device_dma_parameters dma_parms;
/* Pool of descriptors available for the DMAC's channels */
struct list_head desc_pool;
/* To protect desc_pool manipulation */
spinlock_t pool_lock;
/* Size of MicroCode buffers for each channel. */
unsigned mcbufsz;
/* ioremap'ed address of PL330 registers. */
void __iomem *base;
/* Populated by the PL330 core driver during pl330_add */
struct pl330_config pcfg;
spinlock_t lock;
/* Maximum possible events/irqs */
int events[32];
/* BUS address of MicroCode buffer */
dma_addr_t mcode_bus;
/* CPU address of MicroCode buffer */
void *mcode_cpu;
/* List of all Channel threads */
struct pl330_thread *channels;
/* Pointer to the MANAGER thread */
struct pl330_thread *manager;
/* To handle bad news in interrupt */
struct tasklet_struct tasks;
struct _pl330_tbd dmac_tbd;
/* State of DMAC operation */
enum pl330_dmac_state state;
/* Holds list of reqs with due callbacks */
struct list_head req_done;
/* Peripheral channels connected to this DMAC */
unsigned int num_peripherals;
struct dma_pl330_chan *peripherals; /* keep at end */
int quirks;
};
static struct pl330_of_quirks {
char *quirk;
int id;
} of_quirks[] = {
{
.quirk = "arm,pl330-broken-no-flushp",
.id = PL330_QUIRK_BROKEN_NO_FLUSHP,
}
};
struct dma_pl330_desc {
/* To attach to a queue as child */
struct list_head node;
/* Descriptor for the DMA Engine API */
struct dma_async_tx_descriptor txd;
/* Xfer for PL330 core */
struct pl330_xfer px;
struct pl330_reqcfg rqcfg;
enum desc_status status;
int bytes_requested;
bool last;
/* The channel which currently holds this desc */
struct dma_pl330_chan *pchan;
enum dma_transfer_direction rqtype;
/* Index of peripheral for the xfer. */
unsigned peri:5;
/* Hook to attach to DMAC's list of reqs with due callback */
struct list_head rqd;
};
struct _xfer_spec {
u32 ccr;
struct dma_pl330_desc *desc;
};
static inline bool _queue_full(struct pl330_thread *thrd)
{
return thrd->req[0].desc != NULL && thrd->req[1].desc != NULL;
}
static inline bool is_manager(struct pl330_thread *thrd)
{
return thrd->dmac->manager == thrd;
}
/* If manager of the thread is in Non-Secure mode */
static inline bool _manager_ns(struct pl330_thread *thrd)
{
return (thrd->dmac->pcfg.mode & DMAC_MODE_NS) ? true : false;
}
static inline u32 get_revision(u32 periph_id)
{
return (periph_id >> PERIPH_REV_SHIFT) & PERIPH_REV_MASK;
}
static inline u32 _emit_END(unsigned dry_run, u8 buf[])
{
if (dry_run)
return SZ_DMAEND;
buf[0] = CMD_DMAEND;
PL330_DBGCMD_DUMP(SZ_DMAEND, "\tDMAEND\n");
return SZ_DMAEND;
}
static inline u32 _emit_FLUSHP(unsigned dry_run, u8 buf[], u8 peri)
{
if (dry_run)
return SZ_DMAFLUSHP;
buf[0] = CMD_DMAFLUSHP;
peri &= 0x1f;
peri <<= 3;
buf[1] = peri;
PL330_DBGCMD_DUMP(SZ_DMAFLUSHP, "\tDMAFLUSHP %u\n", peri >> 3);
return SZ_DMAFLUSHP;
}
static inline u32 _emit_LD(unsigned dry_run, u8 buf[], enum pl330_cond cond)
{
if (dry_run)
return SZ_DMALD;
buf[0] = CMD_DMALD;
if (cond == SINGLE)
buf[0] |= (0 << 1) | (1 << 0);
else if (cond == BURST)
buf[0] |= (1 << 1) | (1 << 0);
PL330_DBGCMD_DUMP(SZ_DMALD, "\tDMALD%c\n",
cond == SINGLE ? 'S' : (cond == BURST ? 'B' : 'A'));
return SZ_DMALD;
}
static inline u32 _emit_LDP(unsigned dry_run, u8 buf[],
enum pl330_cond cond, u8 peri)
{
if (dry_run)
return SZ_DMALDP;
buf[0] = CMD_DMALDP;
if (cond == BURST)
buf[0] |= (1 << 1);
peri &= 0x1f;
peri <<= 3;
buf[1] = peri;
PL330_DBGCMD_DUMP(SZ_DMALDP, "\tDMALDP%c %u\n",
cond == SINGLE ? 'S' : 'B', peri >> 3);
return SZ_DMALDP;
}
static inline u32 _emit_LP(unsigned dry_run, u8 buf[],
unsigned loop, u8 cnt)
{
if (dry_run)
return SZ_DMALP;
buf[0] = CMD_DMALP;
if (loop)
buf[0] |= (1 << 1);
cnt--; /* DMAC increments by 1 internally */
buf[1] = cnt;
PL330_DBGCMD_DUMP(SZ_DMALP, "\tDMALP_%c %u\n", loop ? '1' : '0', cnt);
return SZ_DMALP;
}
struct _arg_LPEND {
enum pl330_cond cond;
bool forever;
unsigned loop;
u8 bjump;
};
static inline u32 _emit_LPEND(unsigned dry_run, u8 buf[],
const struct _arg_LPEND *arg)
{
enum pl330_cond cond = arg->cond;
bool forever = arg->forever;
unsigned loop = arg->loop;
u8 bjump = arg->bjump;
if (dry_run)
return SZ_DMALPEND;
buf[0] = CMD_DMALPEND;
if (loop)
buf[0] |= (1 << 2);
if (!forever)
buf[0] |= (1 << 4);
if (cond == SINGLE)
buf[0] |= (0 << 1) | (1 << 0);
else if (cond == BURST)
buf[0] |= (1 << 1) | (1 << 0);
buf[1] = bjump;
PL330_DBGCMD_DUMP(SZ_DMALPEND, "\tDMALP%s%c_%c bjmpto_%x\n",
forever ? "FE" : "END",
cond == SINGLE ? 'S' : (cond == BURST ? 'B' : 'A'),
loop ? '1' : '0',
bjump);
return SZ_DMALPEND;
}
static inline u32 _emit_KILL(unsigned dry_run, u8 buf[])
{
if (dry_run)
return SZ_DMAKILL;
buf[0] = CMD_DMAKILL;
return SZ_DMAKILL;
}
static inline u32 _emit_MOV(unsigned dry_run, u8 buf[],
enum dmamov_dst dst, u32 val)
{
if (dry_run)
return SZ_DMAMOV;
buf[0] = CMD_DMAMOV;
buf[1] = dst;
dmaengine: pl330: do not generate unaligned access When PL330 is used with !MMU the following fault is seen: Unhandled fault: alignment exception (0x801) at 0x8f26a002 Internal error: : 801 [#1] ARM Modules linked in: CPU: 0 PID: 640 Comm: dma0chan0-copy0 Not tainted 4.8.0-6a82063-clean+ #1600 Hardware name: ARM-Versatile Express task: 8f1baa80 task.stack: 8e6fe000 PC is at _setup_req+0x4c/0x350 LR is at 0x8f2cbc00 pc : [<801ea538>] lr : [<8f2cbc00>] psr: 60000093 sp : 8e6ffdc0 ip : 00000000 fp : 00000000 r10: 00000000 r9 : 8f2cba10 r8 : 8f2cbc00 r7 : 80000013 r6 : 8f21a050 r5 : 8f21a000 r4 : 8f2ac800 r3 : 8e6ffe18 r2 : 00944251 r1 : ffffffbc r0 : 8f26a000 Flags: nZCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment none Control: 00c5387c Process dma0chan0-copy0 (pid: 640, stack limit = 0x8e6fe210) Stack: (0x8e6ffdc0 to 0x8e700000) fdc0: 00000001 60000093 00000000 8f2cba10 8f26a000 00000004 8f0ae000 8f2cbc00 fde0: 8f0ae000 8f2ac800 8f21a000 8f21a050 80000013 8f2cbc00 8f2cba10 00000000 fe00: 60000093 801ebca0 8e6ffe18 000013ff 40000093 00000000 00944251 8f2ac800 fe20: a0000013 8f2b1320 00001986 00000000 00000001 000013ff 8f1e4f00 8f2cba10 fe40: 8e6fff6c 801e9044 00000003 00000000 fef98c80 002faf07 8e6ffe7c 00000000 fe60: 00000002 00000000 00001986 8f1f158d 8f1e4f00 80568de4 00000002 00000000 fe80: 00001986 8f1f53ff 40000001 80580500 8f1f158d 8001e00c 00000000 cfdfdfdf fea0: fdae2a25 00000001 00000004 8e6fe000 00000008 00000010 00000000 00000005 fec0: 8f2b1330 8f2b1334 8e6ffe80 8e6ffe8c 00001986 00000000 8f21a014 00000001 fee0: 8e6ffe60 8e6ffe78 00000002 00000000 000013ff 00000001 80568de4 8f1e8018 ff00: 0000158d 8055ec30 00000001 803f6b00 00001986 8f2cba10 fdae2a25 00000001 ff20: 8f1baca8 8e6fff24 8e6fff24 00000000 8e6fff24 ac6f3037 00000000 00000000 ff40: 00000000 8e6fe000 8f1e4f40 00000000 8f1e4f40 8f1e4f00 801e84ec 00000000 ff60: 00000000 00000000 00000000 80031714 dfdfdfcf 00000000 dfdfdfcf 8f1e4f00 ff80: 00000000 8e6fff84 8e6fff84 00000000 8e6fff90 8e6fff90 8e6fffac 8f1e4f40 ffa0: 80031640 00000000 00000000 8000f548 00000000 00000000 00000000 00000000 ffc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 ffe0: 00000000 00000000 00000000 00000000 00000013 00000000 dfdfdfcf cfdfdfdf [<801ea538>] (_setup_req) from [<801ebca0>] (pl330_tasklet+0x41c/0x490) [<801ebca0>] (pl330_tasklet) from [<801e9044>] (dmatest_func+0xb58/0x149c) [<801e9044>] (dmatest_func) from [<80031714>] (kthread+0xd4/0xec) [<80031714>] (kthread) from [<8000f548>] (ret_from_fork+0x14/0x2c) Code: e3a03001 e3e01043 e5c03001 e59d3048 (e5802002) This happens because _emit_{ADDH,MOV,GO) accessing to unaligned data while writing to buffer. Fix it with writing to buffer byte by byte. Reviewed-by: Robin Murphy <robin.murphy@arm.com> Tested-by: Robin Murphy <robin.murphy@arm.com> Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2016-12-07 20:17:40 +07:00
buf[2] = val;
buf[3] = val >> 8;
buf[4] = val >> 16;
buf[5] = val >> 24;
PL330_DBGCMD_DUMP(SZ_DMAMOV, "\tDMAMOV %s 0x%x\n",
dst == SAR ? "SAR" : (dst == DAR ? "DAR" : "CCR"), val);
return SZ_DMAMOV;
}
static inline u32 _emit_RMB(unsigned dry_run, u8 buf[])
{
if (dry_run)
return SZ_DMARMB;
buf[0] = CMD_DMARMB;
PL330_DBGCMD_DUMP(SZ_DMARMB, "\tDMARMB\n");
return SZ_DMARMB;
}
static inline u32 _emit_SEV(unsigned dry_run, u8 buf[], u8 ev)
{
if (dry_run)
return SZ_DMASEV;
buf[0] = CMD_DMASEV;
ev &= 0x1f;
ev <<= 3;
buf[1] = ev;
PL330_DBGCMD_DUMP(SZ_DMASEV, "\tDMASEV %u\n", ev >> 3);
return SZ_DMASEV;
}
static inline u32 _emit_ST(unsigned dry_run, u8 buf[], enum pl330_cond cond)
{
if (dry_run)
return SZ_DMAST;
buf[0] = CMD_DMAST;
if (cond == SINGLE)
buf[0] |= (0 << 1) | (1 << 0);
else if (cond == BURST)
buf[0] |= (1 << 1) | (1 << 0);
PL330_DBGCMD_DUMP(SZ_DMAST, "\tDMAST%c\n",
cond == SINGLE ? 'S' : (cond == BURST ? 'B' : 'A'));
return SZ_DMAST;
}
static inline u32 _emit_STP(unsigned dry_run, u8 buf[],
enum pl330_cond cond, u8 peri)
{
if (dry_run)
return SZ_DMASTP;
buf[0] = CMD_DMASTP;
if (cond == BURST)
buf[0] |= (1 << 1);
peri &= 0x1f;
peri <<= 3;
buf[1] = peri;
PL330_DBGCMD_DUMP(SZ_DMASTP, "\tDMASTP%c %u\n",
cond == SINGLE ? 'S' : 'B', peri >> 3);
return SZ_DMASTP;
}
static inline u32 _emit_WFP(unsigned dry_run, u8 buf[],
enum pl330_cond cond, u8 peri)
{
if (dry_run)
return SZ_DMAWFP;
buf[0] = CMD_DMAWFP;
if (cond == SINGLE)
buf[0] |= (0 << 1) | (0 << 0);
else if (cond == BURST)
buf[0] |= (1 << 1) | (0 << 0);
else
buf[0] |= (0 << 1) | (1 << 0);
peri &= 0x1f;
peri <<= 3;
buf[1] = peri;
PL330_DBGCMD_DUMP(SZ_DMAWFP, "\tDMAWFP%c %u\n",
cond == SINGLE ? 'S' : (cond == BURST ? 'B' : 'P'), peri >> 3);
return SZ_DMAWFP;
}
static inline u32 _emit_WMB(unsigned dry_run, u8 buf[])
{
if (dry_run)
return SZ_DMAWMB;
buf[0] = CMD_DMAWMB;
PL330_DBGCMD_DUMP(SZ_DMAWMB, "\tDMAWMB\n");
return SZ_DMAWMB;
}
struct _arg_GO {
u8 chan;
u32 addr;
unsigned ns;
};
static inline u32 _emit_GO(unsigned dry_run, u8 buf[],
const struct _arg_GO *arg)
{
u8 chan = arg->chan;
u32 addr = arg->addr;
unsigned ns = arg->ns;
if (dry_run)
return SZ_DMAGO;
buf[0] = CMD_DMAGO;
buf[0] |= (ns << 1);
buf[1] = chan & 0x7;
dmaengine: pl330: do not generate unaligned access When PL330 is used with !MMU the following fault is seen: Unhandled fault: alignment exception (0x801) at 0x8f26a002 Internal error: : 801 [#1] ARM Modules linked in: CPU: 0 PID: 640 Comm: dma0chan0-copy0 Not tainted 4.8.0-6a82063-clean+ #1600 Hardware name: ARM-Versatile Express task: 8f1baa80 task.stack: 8e6fe000 PC is at _setup_req+0x4c/0x350 LR is at 0x8f2cbc00 pc : [<801ea538>] lr : [<8f2cbc00>] psr: 60000093 sp : 8e6ffdc0 ip : 00000000 fp : 00000000 r10: 00000000 r9 : 8f2cba10 r8 : 8f2cbc00 r7 : 80000013 r6 : 8f21a050 r5 : 8f21a000 r4 : 8f2ac800 r3 : 8e6ffe18 r2 : 00944251 r1 : ffffffbc r0 : 8f26a000 Flags: nZCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment none Control: 00c5387c Process dma0chan0-copy0 (pid: 640, stack limit = 0x8e6fe210) Stack: (0x8e6ffdc0 to 0x8e700000) fdc0: 00000001 60000093 00000000 8f2cba10 8f26a000 00000004 8f0ae000 8f2cbc00 fde0: 8f0ae000 8f2ac800 8f21a000 8f21a050 80000013 8f2cbc00 8f2cba10 00000000 fe00: 60000093 801ebca0 8e6ffe18 000013ff 40000093 00000000 00944251 8f2ac800 fe20: a0000013 8f2b1320 00001986 00000000 00000001 000013ff 8f1e4f00 8f2cba10 fe40: 8e6fff6c 801e9044 00000003 00000000 fef98c80 002faf07 8e6ffe7c 00000000 fe60: 00000002 00000000 00001986 8f1f158d 8f1e4f00 80568de4 00000002 00000000 fe80: 00001986 8f1f53ff 40000001 80580500 8f1f158d 8001e00c 00000000 cfdfdfdf fea0: fdae2a25 00000001 00000004 8e6fe000 00000008 00000010 00000000 00000005 fec0: 8f2b1330 8f2b1334 8e6ffe80 8e6ffe8c 00001986 00000000 8f21a014 00000001 fee0: 8e6ffe60 8e6ffe78 00000002 00000000 000013ff 00000001 80568de4 8f1e8018 ff00: 0000158d 8055ec30 00000001 803f6b00 00001986 8f2cba10 fdae2a25 00000001 ff20: 8f1baca8 8e6fff24 8e6fff24 00000000 8e6fff24 ac6f3037 00000000 00000000 ff40: 00000000 8e6fe000 8f1e4f40 00000000 8f1e4f40 8f1e4f00 801e84ec 00000000 ff60: 00000000 00000000 00000000 80031714 dfdfdfcf 00000000 dfdfdfcf 8f1e4f00 ff80: 00000000 8e6fff84 8e6fff84 00000000 8e6fff90 8e6fff90 8e6fffac 8f1e4f40 ffa0: 80031640 00000000 00000000 8000f548 00000000 00000000 00000000 00000000 ffc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 ffe0: 00000000 00000000 00000000 00000000 00000013 00000000 dfdfdfcf cfdfdfdf [<801ea538>] (_setup_req) from [<801ebca0>] (pl330_tasklet+0x41c/0x490) [<801ebca0>] (pl330_tasklet) from [<801e9044>] (dmatest_func+0xb58/0x149c) [<801e9044>] (dmatest_func) from [<80031714>] (kthread+0xd4/0xec) [<80031714>] (kthread) from [<8000f548>] (ret_from_fork+0x14/0x2c) Code: e3a03001 e3e01043 e5c03001 e59d3048 (e5802002) This happens because _emit_{ADDH,MOV,GO) accessing to unaligned data while writing to buffer. Fix it with writing to buffer byte by byte. Reviewed-by: Robin Murphy <robin.murphy@arm.com> Tested-by: Robin Murphy <robin.murphy@arm.com> Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2016-12-07 20:17:40 +07:00
buf[2] = addr;
buf[3] = addr >> 8;
buf[4] = addr >> 16;
buf[5] = addr >> 24;
return SZ_DMAGO;
}
#define msecs_to_loops(t) (loops_per_jiffy / 1000 * HZ * t)
/* Returns Time-Out */
static bool _until_dmac_idle(struct pl330_thread *thrd)
{
void __iomem *regs = thrd->dmac->base;
unsigned long loops = msecs_to_loops(5);
do {
/* Until Manager is Idle */
if (!(readl(regs + DBGSTATUS) & DBG_BUSY))
break;
cpu_relax();
} while (--loops);
if (!loops)
return true;
return false;
}
static inline void _execute_DBGINSN(struct pl330_thread *thrd,
u8 insn[], bool as_manager)
{
void __iomem *regs = thrd->dmac->base;
u32 val;
val = (insn[0] << 16) | (insn[1] << 24);
if (!as_manager) {
val |= (1 << 0);
val |= (thrd->id << 8); /* Channel Number */
}
writel(val, regs + DBGINST0);
val = le32_to_cpu(*((__le32 *)&insn[2]));
writel(val, regs + DBGINST1);
/* If timed out due to halted state-machine */
if (_until_dmac_idle(thrd)) {
dev_err(thrd->dmac->ddma.dev, "DMAC halted!\n");
return;
}
/* Get going */
writel(0, regs + DBGCMD);
}
static inline u32 _state(struct pl330_thread *thrd)
{
void __iomem *regs = thrd->dmac->base;
u32 val;
if (is_manager(thrd))
val = readl(regs + DS) & 0xf;
else
val = readl(regs + CS(thrd->id)) & 0xf;
switch (val) {
case DS_ST_STOP:
return PL330_STATE_STOPPED;
case DS_ST_EXEC:
return PL330_STATE_EXECUTING;
case DS_ST_CMISS:
return PL330_STATE_CACHEMISS;
case DS_ST_UPDTPC:
return PL330_STATE_UPDTPC;
case DS_ST_WFE:
return PL330_STATE_WFE;
case DS_ST_FAULT:
return PL330_STATE_FAULTING;
case DS_ST_ATBRR:
if (is_manager(thrd))
return PL330_STATE_INVALID;
else
return PL330_STATE_ATBARRIER;
case DS_ST_QBUSY:
if (is_manager(thrd))
return PL330_STATE_INVALID;
else
return PL330_STATE_QUEUEBUSY;
case DS_ST_WFP:
if (is_manager(thrd))
return PL330_STATE_INVALID;
else
return PL330_STATE_WFP;
case DS_ST_KILL:
if (is_manager(thrd))
return PL330_STATE_INVALID;
else
return PL330_STATE_KILLING;
case DS_ST_CMPLT:
if (is_manager(thrd))
return PL330_STATE_INVALID;
else
return PL330_STATE_COMPLETING;
case DS_ST_FLTCMP:
if (is_manager(thrd))
return PL330_STATE_INVALID;
else
return PL330_STATE_FAULT_COMPLETING;
default:
return PL330_STATE_INVALID;
}
}
static void _stop(struct pl330_thread *thrd)
{
void __iomem *regs = thrd->dmac->base;
u8 insn[6] = {0, 0, 0, 0, 0, 0};
if (_state(thrd) == PL330_STATE_FAULT_COMPLETING)
UNTIL(thrd, PL330_STATE_FAULTING | PL330_STATE_KILLING);
/* Return if nothing needs to be done */
if (_state(thrd) == PL330_STATE_COMPLETING
|| _state(thrd) == PL330_STATE_KILLING
|| _state(thrd) == PL330_STATE_STOPPED)
return;
_emit_KILL(0, insn);
/* Stop generating interrupts for SEV */
writel(readl(regs + INTEN) & ~(1 << thrd->ev), regs + INTEN);
_execute_DBGINSN(thrd, insn, is_manager(thrd));
}
/* Start doing req 'idx' of thread 'thrd' */
static bool _trigger(struct pl330_thread *thrd)
{
void __iomem *regs = thrd->dmac->base;
struct _pl330_req *req;
struct dma_pl330_desc *desc;
struct _arg_GO go;
unsigned ns;
u8 insn[6] = {0, 0, 0, 0, 0, 0};
int idx;
/* Return if already ACTIVE */
if (_state(thrd) != PL330_STATE_STOPPED)
return true;
idx = 1 - thrd->lstenq;
if (thrd->req[idx].desc != NULL) {
req = &thrd->req[idx];
} else {
idx = thrd->lstenq;
if (thrd->req[idx].desc != NULL)
req = &thrd->req[idx];
else
req = NULL;
}
/* Return if no request */
if (!req)
return true;
dmaengine: pl330: fix bug that cause start the same descs in cyclic This bug will cause NULL pointer after commit dfac17, and cause wrong package in I2S DMA transfer before commit dfac17. Tested on RK3288-pinky2 board. Detail: I2S DMA transfer(sound/core/pcm_dmaengine.c): dmaengine_pcm_prepare_and_submit --> dmaengine_prep_dma_cyclic --> pl330_prep_dma_cyclic --> the case: 1. pl330_submit_req(desc0): thrd->req[0].desc = desc0, thrd->lstenq = 0 2. pl330_submit_req(desc1): thrd->req[1].desc = desc1, thrd->lstenq = 1 3. _start(desc0) by submit_req: thrd->req_running = 0 because: idx = 1 - thrd->lstenq = 0 4. pl330_update(desc0 OK): thrd->req[0].desc = NULL, desc0 to req_done list because: idx = active = thrd->req_running = 0 5. _start(desc1) by pl330_update: thrd->req_running = 1 because: idx = 1 - thrd->lstenq = 0, but thrd->req[0].desc == NULL, so: idx = thrd->lstenq = 1 6. pl330_submit_req(desc2): thrd->req[0].desc = desc2, thrd->lstenq = 0 7. _start(desc1) by submit_req: thrd->req_running = 1 because: idx = 1 - thrd->lstenq = 1 Note: _start started the same descs _start should start desc2 here, NOT desc1 8. pl330_update(desc1 OK): thrd->req[1].desc = NULL, desc1 to req_done list because: idx = active = thrd->req_running = 1 9. _start(desc2) by pl330_update : thrd->req_running = 0 because: idx = 1 - thrd->lstenq = 0 10.pl330_update(desc1 OK, NOT desc2): thrd->req[0].desc = NULL, desc2 to req_done list because: idx = active = thrd->req_running = 0 11.pl330_submit_req(desc3): thrd->req[0].desc = desc3, thrd->lstenq = 0 12.pl330_submit_req(desc4): thrd->req[1].desc = desc4, thrd->lstenq = 1 13._start(desc3) by submit_req: thrd->req_running = 0 because: idx = 1 - thrd->lstenq = 0 14.pl330_update(desc2 OK NOT desc3): thrd->req[0].desc = NULL desc3 to req_done list because: idx = active = thrd->req_running = 0 15._start(desc4) by pl330_update: thrd->req_running = 1 because: idx = 1 - thrd->lstenq = 0, but thrd->req[0].desc == NULL, so: idx = thrd->lstenq = 1 16.pl330_submit_req(desc5): thrd->req[0].desc = desc5, thrd->lstenq = 0 17._start(desc4) by submit_req: thrd->req_running = 1 because: idx = 1 - thrd->lstenq = 1 18.pl330_update(desc3 OK NOT desc4): thrd->req[1].desc = NULL desc4 to req_done list because: idx = active = thrd->req_running = 1 19._start(desc4) by pl330_update: thrd->req_running = 0 because: idx = 1 - thrd->lstenq = 1, but thrd->req[1].desc == NULL, so: idx = thrd->lstenq = 0 20.pl330_update(desc4 OK): thrd->req[0].desc = NULL, desc5 to req_done list because: idx = active = thrd->req_running = 0 21.pl330_update(desc4 OK): 1) before commit dfac17(set req_running -1 in pl330_update/mark_free()): because: active = -1, abort result: desc0-desc5's callback are all called, but step 10 and step 18 go wrong. 2) before commit dfac17: idx = active = thrd->req_runnig = 0 --> descdone = thrd->req[0] = NULL --> list_add_tail(&descdone->rqd, &pl330->req_done); --> got NULL pointer!!! Signed-off-by: Addy Ke <addy.ke@rock-chips.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-12-08 18:28:20 +07:00
/* Return if req is running */
if (idx == thrd->req_running)
return true;
desc = req->desc;
ns = desc->rqcfg.nonsecure ? 1 : 0;
/* See 'Abort Sources' point-4 at Page 2-25 */
if (_manager_ns(thrd) && !ns)
dev_info(thrd->dmac->ddma.dev, "%s:%d Recipe for ABORT!\n",
__func__, __LINE__);
go.chan = thrd->id;
go.addr = req->mc_bus;
go.ns = ns;
_emit_GO(0, insn, &go);
/* Set to generate interrupts for SEV */
writel(readl(regs + INTEN) | (1 << thrd->ev), regs + INTEN);
/* Only manager can execute GO */
_execute_DBGINSN(thrd, insn, true);
thrd->req_running = idx;
return true;
}
static bool _start(struct pl330_thread *thrd)
{
switch (_state(thrd)) {
case PL330_STATE_FAULT_COMPLETING:
UNTIL(thrd, PL330_STATE_FAULTING | PL330_STATE_KILLING);
if (_state(thrd) == PL330_STATE_KILLING)
UNTIL(thrd, PL330_STATE_STOPPED)
case PL330_STATE_FAULTING:
_stop(thrd);
case PL330_STATE_KILLING:
case PL330_STATE_COMPLETING:
UNTIL(thrd, PL330_STATE_STOPPED)
case PL330_STATE_STOPPED:
return _trigger(thrd);
case PL330_STATE_WFP:
case PL330_STATE_QUEUEBUSY:
case PL330_STATE_ATBARRIER:
case PL330_STATE_UPDTPC:
case PL330_STATE_CACHEMISS:
case PL330_STATE_EXECUTING:
return true;
case PL330_STATE_WFE: /* For RESUME, nothing yet */
default:
return false;
}
}
static inline int _ldst_memtomem(unsigned dry_run, u8 buf[],
const struct _xfer_spec *pxs, int cyc)
{
int off = 0;
struct pl330_config *pcfg = pxs->desc->rqcfg.pcfg;
/* check lock-up free version */
if (get_revision(pcfg->periph_id) >= PERIPH_REV_R1P0) {
while (cyc--) {
off += _emit_LD(dry_run, &buf[off], ALWAYS);
off += _emit_ST(dry_run, &buf[off], ALWAYS);
}
} else {
while (cyc--) {
off += _emit_LD(dry_run, &buf[off], ALWAYS);
off += _emit_RMB(dry_run, &buf[off]);
off += _emit_ST(dry_run, &buf[off], ALWAYS);
off += _emit_WMB(dry_run, &buf[off]);
}
}
return off;
}
static inline int _ldst_devtomem(struct pl330_dmac *pl330, unsigned dry_run,
u8 buf[], const struct _xfer_spec *pxs,
int cyc)
{
int off = 0;
enum pl330_cond cond;
if (pl330->quirks & PL330_QUIRK_BROKEN_NO_FLUSHP)
cond = BURST;
else
cond = SINGLE;
while (cyc--) {
off += _emit_WFP(dry_run, &buf[off], cond, pxs->desc->peri);
off += _emit_LDP(dry_run, &buf[off], cond, pxs->desc->peri);
off += _emit_ST(dry_run, &buf[off], ALWAYS);
if (!(pl330->quirks & PL330_QUIRK_BROKEN_NO_FLUSHP))
off += _emit_FLUSHP(dry_run, &buf[off],
pxs->desc->peri);
}
return off;
}
static inline int _ldst_memtodev(struct pl330_dmac *pl330,
unsigned dry_run, u8 buf[],
const struct _xfer_spec *pxs, int cyc)
{
int off = 0;
enum pl330_cond cond;
if (pl330->quirks & PL330_QUIRK_BROKEN_NO_FLUSHP)
cond = BURST;
else
cond = SINGLE;
while (cyc--) {
off += _emit_WFP(dry_run, &buf[off], cond, pxs->desc->peri);
off += _emit_LD(dry_run, &buf[off], ALWAYS);
off += _emit_STP(dry_run, &buf[off], cond, pxs->desc->peri);
if (!(pl330->quirks & PL330_QUIRK_BROKEN_NO_FLUSHP))
off += _emit_FLUSHP(dry_run, &buf[off],
pxs->desc->peri);
}
return off;
}
static int _bursts(struct pl330_dmac *pl330, unsigned dry_run, u8 buf[],
const struct _xfer_spec *pxs, int cyc)
{
int off = 0;
switch (pxs->desc->rqtype) {
case DMA_MEM_TO_DEV:
off += _ldst_memtodev(pl330, dry_run, &buf[off], pxs, cyc);
break;
case DMA_DEV_TO_MEM:
off += _ldst_devtomem(pl330, dry_run, &buf[off], pxs, cyc);
break;
case DMA_MEM_TO_MEM:
off += _ldst_memtomem(dry_run, &buf[off], pxs, cyc);
break;
default:
off += 0x40000000; /* Scare off the Client */
break;
}
return off;
}
/* Returns bytes consumed and updates bursts */
static inline int _loop(struct pl330_dmac *pl330, unsigned dry_run, u8 buf[],
unsigned long *bursts, const struct _xfer_spec *pxs)
{
int cyc, cycmax, szlp, szlpend, szbrst, off;
unsigned lcnt0, lcnt1, ljmp0, ljmp1;
struct _arg_LPEND lpend;
if (*bursts == 1)
return _bursts(pl330, dry_run, buf, pxs, 1);
/* Max iterations possible in DMALP is 256 */
if (*bursts >= 256*256) {
lcnt1 = 256;
lcnt0 = 256;
cyc = *bursts / lcnt1 / lcnt0;
} else if (*bursts > 256) {
lcnt1 = 256;
lcnt0 = *bursts / lcnt1;
cyc = 1;
} else {
lcnt1 = *bursts;
lcnt0 = 0;
cyc = 1;
}
szlp = _emit_LP(1, buf, 0, 0);
szbrst = _bursts(pl330, 1, buf, pxs, 1);
lpend.cond = ALWAYS;
lpend.forever = false;
lpend.loop = 0;
lpend.bjump = 0;
szlpend = _emit_LPEND(1, buf, &lpend);
if (lcnt0) {
szlp *= 2;
szlpend *= 2;
}
/*
* Max bursts that we can unroll due to limit on the
* size of backward jump that can be encoded in DMALPEND
* which is 8-bits and hence 255
*/
cycmax = (255 - (szlp + szlpend)) / szbrst;
cyc = (cycmax < cyc) ? cycmax : cyc;
off = 0;
if (lcnt0) {
off += _emit_LP(dry_run, &buf[off], 0, lcnt0);
ljmp0 = off;
}
off += _emit_LP(dry_run, &buf[off], 1, lcnt1);
ljmp1 = off;
off += _bursts(pl330, dry_run, &buf[off], pxs, cyc);
lpend.cond = ALWAYS;
lpend.forever = false;
lpend.loop = 1;
lpend.bjump = off - ljmp1;
off += _emit_LPEND(dry_run, &buf[off], &lpend);
if (lcnt0) {
lpend.cond = ALWAYS;
lpend.forever = false;
lpend.loop = 0;
lpend.bjump = off - ljmp0;
off += _emit_LPEND(dry_run, &buf[off], &lpend);
}
*bursts = lcnt1 * cyc;
if (lcnt0)
*bursts *= lcnt0;
return off;
}
static inline int _setup_loops(struct pl330_dmac *pl330,
unsigned dry_run, u8 buf[],
const struct _xfer_spec *pxs)
{
struct pl330_xfer *x = &pxs->desc->px;
u32 ccr = pxs->ccr;
unsigned long c, bursts = BYTE_TO_BURST(x->bytes, ccr);
int off = 0;
while (bursts) {
c = bursts;
off += _loop(pl330, dry_run, &buf[off], &c, pxs);
bursts -= c;
}
return off;
}
static inline int _setup_xfer(struct pl330_dmac *pl330,
unsigned dry_run, u8 buf[],
const struct _xfer_spec *pxs)
{
struct pl330_xfer *x = &pxs->desc->px;
int off = 0;
/* DMAMOV SAR, x->src_addr */
off += _emit_MOV(dry_run, &buf[off], SAR, x->src_addr);
/* DMAMOV DAR, x->dst_addr */
off += _emit_MOV(dry_run, &buf[off], DAR, x->dst_addr);
/* Setup Loop(s) */
off += _setup_loops(pl330, dry_run, &buf[off], pxs);
return off;
}
/*
* A req is a sequence of one or more xfer units.
* Returns the number of bytes taken to setup the MC for the req.
*/
static int _setup_req(struct pl330_dmac *pl330, unsigned dry_run,
struct pl330_thread *thrd, unsigned index,
struct _xfer_spec *pxs)
{
struct _pl330_req *req = &thrd->req[index];
struct pl330_xfer *x;
u8 *buf = req->mc_cpu;
int off = 0;
PL330_DBGMC_START(req->mc_bus);
/* DMAMOV CCR, ccr */
off += _emit_MOV(dry_run, &buf[off], CCR, pxs->ccr);
x = &pxs->desc->px;
/* Error if xfer length is not aligned at burst size */
if (x->bytes % (BRST_SIZE(pxs->ccr) * BRST_LEN(pxs->ccr)))
return -EINVAL;
off += _setup_xfer(pl330, dry_run, &buf[off], pxs);
/* DMASEV peripheral/event */
off += _emit_SEV(dry_run, &buf[off], thrd->ev);
/* DMAEND */
off += _emit_END(dry_run, &buf[off]);
return off;
}
static inline u32 _prepare_ccr(const struct pl330_reqcfg *rqc)
{
u32 ccr = 0;
if (rqc->src_inc)
ccr |= CC_SRCINC;
if (rqc->dst_inc)
ccr |= CC_DSTINC;
/* We set same protection levels for Src and DST for now */
if (rqc->privileged)
ccr |= CC_SRCPRI | CC_DSTPRI;
if (rqc->nonsecure)
ccr |= CC_SRCNS | CC_DSTNS;
if (rqc->insnaccess)
ccr |= CC_SRCIA | CC_DSTIA;
ccr |= (((rqc->brst_len - 1) & 0xf) << CC_SRCBRSTLEN_SHFT);
ccr |= (((rqc->brst_len - 1) & 0xf) << CC_DSTBRSTLEN_SHFT);
ccr |= (rqc->brst_size << CC_SRCBRSTSIZE_SHFT);
ccr |= (rqc->brst_size << CC_DSTBRSTSIZE_SHFT);
ccr |= (rqc->scctl << CC_SRCCCTRL_SHFT);
ccr |= (rqc->dcctl << CC_DSTCCTRL_SHFT);
ccr |= (rqc->swap << CC_SWAP_SHFT);
return ccr;
}
/*
* Submit a list of xfers after which the client wants notification.
* Client is not notified after each xfer unit, just once after all
* xfer units are done or some error occurs.
*/
static int pl330_submit_req(struct pl330_thread *thrd,
struct dma_pl330_desc *desc)
{
struct pl330_dmac *pl330 = thrd->dmac;
struct _xfer_spec xs;
unsigned long flags;
unsigned idx;
u32 ccr;
int ret = 0;
if (pl330->state == DYING
|| pl330->dmac_tbd.reset_chan & (1 << thrd->id)) {
dev_info(thrd->dmac->ddma.dev, "%s:%d\n",
__func__, __LINE__);
return -EAGAIN;
}
/* If request for non-existing peripheral */
if (desc->rqtype != DMA_MEM_TO_MEM &&
desc->peri >= pl330->pcfg.num_peri) {
dev_info(thrd->dmac->ddma.dev,
"%s:%d Invalid peripheral(%u)!\n",
__func__, __LINE__, desc->peri);
return -EINVAL;
}
spin_lock_irqsave(&pl330->lock, flags);
if (_queue_full(thrd)) {
ret = -EAGAIN;
goto xfer_exit;
}
/* Prefer Secure Channel */
if (!_manager_ns(thrd))
desc->rqcfg.nonsecure = 0;
else
desc->rqcfg.nonsecure = 1;
ccr = _prepare_ccr(&desc->rqcfg);
idx = thrd->req[0].desc == NULL ? 0 : 1;
xs.ccr = ccr;
xs.desc = desc;
/* First dry run to check if req is acceptable */
ret = _setup_req(pl330, 1, thrd, idx, &xs);
if (ret < 0)
goto xfer_exit;
if (ret > pl330->mcbufsz / 2) {
dev_info(pl330->ddma.dev, "%s:%d Try increasing mcbufsz (%i/%i)\n",
__func__, __LINE__, ret, pl330->mcbufsz / 2);
ret = -ENOMEM;
goto xfer_exit;
}
/* Hook the request */
thrd->lstenq = idx;
thrd->req[idx].desc = desc;
_setup_req(pl330, 0, thrd, idx, &xs);
ret = 0;
xfer_exit:
spin_unlock_irqrestore(&pl330->lock, flags);
return ret;
}
static void dma_pl330_rqcb(struct dma_pl330_desc *desc, enum pl330_op_err err)
{
struct dma_pl330_chan *pch;
unsigned long flags;
if (!desc)
return;
pch = desc->pchan;
/* If desc aborted */
if (!pch)
return;
spin_lock_irqsave(&pch->lock, flags);
desc->status = DONE;
spin_unlock_irqrestore(&pch->lock, flags);
tasklet_schedule(&pch->task);
}
static void pl330_dotask(unsigned long data)
{
struct pl330_dmac *pl330 = (struct pl330_dmac *) data;
unsigned long flags;
int i;
spin_lock_irqsave(&pl330->lock, flags);
/* The DMAC itself gone nuts */
if (pl330->dmac_tbd.reset_dmac) {
pl330->state = DYING;
/* Reset the manager too */
pl330->dmac_tbd.reset_mngr = true;
/* Clear the reset flag */
pl330->dmac_tbd.reset_dmac = false;
}
if (pl330->dmac_tbd.reset_mngr) {
_stop(pl330->manager);
/* Reset all channels */
pl330->dmac_tbd.reset_chan = (1 << pl330->pcfg.num_chan) - 1;
/* Clear the reset flag */
pl330->dmac_tbd.reset_mngr = false;
}
for (i = 0; i < pl330->pcfg.num_chan; i++) {
if (pl330->dmac_tbd.reset_chan & (1 << i)) {
struct pl330_thread *thrd = &pl330->channels[i];
void __iomem *regs = pl330->base;
enum pl330_op_err err;
_stop(thrd);
if (readl(regs + FSC) & (1 << thrd->id))
err = PL330_ERR_FAIL;
else
err = PL330_ERR_ABORT;
spin_unlock_irqrestore(&pl330->lock, flags);
dma_pl330_rqcb(thrd->req[1 - thrd->lstenq].desc, err);
dma_pl330_rqcb(thrd->req[thrd->lstenq].desc, err);
spin_lock_irqsave(&pl330->lock, flags);
thrd->req[0].desc = NULL;
thrd->req[1].desc = NULL;
thrd->req_running = -1;
/* Clear the reset flag */
pl330->dmac_tbd.reset_chan &= ~(1 << i);
}
}
spin_unlock_irqrestore(&pl330->lock, flags);
return;
}
/* Returns 1 if state was updated, 0 otherwise */
static int pl330_update(struct pl330_dmac *pl330)
{
struct dma_pl330_desc *descdone, *tmp;
unsigned long flags;
void __iomem *regs;
u32 val;
int id, ev, ret = 0;
regs = pl330->base;
spin_lock_irqsave(&pl330->lock, flags);
val = readl(regs + FSM) & 0x1;
if (val)
pl330->dmac_tbd.reset_mngr = true;
else
pl330->dmac_tbd.reset_mngr = false;
val = readl(regs + FSC) & ((1 << pl330->pcfg.num_chan) - 1);
pl330->dmac_tbd.reset_chan |= val;
if (val) {
int i = 0;
while (i < pl330->pcfg.num_chan) {
if (val & (1 << i)) {
dev_info(pl330->ddma.dev,
"Reset Channel-%d\t CS-%x FTC-%x\n",
i, readl(regs + CS(i)),
readl(regs + FTC(i)));
_stop(&pl330->channels[i]);
}
i++;
}
}
/* Check which event happened i.e, thread notified */
val = readl(regs + ES);
if (pl330->pcfg.num_events < 32
&& val & ~((1 << pl330->pcfg.num_events) - 1)) {
pl330->dmac_tbd.reset_dmac = true;
dev_err(pl330->ddma.dev, "%s:%d Unexpected!\n", __func__,
__LINE__);
ret = 1;
goto updt_exit;
}
for (ev = 0; ev < pl330->pcfg.num_events; ev++) {
if (val & (1 << ev)) { /* Event occurred */
struct pl330_thread *thrd;
u32 inten = readl(regs + INTEN);
int active;
/* Clear the event */
if (inten & (1 << ev))
writel(1 << ev, regs + INTCLR);
ret = 1;
id = pl330->events[ev];
thrd = &pl330->channels[id];
active = thrd->req_running;
if (active == -1) /* Aborted */
continue;
/* Detach the req */
descdone = thrd->req[active].desc;
thrd->req[active].desc = NULL;
dmaengine: pl330: fix bug that cause start the same descs in cyclic This bug will cause NULL pointer after commit dfac17, and cause wrong package in I2S DMA transfer before commit dfac17. Tested on RK3288-pinky2 board. Detail: I2S DMA transfer(sound/core/pcm_dmaengine.c): dmaengine_pcm_prepare_and_submit --> dmaengine_prep_dma_cyclic --> pl330_prep_dma_cyclic --> the case: 1. pl330_submit_req(desc0): thrd->req[0].desc = desc0, thrd->lstenq = 0 2. pl330_submit_req(desc1): thrd->req[1].desc = desc1, thrd->lstenq = 1 3. _start(desc0) by submit_req: thrd->req_running = 0 because: idx = 1 - thrd->lstenq = 0 4. pl330_update(desc0 OK): thrd->req[0].desc = NULL, desc0 to req_done list because: idx = active = thrd->req_running = 0 5. _start(desc1) by pl330_update: thrd->req_running = 1 because: idx = 1 - thrd->lstenq = 0, but thrd->req[0].desc == NULL, so: idx = thrd->lstenq = 1 6. pl330_submit_req(desc2): thrd->req[0].desc = desc2, thrd->lstenq = 0 7. _start(desc1) by submit_req: thrd->req_running = 1 because: idx = 1 - thrd->lstenq = 1 Note: _start started the same descs _start should start desc2 here, NOT desc1 8. pl330_update(desc1 OK): thrd->req[1].desc = NULL, desc1 to req_done list because: idx = active = thrd->req_running = 1 9. _start(desc2) by pl330_update : thrd->req_running = 0 because: idx = 1 - thrd->lstenq = 0 10.pl330_update(desc1 OK, NOT desc2): thrd->req[0].desc = NULL, desc2 to req_done list because: idx = active = thrd->req_running = 0 11.pl330_submit_req(desc3): thrd->req[0].desc = desc3, thrd->lstenq = 0 12.pl330_submit_req(desc4): thrd->req[1].desc = desc4, thrd->lstenq = 1 13._start(desc3) by submit_req: thrd->req_running = 0 because: idx = 1 - thrd->lstenq = 0 14.pl330_update(desc2 OK NOT desc3): thrd->req[0].desc = NULL desc3 to req_done list because: idx = active = thrd->req_running = 0 15._start(desc4) by pl330_update: thrd->req_running = 1 because: idx = 1 - thrd->lstenq = 0, but thrd->req[0].desc == NULL, so: idx = thrd->lstenq = 1 16.pl330_submit_req(desc5): thrd->req[0].desc = desc5, thrd->lstenq = 0 17._start(desc4) by submit_req: thrd->req_running = 1 because: idx = 1 - thrd->lstenq = 1 18.pl330_update(desc3 OK NOT desc4): thrd->req[1].desc = NULL desc4 to req_done list because: idx = active = thrd->req_running = 1 19._start(desc4) by pl330_update: thrd->req_running = 0 because: idx = 1 - thrd->lstenq = 1, but thrd->req[1].desc == NULL, so: idx = thrd->lstenq = 0 20.pl330_update(desc4 OK): thrd->req[0].desc = NULL, desc5 to req_done list because: idx = active = thrd->req_running = 0 21.pl330_update(desc4 OK): 1) before commit dfac17(set req_running -1 in pl330_update/mark_free()): because: active = -1, abort result: desc0-desc5's callback are all called, but step 10 and step 18 go wrong. 2) before commit dfac17: idx = active = thrd->req_runnig = 0 --> descdone = thrd->req[0] = NULL --> list_add_tail(&descdone->rqd, &pl330->req_done); --> got NULL pointer!!! Signed-off-by: Addy Ke <addy.ke@rock-chips.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-12-08 18:28:20 +07:00
thrd->req_running = -1;
/* Get going again ASAP */
_start(thrd);
/* For now, just make a list of callbacks to be done */
list_add_tail(&descdone->rqd, &pl330->req_done);
}
}
/* Now that we are in no hurry, do the callbacks */
list_for_each_entry_safe(descdone, tmp, &pl330->req_done, rqd) {
list_del(&descdone->rqd);
spin_unlock_irqrestore(&pl330->lock, flags);
dma_pl330_rqcb(descdone, PL330_ERR_NONE);
spin_lock_irqsave(&pl330->lock, flags);
}
updt_exit:
spin_unlock_irqrestore(&pl330->lock, flags);
if (pl330->dmac_tbd.reset_dmac
|| pl330->dmac_tbd.reset_mngr
|| pl330->dmac_tbd.reset_chan) {
ret = 1;
tasklet_schedule(&pl330->tasks);
}
return ret;
}
/* Reserve an event */
static inline int _alloc_event(struct pl330_thread *thrd)
{
struct pl330_dmac *pl330 = thrd->dmac;
int ev;
for (ev = 0; ev < pl330->pcfg.num_events; ev++)
if (pl330->events[ev] == -1) {
pl330->events[ev] = thrd->id;
return ev;
}
return -1;
}
static bool _chan_ns(const struct pl330_dmac *pl330, int i)
{
return pl330->pcfg.irq_ns & (1 << i);
}
/* Upon success, returns IdentityToken for the
* allocated channel, NULL otherwise.
*/
static struct pl330_thread *pl330_request_channel(struct pl330_dmac *pl330)
{
struct pl330_thread *thrd = NULL;
int chans, i;
if (pl330->state == DYING)
return NULL;
chans = pl330->pcfg.num_chan;
for (i = 0; i < chans; i++) {
thrd = &pl330->channels[i];
if ((thrd->free) && (!_manager_ns(thrd) ||
_chan_ns(pl330, i))) {
thrd->ev = _alloc_event(thrd);
if (thrd->ev >= 0) {
thrd->free = false;
thrd->lstenq = 1;
thrd->req[0].desc = NULL;
thrd->req[1].desc = NULL;
thrd->req_running = -1;
break;
}
}
thrd = NULL;
}
return thrd;
}
/* Release an event */
static inline void _free_event(struct pl330_thread *thrd, int ev)
{
struct pl330_dmac *pl330 = thrd->dmac;
/* If the event is valid and was held by the thread */
if (ev >= 0 && ev < pl330->pcfg.num_events
&& pl330->events[ev] == thrd->id)
pl330->events[ev] = -1;
}
static void pl330_release_channel(struct pl330_thread *thrd)
{
struct pl330_dmac *pl330;
if (!thrd || thrd->free)
return;
_stop(thrd);
dma_pl330_rqcb(thrd->req[1 - thrd->lstenq].desc, PL330_ERR_ABORT);
dma_pl330_rqcb(thrd->req[thrd->lstenq].desc, PL330_ERR_ABORT);
pl330 = thrd->dmac;
_free_event(thrd, thrd->ev);
thrd->free = true;
}
/* Initialize the structure for PL330 configuration, that can be used
* by the client driver the make best use of the DMAC
*/
static void read_dmac_config(struct pl330_dmac *pl330)
{
void __iomem *regs = pl330->base;
u32 val;
val = readl(regs + CRD) >> CRD_DATA_WIDTH_SHIFT;
val &= CRD_DATA_WIDTH_MASK;
pl330->pcfg.data_bus_width = 8 * (1 << val);
val = readl(regs + CRD) >> CRD_DATA_BUFF_SHIFT;
val &= CRD_DATA_BUFF_MASK;
pl330->pcfg.data_buf_dep = val + 1;
val = readl(regs + CR0) >> CR0_NUM_CHANS_SHIFT;
val &= CR0_NUM_CHANS_MASK;
val += 1;
pl330->pcfg.num_chan = val;
val = readl(regs + CR0);
if (val & CR0_PERIPH_REQ_SET) {
val = (val >> CR0_NUM_PERIPH_SHIFT) & CR0_NUM_PERIPH_MASK;
val += 1;
pl330->pcfg.num_peri = val;
pl330->pcfg.peri_ns = readl(regs + CR4);
} else {
pl330->pcfg.num_peri = 0;
}
val = readl(regs + CR0);
if (val & CR0_BOOT_MAN_NS)
pl330->pcfg.mode |= DMAC_MODE_NS;
else
pl330->pcfg.mode &= ~DMAC_MODE_NS;
val = readl(regs + CR0) >> CR0_NUM_EVENTS_SHIFT;
val &= CR0_NUM_EVENTS_MASK;
val += 1;
pl330->pcfg.num_events = val;
pl330->pcfg.irq_ns = readl(regs + CR3);
}
static inline void _reset_thread(struct pl330_thread *thrd)
{
struct pl330_dmac *pl330 = thrd->dmac;
thrd->req[0].mc_cpu = pl330->mcode_cpu
+ (thrd->id * pl330->mcbufsz);
thrd->req[0].mc_bus = pl330->mcode_bus
+ (thrd->id * pl330->mcbufsz);
thrd->req[0].desc = NULL;
thrd->req[1].mc_cpu = thrd->req[0].mc_cpu
+ pl330->mcbufsz / 2;
thrd->req[1].mc_bus = thrd->req[0].mc_bus
+ pl330->mcbufsz / 2;
thrd->req[1].desc = NULL;
thrd->req_running = -1;
}
static int dmac_alloc_threads(struct pl330_dmac *pl330)
{
int chans = pl330->pcfg.num_chan;
struct pl330_thread *thrd;
int i;
/* Allocate 1 Manager and 'chans' Channel threads */
pl330->channels = kzalloc((1 + chans) * sizeof(*thrd),
GFP_KERNEL);
if (!pl330->channels)
return -ENOMEM;
/* Init Channel threads */
for (i = 0; i < chans; i++) {
thrd = &pl330->channels[i];
thrd->id = i;
thrd->dmac = pl330;
_reset_thread(thrd);
thrd->free = true;
}
/* MANAGER is indexed at the end */
thrd = &pl330->channels[chans];
thrd->id = chans;
thrd->dmac = pl330;
thrd->free = false;
pl330->manager = thrd;
return 0;
}
static int dmac_alloc_resources(struct pl330_dmac *pl330)
{
int chans = pl330->pcfg.num_chan;
int ret;
/*
* Alloc MicroCode buffer for 'chans' Channel threads.
* A channel's buffer offset is (Channel_Id * MCODE_BUFF_PERCHAN)
*/
pl330->mcode_cpu = dma_alloc_attrs(pl330->ddma.dev,
chans * pl330->mcbufsz,
&pl330->mcode_bus, GFP_KERNEL,
DMA_ATTR_PRIVILEGED);
if (!pl330->mcode_cpu) {
dev_err(pl330->ddma.dev, "%s:%d Can't allocate memory!\n",
__func__, __LINE__);
return -ENOMEM;
}
ret = dmac_alloc_threads(pl330);
if (ret) {
dev_err(pl330->ddma.dev, "%s:%d Can't to create channels for DMAC!\n",
__func__, __LINE__);
dma_free_coherent(pl330->ddma.dev,
chans * pl330->mcbufsz,
pl330->mcode_cpu, pl330->mcode_bus);
return ret;
}
return 0;
}
static int pl330_add(struct pl330_dmac *pl330)
{
int i, ret;
/* Check if we can handle this DMAC */
if ((pl330->pcfg.periph_id & 0xfffff) != PERIPH_ID_VAL) {
dev_err(pl330->ddma.dev, "PERIPH_ID 0x%x !\n",
pl330->pcfg.periph_id);
return -EINVAL;
}
/* Read the configuration of the DMAC */
read_dmac_config(pl330);
if (pl330->pcfg.num_events == 0) {
dev_err(pl330->ddma.dev, "%s:%d Can't work without events!\n",
__func__, __LINE__);
return -EINVAL;
}
spin_lock_init(&pl330->lock);
INIT_LIST_HEAD(&pl330->req_done);
/* Use default MC buffer size if not provided */
if (!pl330->mcbufsz)
pl330->mcbufsz = MCODE_BUFF_PER_REQ * 2;
/* Mark all events as free */
for (i = 0; i < pl330->pcfg.num_events; i++)
pl330->events[i] = -1;
/* Allocate resources needed by the DMAC */
ret = dmac_alloc_resources(pl330);
if (ret) {
dev_err(pl330->ddma.dev, "Unable to create channels for DMAC\n");
return ret;
}
tasklet_init(&pl330->tasks, pl330_dotask, (unsigned long) pl330);
pl330->state = INIT;
return 0;
}
static int dmac_free_threads(struct pl330_dmac *pl330)
{
struct pl330_thread *thrd;
int i;
/* Release Channel threads */
for (i = 0; i < pl330->pcfg.num_chan; i++) {
thrd = &pl330->channels[i];
pl330_release_channel(thrd);
}
/* Free memory */
kfree(pl330->channels);
return 0;
}
static void pl330_del(struct pl330_dmac *pl330)
{
pl330->state = UNINIT;
tasklet_kill(&pl330->tasks);
/* Free DMAC resources */
dmac_free_threads(pl330);
dma_free_coherent(pl330->ddma.dev,
pl330->pcfg.num_chan * pl330->mcbufsz, pl330->mcode_cpu,
pl330->mcode_bus);
}
/* forward declaration */
static struct amba_driver pl330_driver;
static inline struct dma_pl330_chan *
to_pchan(struct dma_chan *ch)
{
if (!ch)
return NULL;
return container_of(ch, struct dma_pl330_chan, chan);
}
static inline struct dma_pl330_desc *
to_desc(struct dma_async_tx_descriptor *tx)
{
return container_of(tx, struct dma_pl330_desc, txd);
}
static inline void fill_queue(struct dma_pl330_chan *pch)
{
struct dma_pl330_desc *desc;
int ret;
list_for_each_entry(desc, &pch->work_list, node) {
/* If already submitted */
if (desc->status == BUSY)
continue;
ret = pl330_submit_req(pch->thread, desc);
if (!ret) {
desc->status = BUSY;
} else if (ret == -EAGAIN) {
/* QFull or DMAC Dying */
break;
} else {
/* Unacceptable request */
desc->status = DONE;
dev_err(pch->dmac->ddma.dev, "%s:%d Bad Desc(%d)\n",
__func__, __LINE__, desc->txd.cookie);
tasklet_schedule(&pch->task);
}
}
}
static void pl330_tasklet(unsigned long data)
{
struct dma_pl330_chan *pch = (struct dma_pl330_chan *)data;
struct dma_pl330_desc *desc, *_dt;
unsigned long flags;
bool power_down = false;
spin_lock_irqsave(&pch->lock, flags);
/* Pick up ripe tomatoes */
list_for_each_entry_safe(desc, _dt, &pch->work_list, node)
if (desc->status == DONE) {
if (!pch->cyclic)
dma_cookie_complete(&desc->txd);
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
list_move_tail(&desc->node, &pch->completed_list);
}
/* Try to submit a req imm. next to the last completed cookie */
fill_queue(pch);
if (list_empty(&pch->work_list)) {
spin_lock(&pch->thread->dmac->lock);
_stop(pch->thread);
spin_unlock(&pch->thread->dmac->lock);
power_down = true;
pch->active = false;
} else {
/* Make sure the PL330 Channel thread is active */
spin_lock(&pch->thread->dmac->lock);
_start(pch->thread);
spin_unlock(&pch->thread->dmac->lock);
}
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
while (!list_empty(&pch->completed_list)) {
struct dmaengine_desc_callback cb;
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
desc = list_first_entry(&pch->completed_list,
struct dma_pl330_desc, node);
dmaengine_desc_get_callback(&desc->txd, &cb);
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
if (pch->cyclic) {
desc->status = PREP;
list_move_tail(&desc->node, &pch->work_list);
if (power_down) {
pch->active = true;
spin_lock(&pch->thread->dmac->lock);
_start(pch->thread);
spin_unlock(&pch->thread->dmac->lock);
power_down = false;
}
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
} else {
desc->status = FREE;
list_move_tail(&desc->node, &pch->dmac->desc_pool);
}
dma_descriptor_unmap(&desc->txd);
if (dmaengine_desc_callback_valid(&cb)) {
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
spin_unlock_irqrestore(&pch->lock, flags);
dmaengine_desc_callback_invoke(&cb, NULL);
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
spin_lock_irqsave(&pch->lock, flags);
}
}
spin_unlock_irqrestore(&pch->lock, flags);
/* If work list empty, power down */
if (power_down) {
pm_runtime_mark_last_busy(pch->dmac->ddma.dev);
pm_runtime_put_autosuspend(pch->dmac->ddma.dev);
}
}
static struct dma_chan *of_dma_pl330_xlate(struct of_phandle_args *dma_spec,
struct of_dma *ofdma)
{
int count = dma_spec->args_count;
struct pl330_dmac *pl330 = ofdma->of_dma_data;
unsigned int chan_id;
if (!pl330)
return NULL;
if (count != 1)
return NULL;
chan_id = dma_spec->args[0];
if (chan_id >= pl330->num_peripherals)
return NULL;
return dma_get_slave_channel(&pl330->peripherals[chan_id].chan);
}
static int pl330_alloc_chan_resources(struct dma_chan *chan)
{
struct dma_pl330_chan *pch = to_pchan(chan);
struct pl330_dmac *pl330 = pch->dmac;
unsigned long flags;
dmaengine: pl330: fix double lock The static bug finder EBA (http://www.iagoabal.eu/eba/) reported the following double-lock bug: Double lock: 1. spin_lock_irqsave(pch->lock, flags) at pl330_free_chan_resources:2236; 2. call to function `pl330_release_channel' immediately after; 3. call to function `dma_pl330_rqcb' in line 1753; 4. spin_lock_irqsave(pch->lock, flags) at dma_pl330_rqcb:1505. I have fixed it as suggested by Marek Szyprowski. First, I have replaced `pch->lock' with `pl330->lock' in functions `pl330_alloc_chan_resources' and `pl330_free_chan_resources'. This avoids the double-lock by acquiring a different lock than `dma_pl330_rqcb'. NOTE that, as a result, `pl330_free_chan_resources' executes `list_splice_tail_init' on `pch->work_list' under lock `pl330->lock', whereas in the rest of the code `pch->work_list' is protected by `pch->lock'. I don't know if this may cause race conditions. Similarly `pch->cyclic' is written by `pl330_alloc_chan_resources' under `pl330->lock' but read by `pl330_tx_submit' under `pch->lock'. Second, I have removed locking from `pl330_request_channel' and `pl330_release_channel' functions. Function `pl330_request_channel' is only called from `pl330_alloc_chan_resources', so the lock is already held. Function `pl330_release_channel' is called from `pl330_free_chan_resources', which already holds the lock, and from `pl330_del'. Function `pl330_del' is called in an error path of `pl330_probe' and at the end of `pl330_remove', but I assume that there cannot be concurrent accesses to the protected data at those points. Signed-off-by: Iago Abal <mail@iagoabal.eu> Reviewed-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-11 20:00:21 +07:00
spin_lock_irqsave(&pl330->lock, flags);
dma_cookie_init(chan);
pch->cyclic = false;
pch->thread = pl330_request_channel(pl330);
if (!pch->thread) {
dmaengine: pl330: fix double lock The static bug finder EBA (http://www.iagoabal.eu/eba/) reported the following double-lock bug: Double lock: 1. spin_lock_irqsave(pch->lock, flags) at pl330_free_chan_resources:2236; 2. call to function `pl330_release_channel' immediately after; 3. call to function `dma_pl330_rqcb' in line 1753; 4. spin_lock_irqsave(pch->lock, flags) at dma_pl330_rqcb:1505. I have fixed it as suggested by Marek Szyprowski. First, I have replaced `pch->lock' with `pl330->lock' in functions `pl330_alloc_chan_resources' and `pl330_free_chan_resources'. This avoids the double-lock by acquiring a different lock than `dma_pl330_rqcb'. NOTE that, as a result, `pl330_free_chan_resources' executes `list_splice_tail_init' on `pch->work_list' under lock `pl330->lock', whereas in the rest of the code `pch->work_list' is protected by `pch->lock'. I don't know if this may cause race conditions. Similarly `pch->cyclic' is written by `pl330_alloc_chan_resources' under `pl330->lock' but read by `pl330_tx_submit' under `pch->lock'. Second, I have removed locking from `pl330_request_channel' and `pl330_release_channel' functions. Function `pl330_request_channel' is only called from `pl330_alloc_chan_resources', so the lock is already held. Function `pl330_release_channel' is called from `pl330_free_chan_resources', which already holds the lock, and from `pl330_del'. Function `pl330_del' is called in an error path of `pl330_probe' and at the end of `pl330_remove', but I assume that there cannot be concurrent accesses to the protected data at those points. Signed-off-by: Iago Abal <mail@iagoabal.eu> Reviewed-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-11 20:00:21 +07:00
spin_unlock_irqrestore(&pl330->lock, flags);
return -ENOMEM;
}
tasklet_init(&pch->task, pl330_tasklet, (unsigned long) pch);
dmaengine: pl330: fix double lock The static bug finder EBA (http://www.iagoabal.eu/eba/) reported the following double-lock bug: Double lock: 1. spin_lock_irqsave(pch->lock, flags) at pl330_free_chan_resources:2236; 2. call to function `pl330_release_channel' immediately after; 3. call to function `dma_pl330_rqcb' in line 1753; 4. spin_lock_irqsave(pch->lock, flags) at dma_pl330_rqcb:1505. I have fixed it as suggested by Marek Szyprowski. First, I have replaced `pch->lock' with `pl330->lock' in functions `pl330_alloc_chan_resources' and `pl330_free_chan_resources'. This avoids the double-lock by acquiring a different lock than `dma_pl330_rqcb'. NOTE that, as a result, `pl330_free_chan_resources' executes `list_splice_tail_init' on `pch->work_list' under lock `pl330->lock', whereas in the rest of the code `pch->work_list' is protected by `pch->lock'. I don't know if this may cause race conditions. Similarly `pch->cyclic' is written by `pl330_alloc_chan_resources' under `pl330->lock' but read by `pl330_tx_submit' under `pch->lock'. Second, I have removed locking from `pl330_request_channel' and `pl330_release_channel' functions. Function `pl330_request_channel' is only called from `pl330_alloc_chan_resources', so the lock is already held. Function `pl330_release_channel' is called from `pl330_free_chan_resources', which already holds the lock, and from `pl330_del'. Function `pl330_del' is called in an error path of `pl330_probe' and at the end of `pl330_remove', but I assume that there cannot be concurrent accesses to the protected data at those points. Signed-off-by: Iago Abal <mail@iagoabal.eu> Reviewed-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-11 20:00:21 +07:00
spin_unlock_irqrestore(&pl330->lock, flags);
return 1;
}
/*
* We need the data direction between the DMAC (the dma-mapping "device") and
* the FIFO (the dmaengine "dev"), from the FIFO's point of view. Confusing!
*/
static enum dma_data_direction
pl330_dma_slave_map_dir(enum dma_transfer_direction dir)
{
switch (dir) {
case DMA_MEM_TO_DEV:
return DMA_FROM_DEVICE;
case DMA_DEV_TO_MEM:
return DMA_TO_DEVICE;
case DMA_DEV_TO_DEV:
return DMA_BIDIRECTIONAL;
default:
return DMA_NONE;
}
}
static void pl330_unprep_slave_fifo(struct dma_pl330_chan *pch)
{
if (pch->dir != DMA_NONE)
dma_unmap_resource(pch->chan.device->dev, pch->fifo_dma,
1 << pch->burst_sz, pch->dir, 0);
pch->dir = DMA_NONE;
}
static bool pl330_prep_slave_fifo(struct dma_pl330_chan *pch,
enum dma_transfer_direction dir)
{
struct device *dev = pch->chan.device->dev;
enum dma_data_direction dma_dir = pl330_dma_slave_map_dir(dir);
/* Already mapped for this config? */
if (pch->dir == dma_dir)
return true;
pl330_unprep_slave_fifo(pch);
pch->fifo_dma = dma_map_resource(dev, pch->fifo_addr,
1 << pch->burst_sz, dma_dir, 0);
if (dma_mapping_error(dev, pch->fifo_dma))
return false;
pch->dir = dma_dir;
return true;
}
static int pl330_config(struct dma_chan *chan,
struct dma_slave_config *slave_config)
{
struct dma_pl330_chan *pch = to_pchan(chan);
pl330_unprep_slave_fifo(pch);
if (slave_config->direction == DMA_MEM_TO_DEV) {
if (slave_config->dst_addr)
pch->fifo_addr = slave_config->dst_addr;
if (slave_config->dst_addr_width)
pch->burst_sz = __ffs(slave_config->dst_addr_width);
if (slave_config->dst_maxburst)
pch->burst_len = slave_config->dst_maxburst;
} else if (slave_config->direction == DMA_DEV_TO_MEM) {
if (slave_config->src_addr)
pch->fifo_addr = slave_config->src_addr;
if (slave_config->src_addr_width)
pch->burst_sz = __ffs(slave_config->src_addr_width);
if (slave_config->src_maxburst)
pch->burst_len = slave_config->src_maxburst;
}
return 0;
}
static int pl330_terminate_all(struct dma_chan *chan)
{
struct dma_pl330_chan *pch = to_pchan(chan);
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
struct dma_pl330_desc *desc;
unsigned long flags;
struct pl330_dmac *pl330 = pch->dmac;
LIST_HEAD(list);
bool power_down = false;
pm_runtime_get_sync(pl330->ddma.dev);
spin_lock_irqsave(&pch->lock, flags);
spin_lock(&pl330->lock);
_stop(pch->thread);
spin_unlock(&pl330->lock);
pch->thread->req[0].desc = NULL;
pch->thread->req[1].desc = NULL;
pch->thread->req_running = -1;
power_down = pch->active;
pch->active = false;
/* Mark all desc done */
list_for_each_entry(desc, &pch->submitted_list, node) {
desc->status = FREE;
dma_cookie_complete(&desc->txd);
}
list_for_each_entry(desc, &pch->work_list , node) {
desc->status = FREE;
dma_cookie_complete(&desc->txd);
}
list_splice_tail_init(&pch->submitted_list, &pl330->desc_pool);
list_splice_tail_init(&pch->work_list, &pl330->desc_pool);
list_splice_tail_init(&pch->completed_list, &pl330->desc_pool);
spin_unlock_irqrestore(&pch->lock, flags);
pm_runtime_mark_last_busy(pl330->ddma.dev);
if (power_down)
pm_runtime_put_autosuspend(pl330->ddma.dev);
pm_runtime_put_autosuspend(pl330->ddma.dev);
return 0;
}
/*
* We don't support DMA_RESUME command because of hardware
* limitations, so after pausing the channel we cannot restore
* it to active state. We have to terminate channel and setup
* DMA transfer again. This pause feature was implemented to
* allow safely read residue before channel termination.
*/
static int pl330_pause(struct dma_chan *chan)
{
struct dma_pl330_chan *pch = to_pchan(chan);
struct pl330_dmac *pl330 = pch->dmac;
unsigned long flags;
pm_runtime_get_sync(pl330->ddma.dev);
spin_lock_irqsave(&pch->lock, flags);
spin_lock(&pl330->lock);
_stop(pch->thread);
spin_unlock(&pl330->lock);
spin_unlock_irqrestore(&pch->lock, flags);
pm_runtime_mark_last_busy(pl330->ddma.dev);
pm_runtime_put_autosuspend(pl330->ddma.dev);
return 0;
}
static void pl330_free_chan_resources(struct dma_chan *chan)
{
struct dma_pl330_chan *pch = to_pchan(chan);
dmaengine: pl330: fix double lock The static bug finder EBA (http://www.iagoabal.eu/eba/) reported the following double-lock bug: Double lock: 1. spin_lock_irqsave(pch->lock, flags) at pl330_free_chan_resources:2236; 2. call to function `pl330_release_channel' immediately after; 3. call to function `dma_pl330_rqcb' in line 1753; 4. spin_lock_irqsave(pch->lock, flags) at dma_pl330_rqcb:1505. I have fixed it as suggested by Marek Szyprowski. First, I have replaced `pch->lock' with `pl330->lock' in functions `pl330_alloc_chan_resources' and `pl330_free_chan_resources'. This avoids the double-lock by acquiring a different lock than `dma_pl330_rqcb'. NOTE that, as a result, `pl330_free_chan_resources' executes `list_splice_tail_init' on `pch->work_list' under lock `pl330->lock', whereas in the rest of the code `pch->work_list' is protected by `pch->lock'. I don't know if this may cause race conditions. Similarly `pch->cyclic' is written by `pl330_alloc_chan_resources' under `pl330->lock' but read by `pl330_tx_submit' under `pch->lock'. Second, I have removed locking from `pl330_request_channel' and `pl330_release_channel' functions. Function `pl330_request_channel' is only called from `pl330_alloc_chan_resources', so the lock is already held. Function `pl330_release_channel' is called from `pl330_free_chan_resources', which already holds the lock, and from `pl330_del'. Function `pl330_del' is called in an error path of `pl330_probe' and at the end of `pl330_remove', but I assume that there cannot be concurrent accesses to the protected data at those points. Signed-off-by: Iago Abal <mail@iagoabal.eu> Reviewed-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-11 20:00:21 +07:00
struct pl330_dmac *pl330 = pch->dmac;
unsigned long flags;
tasklet_kill(&pch->task);
pm_runtime_get_sync(pch->dmac->ddma.dev);
dmaengine: pl330: fix double lock The static bug finder EBA (http://www.iagoabal.eu/eba/) reported the following double-lock bug: Double lock: 1. spin_lock_irqsave(pch->lock, flags) at pl330_free_chan_resources:2236; 2. call to function `pl330_release_channel' immediately after; 3. call to function `dma_pl330_rqcb' in line 1753; 4. spin_lock_irqsave(pch->lock, flags) at dma_pl330_rqcb:1505. I have fixed it as suggested by Marek Szyprowski. First, I have replaced `pch->lock' with `pl330->lock' in functions `pl330_alloc_chan_resources' and `pl330_free_chan_resources'. This avoids the double-lock by acquiring a different lock than `dma_pl330_rqcb'. NOTE that, as a result, `pl330_free_chan_resources' executes `list_splice_tail_init' on `pch->work_list' under lock `pl330->lock', whereas in the rest of the code `pch->work_list' is protected by `pch->lock'. I don't know if this may cause race conditions. Similarly `pch->cyclic' is written by `pl330_alloc_chan_resources' under `pl330->lock' but read by `pl330_tx_submit' under `pch->lock'. Second, I have removed locking from `pl330_request_channel' and `pl330_release_channel' functions. Function `pl330_request_channel' is only called from `pl330_alloc_chan_resources', so the lock is already held. Function `pl330_release_channel' is called from `pl330_free_chan_resources', which already holds the lock, and from `pl330_del'. Function `pl330_del' is called in an error path of `pl330_probe' and at the end of `pl330_remove', but I assume that there cannot be concurrent accesses to the protected data at those points. Signed-off-by: Iago Abal <mail@iagoabal.eu> Reviewed-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-11 20:00:21 +07:00
spin_lock_irqsave(&pl330->lock, flags);
pl330_release_channel(pch->thread);
pch->thread = NULL;
if (pch->cyclic)
list_splice_tail_init(&pch->work_list, &pch->dmac->desc_pool);
dmaengine: pl330: fix double lock The static bug finder EBA (http://www.iagoabal.eu/eba/) reported the following double-lock bug: Double lock: 1. spin_lock_irqsave(pch->lock, flags) at pl330_free_chan_resources:2236; 2. call to function `pl330_release_channel' immediately after; 3. call to function `dma_pl330_rqcb' in line 1753; 4. spin_lock_irqsave(pch->lock, flags) at dma_pl330_rqcb:1505. I have fixed it as suggested by Marek Szyprowski. First, I have replaced `pch->lock' with `pl330->lock' in functions `pl330_alloc_chan_resources' and `pl330_free_chan_resources'. This avoids the double-lock by acquiring a different lock than `dma_pl330_rqcb'. NOTE that, as a result, `pl330_free_chan_resources' executes `list_splice_tail_init' on `pch->work_list' under lock `pl330->lock', whereas in the rest of the code `pch->work_list' is protected by `pch->lock'. I don't know if this may cause race conditions. Similarly `pch->cyclic' is written by `pl330_alloc_chan_resources' under `pl330->lock' but read by `pl330_tx_submit' under `pch->lock'. Second, I have removed locking from `pl330_request_channel' and `pl330_release_channel' functions. Function `pl330_request_channel' is only called from `pl330_alloc_chan_resources', so the lock is already held. Function `pl330_release_channel' is called from `pl330_free_chan_resources', which already holds the lock, and from `pl330_del'. Function `pl330_del' is called in an error path of `pl330_probe' and at the end of `pl330_remove', but I assume that there cannot be concurrent accesses to the protected data at those points. Signed-off-by: Iago Abal <mail@iagoabal.eu> Reviewed-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2017-01-11 20:00:21 +07:00
spin_unlock_irqrestore(&pl330->lock, flags);
pm_runtime_mark_last_busy(pch->dmac->ddma.dev);
pm_runtime_put_autosuspend(pch->dmac->ddma.dev);
pl330_unprep_slave_fifo(pch);
}
static int pl330_get_current_xferred_count(struct dma_pl330_chan *pch,
struct dma_pl330_desc *desc)
{
struct pl330_thread *thrd = pch->thread;
struct pl330_dmac *pl330 = pch->dmac;
void __iomem *regs = thrd->dmac->base;
u32 val, addr;
pm_runtime_get_sync(pl330->ddma.dev);
val = addr = 0;
if (desc->rqcfg.src_inc) {
val = readl(regs + SA(thrd->id));
addr = desc->px.src_addr;
} else {
val = readl(regs + DA(thrd->id));
addr = desc->px.dst_addr;
}
pm_runtime_mark_last_busy(pch->dmac->ddma.dev);
pm_runtime_put_autosuspend(pl330->ddma.dev);
/* If DMAMOV hasn't finished yet, SAR/DAR can be zero */
if (!val)
return 0;
return val - addr;
}
static enum dma_status
pl330_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
struct dma_tx_state *txstate)
{
enum dma_status ret;
unsigned long flags;
struct dma_pl330_desc *desc, *running = NULL, *last_enq = NULL;
struct dma_pl330_chan *pch = to_pchan(chan);
unsigned int transferred, residual = 0;
ret = dma_cookie_status(chan, cookie, txstate);
if (!txstate)
return ret;
if (ret == DMA_COMPLETE)
goto out;
spin_lock_irqsave(&pch->lock, flags);
spin_lock(&pch->thread->dmac->lock);
if (pch->thread->req_running != -1)
running = pch->thread->req[pch->thread->req_running].desc;
last_enq = pch->thread->req[pch->thread->lstenq].desc;
/* Check in pending list */
list_for_each_entry(desc, &pch->work_list, node) {
if (desc->status == DONE)
transferred = desc->bytes_requested;
else if (running && desc == running)
transferred =
pl330_get_current_xferred_count(pch, desc);
else if (desc->status == BUSY)
/*
* Busy but not running means either just enqueued,
* or finished and not yet marked done
*/
if (desc == last_enq)
transferred = 0;
else
transferred = desc->bytes_requested;
else
transferred = 0;
residual += desc->bytes_requested - transferred;
if (desc->txd.cookie == cookie) {
switch (desc->status) {
case DONE:
ret = DMA_COMPLETE;
break;
case PREP:
case BUSY:
ret = DMA_IN_PROGRESS;
break;
default:
WARN_ON(1);
}
break;
}
if (desc->last)
residual = 0;
}
spin_unlock(&pch->thread->dmac->lock);
spin_unlock_irqrestore(&pch->lock, flags);
out:
dma_set_residue(txstate, residual);
return ret;
}
static void pl330_issue_pending(struct dma_chan *chan)
{
struct dma_pl330_chan *pch = to_pchan(chan);
unsigned long flags;
spin_lock_irqsave(&pch->lock, flags);
if (list_empty(&pch->work_list)) {
/*
* Warn on nothing pending. Empty submitted_list may
* break our pm_runtime usage counter as it is
* updated on work_list emptiness status.
*/
WARN_ON(list_empty(&pch->submitted_list));
pch->active = true;
pm_runtime_get_sync(pch->dmac->ddma.dev);
}
list_splice_tail_init(&pch->submitted_list, &pch->work_list);
spin_unlock_irqrestore(&pch->lock, flags);
pl330_tasklet((unsigned long)pch);
}
/*
* We returned the last one of the circular list of descriptor(s)
* from prep_xxx, so the argument to submit corresponds to the last
* descriptor of the list.
*/
static dma_cookie_t pl330_tx_submit(struct dma_async_tx_descriptor *tx)
{
struct dma_pl330_desc *desc, *last = to_desc(tx);
struct dma_pl330_chan *pch = to_pchan(tx->chan);
dma_cookie_t cookie;
unsigned long flags;
spin_lock_irqsave(&pch->lock, flags);
/* Assign cookies to all nodes */
while (!list_empty(&last->node)) {
desc = list_entry(last->node.next, struct dma_pl330_desc, node);
if (pch->cyclic) {
desc->txd.callback = last->txd.callback;
desc->txd.callback_param = last->txd.callback_param;
}
desc->last = false;
dma_cookie_assign(&desc->txd);
list_move_tail(&desc->node, &pch->submitted_list);
}
last->last = true;
cookie = dma_cookie_assign(&last->txd);
list_add_tail(&last->node, &pch->submitted_list);
spin_unlock_irqrestore(&pch->lock, flags);
return cookie;
}
static inline void _init_desc(struct dma_pl330_desc *desc)
{
desc->rqcfg.swap = SWAP_NO;
desc->rqcfg.scctl = CCTRL0;
desc->rqcfg.dcctl = CCTRL0;
desc->txd.tx_submit = pl330_tx_submit;
INIT_LIST_HEAD(&desc->node);
}
/* Returns the number of descriptors added to the DMAC pool */
static int add_desc(struct list_head *pool, spinlock_t *lock,
gfp_t flg, int count)
{
struct dma_pl330_desc *desc;
unsigned long flags;
int i;
dma: pl330: ensure DMA descriptors are zero-initialised I see the following splat with 3.13-rc1 when attempting to perform DMA: [ 253.004516] Alignment trap: not handling instruction e1902f9f at [<c0204b40>] [ 253.004583] Unhandled fault: alignment exception (0x221) at 0xdfdfdfd7 [ 253.004646] Internal error: : 221 [#1] PREEMPT SMP ARM [ 253.004691] Modules linked in: dmatest(+) [last unloaded: dmatest] [ 253.004798] CPU: 0 PID: 671 Comm: kthreadd Not tainted 3.13.0-rc1+ #2 [ 253.004864] task: df9b0900 ti: df03e000 task.ti: df03e000 [ 253.004937] PC is at dmaengine_unmap_put+0x14/0x34 [ 253.005010] LR is at pl330_tasklet+0x3c8/0x550 [ 253.005087] pc : [<c0204b44>] lr : [<c0207478>] psr: a00e0193 [ 253.005087] sp : df03fe48 ip : 00000000 fp : df03bf18 [ 253.005178] r10: bf00e108 r9 : 00000001 r8 : 00000000 [ 253.005245] r7 : df837040 r6 : dfb41800 r5 : df837048 r4 : df837000 [ 253.005316] r3 : dfdfdfcf r2 : dfb41f80 r1 : df837048 r0 : dfdfdfd7 [ 253.005384] Flags: NzCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment kernel [ 253.005459] Control: 30c5387d Table: 9fb9ba80 DAC: fffffffd [ 253.005520] Process kthreadd (pid: 671, stack limit = 0xdf03e248) This is due to desc->txd.unmap containing garbage (uninitialised memory). Rather than add another dummy initialisation to _init_desc, instead ensure that the descriptors are zero-initialised during allocation and remove the dummy, per-field initialisation. Cc: Andriy Shevchenko <andriy.shevchenko@intel.com> Acked-by: Jassi Brar <jassisinghbrar@gmail.com> Signed-off-by: Will Deacon <will.deacon@arm.com> Acked-by: Vinod Koul <vinod.koul@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2013-12-03 01:01:30 +07:00
desc = kcalloc(count, sizeof(*desc), flg);
if (!desc)
return 0;
spin_lock_irqsave(lock, flags);
for (i = 0; i < count; i++) {
_init_desc(&desc[i]);
list_add_tail(&desc[i].node, pool);
}
spin_unlock_irqrestore(lock, flags);
return count;
}
static struct dma_pl330_desc *pluck_desc(struct list_head *pool,
spinlock_t *lock)
{
struct dma_pl330_desc *desc = NULL;
unsigned long flags;
spin_lock_irqsave(lock, flags);
if (!list_empty(pool)) {
desc = list_entry(pool->next,
struct dma_pl330_desc, node);
list_del_init(&desc->node);
desc->status = PREP;
desc->txd.callback = NULL;
}
spin_unlock_irqrestore(lock, flags);
return desc;
}
static struct dma_pl330_desc *pl330_get_desc(struct dma_pl330_chan *pch)
{
struct pl330_dmac *pl330 = pch->dmac;
u8 *peri_id = pch->chan.private;
struct dma_pl330_desc *desc;
/* Pluck one desc from the pool of DMAC */
desc = pluck_desc(&pl330->desc_pool, &pl330->pool_lock);
/* If the DMAC pool is empty, alloc new */
if (!desc) {
DEFINE_SPINLOCK(lock);
LIST_HEAD(pool);
if (!add_desc(&pool, &lock, GFP_ATOMIC, 1))
return NULL;
desc = pluck_desc(&pool, &lock);
WARN_ON(!desc || !list_empty(&pool));
}
/* Initialize the descriptor */
desc->pchan = pch;
desc->txd.cookie = 0;
async_tx_ack(&desc->txd);
desc->peri = peri_id ? pch->chan.chan_id : 0;
desc->rqcfg.pcfg = &pch->dmac->pcfg;
dma_async_tx_descriptor_init(&desc->txd, &pch->chan);
return desc;
}
static inline void fill_px(struct pl330_xfer *px,
dma_addr_t dst, dma_addr_t src, size_t len)
{
px->bytes = len;
px->dst_addr = dst;
px->src_addr = src;
}
static struct dma_pl330_desc *
__pl330_prep_dma_memcpy(struct dma_pl330_chan *pch, dma_addr_t dst,
dma_addr_t src, size_t len)
{
struct dma_pl330_desc *desc = pl330_get_desc(pch);
if (!desc) {
dev_err(pch->dmac->ddma.dev, "%s:%d Unable to fetch desc\n",
__func__, __LINE__);
return NULL;
}
/*
* Ideally we should lookout for reqs bigger than
* those that can be programmed with 256 bytes of
* MC buffer, but considering a req size is seldom
* going to be word-unaligned and more than 200MB,
* we take it easy.
* Also, should the limit is reached we'd rather
* have the platform increase MC buffer size than
* complicating this API driver.
*/
fill_px(&desc->px, dst, src, len);
return desc;
}
/* Call after fixing burst size */
static inline int get_burst_len(struct dma_pl330_desc *desc, size_t len)
{
struct dma_pl330_chan *pch = desc->pchan;
struct pl330_dmac *pl330 = pch->dmac;
int burst_len;
burst_len = pl330->pcfg.data_bus_width / 8;
burst_len *= pl330->pcfg.data_buf_dep / pl330->pcfg.num_chan;
burst_len >>= desc->rqcfg.brst_size;
/* src/dst_burst_len can't be more than 16 */
if (burst_len > 16)
burst_len = 16;
while (burst_len > 1) {
if (!(len % (burst_len << desc->rqcfg.brst_size)))
break;
burst_len--;
}
return burst_len;
}
static struct dma_async_tx_descriptor *pl330_prep_dma_cyclic(
struct dma_chan *chan, dma_addr_t dma_addr, size_t len,
size_t period_len, enum dma_transfer_direction direction,
unsigned long flags)
{
struct dma_pl330_desc *desc = NULL, *first = NULL;
struct dma_pl330_chan *pch = to_pchan(chan);
struct pl330_dmac *pl330 = pch->dmac;
unsigned int i;
dma_addr_t dst;
dma_addr_t src;
if (len % period_len != 0)
return NULL;
if (!is_slave_direction(direction)) {
dev_err(pch->dmac->ddma.dev, "%s:%d Invalid dma direction\n",
__func__, __LINE__);
return NULL;
}
if (!pl330_prep_slave_fifo(pch, direction))
return NULL;
for (i = 0; i < len / period_len; i++) {
desc = pl330_get_desc(pch);
if (!desc) {
dev_err(pch->dmac->ddma.dev, "%s:%d Unable to fetch desc\n",
__func__, __LINE__);
if (!first)
return NULL;
spin_lock_irqsave(&pl330->pool_lock, flags);
while (!list_empty(&first->node)) {
desc = list_entry(first->node.next,
struct dma_pl330_desc, node);
list_move_tail(&desc->node, &pl330->desc_pool);
}
list_move_tail(&first->node, &pl330->desc_pool);
spin_unlock_irqrestore(&pl330->pool_lock, flags);
return NULL;
}
switch (direction) {
case DMA_MEM_TO_DEV:
desc->rqcfg.src_inc = 1;
desc->rqcfg.dst_inc = 0;
src = dma_addr;
dst = pch->fifo_dma;
break;
case DMA_DEV_TO_MEM:
desc->rqcfg.src_inc = 0;
desc->rqcfg.dst_inc = 1;
src = pch->fifo_dma;
dst = dma_addr;
break;
default:
break;
}
desc->rqtype = direction;
desc->rqcfg.brst_size = pch->burst_sz;
desc->rqcfg.brst_len = 1;
desc->bytes_requested = period_len;
fill_px(&desc->px, dst, src, period_len);
if (!first)
first = desc;
else
list_add_tail(&desc->node, &first->node);
dma_addr += period_len;
}
if (!desc)
return NULL;
pch->cyclic = true;
desc->txd.flags = flags;
return &desc->txd;
}
static struct dma_async_tx_descriptor *
pl330_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dst,
dma_addr_t src, size_t len, unsigned long flags)
{
struct dma_pl330_desc *desc;
struct dma_pl330_chan *pch = to_pchan(chan);
struct pl330_dmac *pl330;
int burst;
if (unlikely(!pch || !len))
return NULL;
pl330 = pch->dmac;
desc = __pl330_prep_dma_memcpy(pch, dst, src, len);
if (!desc)
return NULL;
desc->rqcfg.src_inc = 1;
desc->rqcfg.dst_inc = 1;
desc->rqtype = DMA_MEM_TO_MEM;
/* Select max possible burst size */
burst = pl330->pcfg.data_bus_width / 8;
/*
* Make sure we use a burst size that aligns with all the memcpy
* parameters because our DMA programming algorithm doesn't cope with
* transfers which straddle an entry in the DMA device's MFIFO.
*/
while ((src | dst | len) & (burst - 1))
burst /= 2;
desc->rqcfg.brst_size = 0;
while (burst != (1 << desc->rqcfg.brst_size))
desc->rqcfg.brst_size++;
/*
* If burst size is smaller than bus width then make sure we only
* transfer one at a time to avoid a burst stradling an MFIFO entry.
*/
if (desc->rqcfg.brst_size * 8 < pl330->pcfg.data_bus_width)
desc->rqcfg.brst_len = 1;
desc->rqcfg.brst_len = get_burst_len(desc, len);
desc->bytes_requested = len;
desc->txd.flags = flags;
return &desc->txd;
}
static void __pl330_giveback_desc(struct pl330_dmac *pl330,
struct dma_pl330_desc *first)
{
unsigned long flags;
struct dma_pl330_desc *desc;
if (!first)
return;
spin_lock_irqsave(&pl330->pool_lock, flags);
while (!list_empty(&first->node)) {
desc = list_entry(first->node.next,
struct dma_pl330_desc, node);
list_move_tail(&desc->node, &pl330->desc_pool);
}
list_move_tail(&first->node, &pl330->desc_pool);
spin_unlock_irqrestore(&pl330->pool_lock, flags);
}
static struct dma_async_tx_descriptor *
pl330_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
unsigned int sg_len, enum dma_transfer_direction direction,
unsigned long flg, void *context)
{
struct dma_pl330_desc *first, *desc = NULL;
struct dma_pl330_chan *pch = to_pchan(chan);
struct scatterlist *sg;
int i;
if (unlikely(!pch || !sgl || !sg_len))
return NULL;
if (!pl330_prep_slave_fifo(pch, direction))
return NULL;
first = NULL;
for_each_sg(sgl, sg, sg_len, i) {
desc = pl330_get_desc(pch);
if (!desc) {
struct pl330_dmac *pl330 = pch->dmac;
dev_err(pch->dmac->ddma.dev,
"%s:%d Unable to fetch desc\n",
__func__, __LINE__);
__pl330_giveback_desc(pl330, first);
return NULL;
}
if (!first)
first = desc;
else
list_add_tail(&desc->node, &first->node);
if (direction == DMA_MEM_TO_DEV) {
desc->rqcfg.src_inc = 1;
desc->rqcfg.dst_inc = 0;
fill_px(&desc->px, pch->fifo_dma, sg_dma_address(sg),
sg_dma_len(sg));
} else {
desc->rqcfg.src_inc = 0;
desc->rqcfg.dst_inc = 1;
fill_px(&desc->px, sg_dma_address(sg), pch->fifo_dma,
sg_dma_len(sg));
}
desc->rqcfg.brst_size = pch->burst_sz;
desc->rqcfg.brst_len = 1;
desc->rqtype = direction;
desc->bytes_requested = sg_dma_len(sg);
}
/* Return the last desc in the chain */
desc->txd.flags = flg;
return &desc->txd;
}
static irqreturn_t pl330_irq_handler(int irq, void *data)
{
if (pl330_update(data))
return IRQ_HANDLED;
else
return IRQ_NONE;
}
#define PL330_DMA_BUSWIDTHS \
BIT(DMA_SLAVE_BUSWIDTH_UNDEFINED) | \
BIT(DMA_SLAVE_BUSWIDTH_1_BYTE) | \
BIT(DMA_SLAVE_BUSWIDTH_2_BYTES) | \
BIT(DMA_SLAVE_BUSWIDTH_4_BYTES) | \
BIT(DMA_SLAVE_BUSWIDTH_8_BYTES)
/*
* Runtime PM callbacks are provided by amba/bus.c driver.
*
* It is assumed here that IRQ safe runtime PM is chosen in probe and amba
* bus driver will only disable/enable the clock in runtime PM callbacks.
*/
static int __maybe_unused pl330_suspend(struct device *dev)
{
struct amba_device *pcdev = to_amba_device(dev);
pm_runtime_disable(dev);
if (!pm_runtime_status_suspended(dev)) {
/* amba did not disable the clock */
amba_pclk_disable(pcdev);
}
amba_pclk_unprepare(pcdev);
return 0;
}
static int __maybe_unused pl330_resume(struct device *dev)
{
struct amba_device *pcdev = to_amba_device(dev);
int ret;
ret = amba_pclk_prepare(pcdev);
if (ret)
return ret;
if (!pm_runtime_status_suspended(dev))
ret = amba_pclk_enable(pcdev);
pm_runtime_enable(dev);
return ret;
}
static SIMPLE_DEV_PM_OPS(pl330_pm, pl330_suspend, pl330_resume);
static int
pl330_probe(struct amba_device *adev, const struct amba_id *id)
{
struct pl330_config *pcfg;
struct pl330_dmac *pl330;
struct dma_pl330_chan *pch, *_p;
struct dma_device *pd;
struct resource *res;
int i, ret, irq;
int num_chan;
struct device_node *np = adev->dev.of_node;
ret = dma_set_mask_and_coherent(&adev->dev, DMA_BIT_MASK(32));
if (ret)
return ret;
/* Allocate a new DMAC and its Channels */
pl330 = devm_kzalloc(&adev->dev, sizeof(*pl330), GFP_KERNEL);
if (!pl330)
return -ENOMEM;
dmaengine: pl330: Correct device assignment. Commit f6f2421c0a1c removed pl330_info structure by embedding it into pl330_dmac structure, but did not ensure that the dmac->ddma.dev pointer gets initialised before use. When dma_alloc_coherent() gets called on arm64 a WARN() gets triggered due to dev being NULL. ------------[ cut here ]------------ WARNING: CPU: 1 PID: 1 at arch/arm64/mm/dma-mapping.c:49 __dma_alloc_coherent+0xd0/0xe0() Use an actual device structure for DMA allocation Modules linked in: CPU: 1 PID: 1 Comm: swapper/0 Not tainted 3.17.0+ #5 Call trace: [<ffffffc000087f24>] dump_backtrace+0x0/0x130 [<ffffffc000088064>] show_stack+0x10/0x1c [<ffffffc0004e8af8>] dump_stack+0x74/0xb8 [<ffffffc0000aa444>] warn_slowpath_common+0x8c/0xb4 [<ffffffc0000aa4b8>] warn_slowpath_fmt+0x4c/0x58 [<ffffffc000092580>] __dma_alloc_coherent+0xcc/0xe0 [<ffffffc000092734>] __dma_alloc_noncoherent+0x64/0x158 [<ffffffc000312cd8>] pl330_probe+0x650/0x8f0 [<ffffffc00030e1d4>] amba_probe+0xa0/0xc8 [<ffffffc000350240>] really_probe+0xc4/0x22c [<ffffffc0003504b4>] __driver_attach+0xa0/0xa8 [<ffffffc00034e5fc>] bus_for_each_dev+0x54/0x98 [<ffffffc00034fd8c>] driver_attach+0x1c/0x28 [<ffffffc00034fa08>] bus_add_driver+0x14c/0x204 [<ffffffc000350b84>] driver_register+0x64/0x130 [<ffffffc00030dcf8>] amba_driver_register+0x50/0x5c [<ffffffc0006a60d0>] pl330_driver_init+0x10/0x1c [<ffffffc0000814ac>] do_one_initcall+0x88/0x19c [<ffffffc00068dab8>] kernel_init_freeable+0x140/0x1e0 [<ffffffc0004e5e18>] kernel_init+0x10/0xd4 ---[ end trace 76f2d47a444e523e ]--- (NULL device *): dmac_alloc_resources:1821 Can't allocate memory! (NULL device *): Unable to create channels for DMAC This patch will also ensure that any dev_err messages are printed with the appropriate device name. Signed-off-by: Liviu Dudau <Liviu.Dudau@arm.com> Signed-off-by: Andrew Jackson <Andrew.Jackson@arm.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-11-06 18:39:47 +07:00
pd = &pl330->ddma;
pd->dev = &adev->dev;
pl330->mcbufsz = 0;
/* get quirk */
for (i = 0; i < ARRAY_SIZE(of_quirks); i++)
if (of_property_read_bool(np, of_quirks[i].quirk))
pl330->quirks |= of_quirks[i].id;
res = &adev->res;
pl330->base = devm_ioremap_resource(&adev->dev, res);
if (IS_ERR(pl330->base))
return PTR_ERR(pl330->base);
amba_set_drvdata(adev, pl330);
for (i = 0; i < AMBA_NR_IRQS; i++) {
irq = adev->irq[i];
if (irq) {
ret = devm_request_irq(&adev->dev, irq,
pl330_irq_handler, 0,
dev_name(&adev->dev), pl330);
if (ret)
return ret;
} else {
break;
}
}
pcfg = &pl330->pcfg;
pcfg->periph_id = adev->periphid;
ret = pl330_add(pl330);
if (ret)
return ret;
INIT_LIST_HEAD(&pl330->desc_pool);
spin_lock_init(&pl330->pool_lock);
/* Create a descriptor pool of default size */
if (!add_desc(&pl330->desc_pool, &pl330->pool_lock,
GFP_KERNEL, NR_DEFAULT_DESC))
dev_warn(&adev->dev, "unable to allocate desc\n");
INIT_LIST_HEAD(&pd->channels);
/* Initialize channel parameters */
num_chan = max_t(int, pcfg->num_peri, pcfg->num_chan);
pl330->num_peripherals = num_chan;
pl330->peripherals = kzalloc(num_chan * sizeof(*pch), GFP_KERNEL);
if (!pl330->peripherals) {
ret = -ENOMEM;
goto probe_err2;
}
for (i = 0; i < num_chan; i++) {
pch = &pl330->peripherals[i];
pch->chan.private = adev->dev.of_node;
INIT_LIST_HEAD(&pch->submitted_list);
INIT_LIST_HEAD(&pch->work_list);
dma: pl330: Fix handling of TERMINATE_ALL while processing completed descriptors The pl330 DMA driver is broken in regard to handling a terminate all request while it is processing the list of completed descriptors. This is most visible when calling dmaengine_terminate_all() from within the descriptors callback for cyclic transfers. In this case the TERMINATE_ALL transfer will clear the work_list and stop the transfer. But after all callbacks for all completed descriptors have been handled the descriptors will be re-enqueued into the (now empty) work_list. So the next time dma_async_issue_pending() is called for the channel these descriptors will be transferred again which will cause data corruption. Similar issues can occur if dmaengine_terminate_all() is not called from within the descriptor callback but runs on a different CPU at the same time as the completed descriptor list is processed. This patch introduces a new per channel list which will hold the completed descriptors. While processing the list the channel's lock will be held to avoid racing against dmaengine_terminate_all(). The lock will be released when calling the descriptors callback though. Since the list of completed descriptors might be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can not use the normal list iterator macros. Instead we'll need to check for each loop iteration again if there are still items in the list. The drivers TERMINATE_ALL implementation is updated to move descriptors from both the work_list as well the new completed_list back to the descriptor pool. This makes sure that none of the descripts finds its way back into the work list and also that we do not call any futher complete callbacks after dmaengine_terminate_all() has been called. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2013-08-28 01:34:05 +07:00
INIT_LIST_HEAD(&pch->completed_list);
spin_lock_init(&pch->lock);
pch->thread = NULL;
pch->chan.device = pd;
pch->dmac = pl330;
pch->dir = DMA_NONE;
/* Add the channel to the DMAC list */
list_add_tail(&pch->chan.device_node, &pd->channels);
}
dma_cap_set(DMA_MEMCPY, pd->cap_mask);
if (pcfg->num_peri) {
dma_cap_set(DMA_SLAVE, pd->cap_mask);
dma_cap_set(DMA_CYCLIC, pd->cap_mask);
dma_cap_set(DMA_PRIVATE, pd->cap_mask);
}
pd->device_alloc_chan_resources = pl330_alloc_chan_resources;
pd->device_free_chan_resources = pl330_free_chan_resources;
pd->device_prep_dma_memcpy = pl330_prep_dma_memcpy;
pd->device_prep_dma_cyclic = pl330_prep_dma_cyclic;
pd->device_tx_status = pl330_tx_status;
pd->device_prep_slave_sg = pl330_prep_slave_sg;
pd->device_config = pl330_config;
pd->device_pause = pl330_pause;
pd->device_terminate_all = pl330_terminate_all;
pd->device_issue_pending = pl330_issue_pending;
pd->src_addr_widths = PL330_DMA_BUSWIDTHS;
pd->dst_addr_widths = PL330_DMA_BUSWIDTHS;
pd->directions = BIT(DMA_DEV_TO_MEM) | BIT(DMA_MEM_TO_DEV);
pd->residue_granularity = DMA_RESIDUE_GRANULARITY_SEGMENT;
pd->max_burst = ((pl330->quirks & PL330_QUIRK_BROKEN_NO_FLUSHP) ?
1 : PL330_MAX_BURST);
ret = dma_async_device_register(pd);
if (ret) {
dev_err(&adev->dev, "unable to register DMAC\n");
goto probe_err3;
}
if (adev->dev.of_node) {
ret = of_dma_controller_register(adev->dev.of_node,
of_dma_pl330_xlate, pl330);
if (ret) {
dev_err(&adev->dev,
"unable to register DMA to the generic DT DMA helpers\n");
}
}
adev->dev.dma_parms = &pl330->dma_parms;
/*
* This is the limit for transfers with a buswidth of 1, larger
* buswidths will have larger limits.
*/
ret = dma_set_max_seg_size(&adev->dev, 1900800);
if (ret)
dev_err(&adev->dev, "unable to set the seg size\n");
dev_info(&adev->dev,
"Loaded driver for PL330 DMAC-%x\n", adev->periphid);
dev_info(&adev->dev,
"\tDBUFF-%ux%ubytes Num_Chans-%u Num_Peri-%u Num_Events-%u\n",
pcfg->data_buf_dep, pcfg->data_bus_width / 8, pcfg->num_chan,
pcfg->num_peri, pcfg->num_events);
pm_runtime_irq_safe(&adev->dev);
pm_runtime_use_autosuspend(&adev->dev);
pm_runtime_set_autosuspend_delay(&adev->dev, PL330_AUTOSUSPEND_DELAY);
pm_runtime_mark_last_busy(&adev->dev);
pm_runtime_put_autosuspend(&adev->dev);
return 0;
probe_err3:
/* Idle the DMAC */
list_for_each_entry_safe(pch, _p, &pl330->ddma.channels,
chan.device_node) {
/* Remove the channel */
list_del(&pch->chan.device_node);
/* Flush the channel */
dmaengine: pl330: Fix NULL pointer dereference on probe failure If dma_async_device_register() returns error and probe should clean up and return error, a NULL pointer exception happens because of dereference of not allocated channel thread: Dmesg log (from early printk): dma-pl330 12680000.pdma: unable to register DMAC DMA pl330_control: removing pch: eeac4000, chan: eeac4014, thread: (null) Unable to handle kernel NULL pointer dereference at virtual address 0000000c pgd = c0004000 [0000000c] *pgd=00000000 Internal error: Oops: 5 [#1] PREEMPT SMP ARM Modules linked in: CPU: 2 PID: 1 Comm: swapper/0 Not tainted 3.17.0-rc3-next-20140904-00005-g6cc4c1937d90-dirty #427 task: ee80a800 ti: ee888000 task.ti: ee888000 PC is at _stop+0x8/0x2c8 LR is at pl330_control+0x70/0x2e8 pc : [<c0205dc8>] lr : [<c020623c>] psr: 60000193 sp : ee889df8 ip : 00000002 fp : 00000000 r10: eeac4014 r9 : ee0e62bc r8 : 00000000 r7 : eeac405c r6 : 60000113 r5 : ee0e6210 r4 : eeac4000 r3 : 00000002 r2 : 00000002 r1 : 00010000 r0 : 00000000 Flags: nZCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment kernel Control: 10c5387d Table: 4000404a DAC: 00000015 Process swapper/0 (pid: 1, stack limit = 0xee888240) Stack: (0xee889df8 to 0xee88a000) 9de0: 00000002 eeac4000 9e00: ee0e6210 eeac4000 ee0e6210 60000113 eeac405c c020623c 00000000 c020725c 9e20: ee889e20 ee889e20 ee0e6210 eeac4080 00200200 00100100 eeac4014 00000020 9e40: ee0e6218 c0208374 00000000 ee9bb340 ee0e6210 00000000 00000000 c0605cd8 9e60: ee970000 c0605c84 ee9700f8 00000000 c05c4270 00000000 00000000 c0203b3c 9e80: ee970000 c06624a8 00000000 c0605c84 00000000 c023f890 ee970000 c0605c84 9ea0: ee970034 00000000 c05b23d0 c023fa3c 00000000 c0605c84 c023f9b0 c023e0d4 9ec0: ee947e78 ee9b9440 c0605c84 eea1e780 c0605acc c023f094 c0513b50 c0605c84 9ee0: c05ecbd8 c0605c84 c05ecbd8 ee11ba40 c0626500 c0240064 00000000 c05ecbd8 9f00: c05ecbd8 c0008964 c040f13c 0000009f c0626500 c057465c ee80a800 60000113 9f20: 00000000 c05efdb0 60000113 00000000 ef7fc89d c0421168 0000008f c003787c 9f40: c0573d6c 00000006 ef7fc8bb 00000006 c05efd50 ef7fc800 c05dfbc4 00000006 9f60: c05c4264 c0626500 0000008f c05c4270 c059b518 c059bcb4 00000006 00000006 9f80: c059b518 c003c08c 00000000 c040091c 00000000 00000000 00000000 00000000 9fa0: 00000000 c0400924 00000000 c000e7b8 00000000 00000000 00000000 00000000 9fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 9fe0: 00000000 00000000 00000000 00000000 00000013 00000000 c0c0c0c0 c0c0c0c0 [<c0205dc8>] (_stop) from [<c020623c>] (pl330_control+0x70/0x2e8) [<c020623c>] (pl330_control) from [<c0208374>] (pl330_probe+0x594/0x75c) [<c0208374>] (pl330_probe) from [<c0203b3c>] (amba_probe+0xb8/0x120) [<c0203b3c>] (amba_probe) from [<c023f890>] (driver_probe_device+0x10c/0x22c) [<c023f890>] (driver_probe_device) from [<c023fa3c>] (__driver_attach+0x8c/0x90) [<c023fa3c>] (__driver_attach) from [<c023e0d4>] (bus_for_each_dev+0x54/0x88) [<c023e0d4>] (bus_for_each_dev) from [<c023f094>] (bus_add_driver+0xd4/0x1d0) [<c023f094>] (bus_add_driver) from [<c0240064>] (driver_register+0x78/0xf4) [<c0240064>] (driver_register) from [<c0008964>] (do_one_initcall+0x80/0x1d0) [<c0008964>] (do_one_initcall) from [<c059bcb4>] (kernel_init_freeable+0x108/0x1d4) [<c059bcb4>] (kernel_init_freeable) from [<c0400924>] (kernel_init+0x8/0xec) [<c0400924>] (kernel_init) from [<c000e7b8>] (ret_from_fork+0x14/0x3c) Code: e5813010 e12fff1e e92d40f0 e24dd00c (e590200c) ---[ end trace c94b2f4f38dff3bf ]--- This happens because the necessary resources were not yet allocated - no call to pl330_alloc_chan_resources(). Terminate the thread and free channel resource only if channel thread is not NULL. Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> Cc: <stable@vger.kernel.org> Fixes: 0b94c5771705 ("DMA: PL330: Add check if device tree compatible") Reviewed-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-09-29 19:42:20 +07:00
if (pch->thread) {
pl330_terminate_all(&pch->chan);
dmaengine: pl330: Fix NULL pointer dereference on probe failure If dma_async_device_register() returns error and probe should clean up and return error, a NULL pointer exception happens because of dereference of not allocated channel thread: Dmesg log (from early printk): dma-pl330 12680000.pdma: unable to register DMAC DMA pl330_control: removing pch: eeac4000, chan: eeac4014, thread: (null) Unable to handle kernel NULL pointer dereference at virtual address 0000000c pgd = c0004000 [0000000c] *pgd=00000000 Internal error: Oops: 5 [#1] PREEMPT SMP ARM Modules linked in: CPU: 2 PID: 1 Comm: swapper/0 Not tainted 3.17.0-rc3-next-20140904-00005-g6cc4c1937d90-dirty #427 task: ee80a800 ti: ee888000 task.ti: ee888000 PC is at _stop+0x8/0x2c8 LR is at pl330_control+0x70/0x2e8 pc : [<c0205dc8>] lr : [<c020623c>] psr: 60000193 sp : ee889df8 ip : 00000002 fp : 00000000 r10: eeac4014 r9 : ee0e62bc r8 : 00000000 r7 : eeac405c r6 : 60000113 r5 : ee0e6210 r4 : eeac4000 r3 : 00000002 r2 : 00000002 r1 : 00010000 r0 : 00000000 Flags: nZCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment kernel Control: 10c5387d Table: 4000404a DAC: 00000015 Process swapper/0 (pid: 1, stack limit = 0xee888240) Stack: (0xee889df8 to 0xee88a000) 9de0: 00000002 eeac4000 9e00: ee0e6210 eeac4000 ee0e6210 60000113 eeac405c c020623c 00000000 c020725c 9e20: ee889e20 ee889e20 ee0e6210 eeac4080 00200200 00100100 eeac4014 00000020 9e40: ee0e6218 c0208374 00000000 ee9bb340 ee0e6210 00000000 00000000 c0605cd8 9e60: ee970000 c0605c84 ee9700f8 00000000 c05c4270 00000000 00000000 c0203b3c 9e80: ee970000 c06624a8 00000000 c0605c84 00000000 c023f890 ee970000 c0605c84 9ea0: ee970034 00000000 c05b23d0 c023fa3c 00000000 c0605c84 c023f9b0 c023e0d4 9ec0: ee947e78 ee9b9440 c0605c84 eea1e780 c0605acc c023f094 c0513b50 c0605c84 9ee0: c05ecbd8 c0605c84 c05ecbd8 ee11ba40 c0626500 c0240064 00000000 c05ecbd8 9f00: c05ecbd8 c0008964 c040f13c 0000009f c0626500 c057465c ee80a800 60000113 9f20: 00000000 c05efdb0 60000113 00000000 ef7fc89d c0421168 0000008f c003787c 9f40: c0573d6c 00000006 ef7fc8bb 00000006 c05efd50 ef7fc800 c05dfbc4 00000006 9f60: c05c4264 c0626500 0000008f c05c4270 c059b518 c059bcb4 00000006 00000006 9f80: c059b518 c003c08c 00000000 c040091c 00000000 00000000 00000000 00000000 9fa0: 00000000 c0400924 00000000 c000e7b8 00000000 00000000 00000000 00000000 9fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 9fe0: 00000000 00000000 00000000 00000000 00000013 00000000 c0c0c0c0 c0c0c0c0 [<c0205dc8>] (_stop) from [<c020623c>] (pl330_control+0x70/0x2e8) [<c020623c>] (pl330_control) from [<c0208374>] (pl330_probe+0x594/0x75c) [<c0208374>] (pl330_probe) from [<c0203b3c>] (amba_probe+0xb8/0x120) [<c0203b3c>] (amba_probe) from [<c023f890>] (driver_probe_device+0x10c/0x22c) [<c023f890>] (driver_probe_device) from [<c023fa3c>] (__driver_attach+0x8c/0x90) [<c023fa3c>] (__driver_attach) from [<c023e0d4>] (bus_for_each_dev+0x54/0x88) [<c023e0d4>] (bus_for_each_dev) from [<c023f094>] (bus_add_driver+0xd4/0x1d0) [<c023f094>] (bus_add_driver) from [<c0240064>] (driver_register+0x78/0xf4) [<c0240064>] (driver_register) from [<c0008964>] (do_one_initcall+0x80/0x1d0) [<c0008964>] (do_one_initcall) from [<c059bcb4>] (kernel_init_freeable+0x108/0x1d4) [<c059bcb4>] (kernel_init_freeable) from [<c0400924>] (kernel_init+0x8/0xec) [<c0400924>] (kernel_init) from [<c000e7b8>] (ret_from_fork+0x14/0x3c) Code: e5813010 e12fff1e e92d40f0 e24dd00c (e590200c) ---[ end trace c94b2f4f38dff3bf ]--- This happens because the necessary resources were not yet allocated - no call to pl330_alloc_chan_resources(). Terminate the thread and free channel resource only if channel thread is not NULL. Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> Cc: <stable@vger.kernel.org> Fixes: 0b94c5771705 ("DMA: PL330: Add check if device tree compatible") Reviewed-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-09-29 19:42:20 +07:00
pl330_free_chan_resources(&pch->chan);
}
}
probe_err2:
pl330_del(pl330);
return ret;
}
static int pl330_remove(struct amba_device *adev)
{
struct pl330_dmac *pl330 = amba_get_drvdata(adev);
struct dma_pl330_chan *pch, *_p;
int i, irq;
pm_runtime_get_noresume(pl330->ddma.dev);
if (adev->dev.of_node)
of_dma_controller_free(adev->dev.of_node);
for (i = 0; i < AMBA_NR_IRQS; i++) {
irq = adev->irq[i];
if (irq)
devm_free_irq(&adev->dev, irq, pl330);
}
dma_async_device_unregister(&pl330->ddma);
/* Idle the DMAC */
list_for_each_entry_safe(pch, _p, &pl330->ddma.channels,
chan.device_node) {
/* Remove the channel */
list_del(&pch->chan.device_node);
/* Flush the channel */
dmaengine: pl330: Fix NULL pointer dereference on driver unbind Fix a NULL pointer dereference after unbinding the driver, if channel resources were not yet allocated (no call to pl330_alloc_chan_resources()): $ echo 12850000.mdma > /sys/bus/amba/drivers/dma-pl330/unbind [ 13.606533] DMA pl330_control: removing pch: eeab6800, chan: eeab6814, thread: (null) [ 13.614472] Unable to handle kernel NULL pointer dereference at virtual address 0000000c [ 13.622537] pgd = ee284000 [ 13.625228] [0000000c] *pgd=6e1e4831, *pte=00000000, *ppte=00000000 [ 13.631482] Internal error: Oops: 17 [#1] PREEMPT SMP ARM [ 13.636859] Modules linked in: [ 13.639903] CPU: 0 PID: 1 Comm: sh Not tainted 3.17.0-rc3-next-20140904-00004-g7020ffc33ca3-dirty #420 [ 13.649187] task: ee80a800 ti: ee888000 task.ti: ee888000 [ 13.654589] PC is at _stop+0x8/0x2c8 [ 13.658131] LR is at pl330_control+0x70/0x2e8 [ 13.662468] pc : [<c0206028>] lr : [<c020649c>] psr: 60000093 [ 13.662468] sp : ee889e58 ip : 00000001 fp : 000bab70 [ 13.673922] r10: eeab6814 r9 : ee16debc r8 : 00000000 [ 13.679131] r7 : eeab685c r6 : 60000013 r5 : ee16de10 r4 : eeab6800 [ 13.685641] r3 : 00000002 r2 : 00000000 r1 : 00010000 r0 : 00000000 [ 13.692153] Flags: nZCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment user [ 13.699357] Control: 10c5387d Table: 6e28404a DAC: 00000015 [ 13.705085] Process sh (pid: 1, stack limit = 0xee888240) [ 13.710466] Stack: (0xee889e58 to 0xee88a000) [ 13.714808] 9e40: 00000002 eeab6800 [ 13.722969] 9e60: ee16de10 eeab6800 ee16de10 60000013 eeab685c c020649c 00000000 c040280c [ 13.731128] 9e80: ee889e80 ee889e80 ee16de18 ee16de10 eeab6880 eeab6814 00200200 eeab68a8 [ 13.739287] 9ea0: 00100100 c0208048 00000000 c0409fc4 eea80800 eea808f8 c0605c44 0000000e [ 13.747446] 9ec0: 0000000e eeb3960c eeb39600 c0203c48 eea80800 c0605c44 c0605a8c c023f694 [ 13.755605] 9ee0: ee80a800 eea80834 eea80800 c023f704 ee80a800 eea80800 c0605c44 c023e8ec [ 13.763764] 9f00: 0000000e ee149780 ee29e580 ee889f80 ee29e580 c023e19c 0000000e c01167e4 [ 13.771923] 9f20: c01167a0 00000000 00000000 c0115e88 00000000 00000000 ee0b1a00 0000000e [ 13.780082] 9f40: b6f48000 ee889f80 0000000e ee888000 b6f48000 c00bfadc 00000000 00000003 [ 13.788241] 9f60: 00000000 00000000 00000000 ee0b1a00 ee0b1a00 0000000e b6f48000 c00bfdf4 [ 13.796401] 9f80: 00000000 00000000 ffffffff 0000000e b6f48000 b6edc5d0 00000004 c000e7a4 [ 13.804560] 9fa0: 00000000 c000e620 0000000e b6f48000 00000001 b6f48000 0000000e 00000000 [ 13.812719] 9fc0: 0000000e b6f48000 b6edc5d0 00000004 0000000e b6f4c8c0 000c3470 000bab70 [ 13.820879] 9fe0: 00000000 bed2aa50 b6e18bdc b6e6b52c 60000010 00000001 c0c0c0c0 c0c0c0c0 [ 13.829058] [<c0206028>] (_stop) from [<c020649c>] (pl330_control+0x70/0x2e8) [ 13.836165] [<c020649c>] (pl330_control) from [<c0208048>] (pl330_remove+0xb0/0xdc) [ 13.843800] [<c0208048>] (pl330_remove) from [<c0203c48>] (amba_remove+0x24/0xc0) [ 13.851272] [<c0203c48>] (amba_remove) from [<c023f694>] (__device_release_driver+0x70/0xc4) [ 13.859685] [<c023f694>] (__device_release_driver) from [<c023f704>] (device_release_driver+0x1c/0x28) [ 13.868971] [<c023f704>] (device_release_driver) from [<c023e8ec>] (unbind_store+0x58/0x90) [ 13.877303] [<c023e8ec>] (unbind_store) from [<c023e19c>] (drv_attr_store+0x20/0x2c) [ 13.885036] [<c023e19c>] (drv_attr_store) from [<c01167e4>] (sysfs_kf_write+0x44/0x48) [ 13.892928] [<c01167e4>] (sysfs_kf_write) from [<c0115e88>] (kernfs_fop_write+0xc0/0x17c) [ 13.901090] [<c0115e88>] (kernfs_fop_write) from [<c00bfadc>] (vfs_write+0xa0/0x1a8) [ 13.908812] [<c00bfadc>] (vfs_write) from [<c00bfdf4>] (SyS_write+0x40/0x8c) [ 13.915850] [<c00bfdf4>] (SyS_write) from [<c000e620>] (ret_fast_syscall+0x0/0x30) [ 13.923392] Code: e5813010 e12fff1e e92d40f0 e24dd00c (e590200c) [ 13.929467] ---[ end trace 10064e15a5929cf8 ]--- Terminate the thread and free channel resource only if channel resources were allocated (thread is not NULL). Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> Cc: <stable@vger.kernel.org> Fixes: b3040e40675e ("DMA: PL330: Add dma api driver") Reviewed-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-09-29 19:42:21 +07:00
if (pch->thread) {
pl330_terminate_all(&pch->chan);
dmaengine: pl330: Fix NULL pointer dereference on driver unbind Fix a NULL pointer dereference after unbinding the driver, if channel resources were not yet allocated (no call to pl330_alloc_chan_resources()): $ echo 12850000.mdma > /sys/bus/amba/drivers/dma-pl330/unbind [ 13.606533] DMA pl330_control: removing pch: eeab6800, chan: eeab6814, thread: (null) [ 13.614472] Unable to handle kernel NULL pointer dereference at virtual address 0000000c [ 13.622537] pgd = ee284000 [ 13.625228] [0000000c] *pgd=6e1e4831, *pte=00000000, *ppte=00000000 [ 13.631482] Internal error: Oops: 17 [#1] PREEMPT SMP ARM [ 13.636859] Modules linked in: [ 13.639903] CPU: 0 PID: 1 Comm: sh Not tainted 3.17.0-rc3-next-20140904-00004-g7020ffc33ca3-dirty #420 [ 13.649187] task: ee80a800 ti: ee888000 task.ti: ee888000 [ 13.654589] PC is at _stop+0x8/0x2c8 [ 13.658131] LR is at pl330_control+0x70/0x2e8 [ 13.662468] pc : [<c0206028>] lr : [<c020649c>] psr: 60000093 [ 13.662468] sp : ee889e58 ip : 00000001 fp : 000bab70 [ 13.673922] r10: eeab6814 r9 : ee16debc r8 : 00000000 [ 13.679131] r7 : eeab685c r6 : 60000013 r5 : ee16de10 r4 : eeab6800 [ 13.685641] r3 : 00000002 r2 : 00000000 r1 : 00010000 r0 : 00000000 [ 13.692153] Flags: nZCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment user [ 13.699357] Control: 10c5387d Table: 6e28404a DAC: 00000015 [ 13.705085] Process sh (pid: 1, stack limit = 0xee888240) [ 13.710466] Stack: (0xee889e58 to 0xee88a000) [ 13.714808] 9e40: 00000002 eeab6800 [ 13.722969] 9e60: ee16de10 eeab6800 ee16de10 60000013 eeab685c c020649c 00000000 c040280c [ 13.731128] 9e80: ee889e80 ee889e80 ee16de18 ee16de10 eeab6880 eeab6814 00200200 eeab68a8 [ 13.739287] 9ea0: 00100100 c0208048 00000000 c0409fc4 eea80800 eea808f8 c0605c44 0000000e [ 13.747446] 9ec0: 0000000e eeb3960c eeb39600 c0203c48 eea80800 c0605c44 c0605a8c c023f694 [ 13.755605] 9ee0: ee80a800 eea80834 eea80800 c023f704 ee80a800 eea80800 c0605c44 c023e8ec [ 13.763764] 9f00: 0000000e ee149780 ee29e580 ee889f80 ee29e580 c023e19c 0000000e c01167e4 [ 13.771923] 9f20: c01167a0 00000000 00000000 c0115e88 00000000 00000000 ee0b1a00 0000000e [ 13.780082] 9f40: b6f48000 ee889f80 0000000e ee888000 b6f48000 c00bfadc 00000000 00000003 [ 13.788241] 9f60: 00000000 00000000 00000000 ee0b1a00 ee0b1a00 0000000e b6f48000 c00bfdf4 [ 13.796401] 9f80: 00000000 00000000 ffffffff 0000000e b6f48000 b6edc5d0 00000004 c000e7a4 [ 13.804560] 9fa0: 00000000 c000e620 0000000e b6f48000 00000001 b6f48000 0000000e 00000000 [ 13.812719] 9fc0: 0000000e b6f48000 b6edc5d0 00000004 0000000e b6f4c8c0 000c3470 000bab70 [ 13.820879] 9fe0: 00000000 bed2aa50 b6e18bdc b6e6b52c 60000010 00000001 c0c0c0c0 c0c0c0c0 [ 13.829058] [<c0206028>] (_stop) from [<c020649c>] (pl330_control+0x70/0x2e8) [ 13.836165] [<c020649c>] (pl330_control) from [<c0208048>] (pl330_remove+0xb0/0xdc) [ 13.843800] [<c0208048>] (pl330_remove) from [<c0203c48>] (amba_remove+0x24/0xc0) [ 13.851272] [<c0203c48>] (amba_remove) from [<c023f694>] (__device_release_driver+0x70/0xc4) [ 13.859685] [<c023f694>] (__device_release_driver) from [<c023f704>] (device_release_driver+0x1c/0x28) [ 13.868971] [<c023f704>] (device_release_driver) from [<c023e8ec>] (unbind_store+0x58/0x90) [ 13.877303] [<c023e8ec>] (unbind_store) from [<c023e19c>] (drv_attr_store+0x20/0x2c) [ 13.885036] [<c023e19c>] (drv_attr_store) from [<c01167e4>] (sysfs_kf_write+0x44/0x48) [ 13.892928] [<c01167e4>] (sysfs_kf_write) from [<c0115e88>] (kernfs_fop_write+0xc0/0x17c) [ 13.901090] [<c0115e88>] (kernfs_fop_write) from [<c00bfadc>] (vfs_write+0xa0/0x1a8) [ 13.908812] [<c00bfadc>] (vfs_write) from [<c00bfdf4>] (SyS_write+0x40/0x8c) [ 13.915850] [<c00bfdf4>] (SyS_write) from [<c000e620>] (ret_fast_syscall+0x0/0x30) [ 13.923392] Code: e5813010 e12fff1e e92d40f0 e24dd00c (e590200c) [ 13.929467] ---[ end trace 10064e15a5929cf8 ]--- Terminate the thread and free channel resource only if channel resources were allocated (thread is not NULL). Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> Cc: <stable@vger.kernel.org> Fixes: b3040e40675e ("DMA: PL330: Add dma api driver") Reviewed-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-09-29 19:42:21 +07:00
pl330_free_chan_resources(&pch->chan);
}
}
pl330_del(pl330);
return 0;
}
static const struct amba_id pl330_ids[] = {
{
.id = 0x00041330,
.mask = 0x000fffff,
},
{ 0, 0 },
};
MODULE_DEVICE_TABLE(amba, pl330_ids);
static struct amba_driver pl330_driver = {
.drv = {
.owner = THIS_MODULE,
.name = "dma-pl330",
.pm = &pl330_pm,
},
.id_table = pl330_ids,
.probe = pl330_probe,
.remove = pl330_remove,
};
module_amba_driver(pl330_driver);
MODULE_AUTHOR("Jaswinder Singh <jassisinghbrar@gmail.com>");
MODULE_DESCRIPTION("API Driver for PL330 DMAC");
MODULE_LICENSE("GPL");