mirror of
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git
synced 2024-11-01 17:08:10 +00:00
c4e773764c
The use of a memcpy() during a spinlock operation will cause very long thread context switch delays if the flash chip bandwidth is low and the data to be copied large, because a spinlock will disable preemption. For example: A flash with 6,5 MB/s bandwidth will cause under ubifs, which request sometimes 128 KiB (the flash erase size), a preemption delay of 20 milliseconds. High priority threads will not be served during this time, regardless whether this threads access the flash or not. This behavior breaks real time. The patch changes all the use of spin_lock operations for xxxx->mutex into mutex operations, which is exact what the name says and means. I have checked the code of the drivers and there is no use of atomic pathes like interrupt or timers. The mtdoops facility will also not be used by this drivers. So it is dave to replace the spin_lock against mutex. There is no performance regression since the mutex is normally not acquired. Changelog: 06.03.2010 First release 26.03.2010 Fix mutex[1] issue and tested it for compile failure Signed-off-by: Stefani Seibold <stefani@seibold.net> Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com> Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
101 lines
2.4 KiB
C
101 lines
2.4 KiB
C
|
|
/*
|
|
* struct flchip definition
|
|
*
|
|
* Contains information about the location and state of a given flash device
|
|
*
|
|
* (C) 2000 Red Hat. GPLd.
|
|
*/
|
|
|
|
#ifndef __MTD_FLASHCHIP_H__
|
|
#define __MTD_FLASHCHIP_H__
|
|
|
|
/* For spinlocks. sched.h includes spinlock.h from whichever directory it
|
|
* happens to be in - so we don't have to care whether we're on 2.2, which
|
|
* has asm/spinlock.h, or 2.4, which has linux/spinlock.h
|
|
*/
|
|
#include <linux/sched.h>
|
|
#include <linux/mutex.h>
|
|
|
|
typedef enum {
|
|
FL_READY,
|
|
FL_STATUS,
|
|
FL_CFI_QUERY,
|
|
FL_JEDEC_QUERY,
|
|
FL_ERASING,
|
|
FL_ERASE_SUSPENDING,
|
|
FL_ERASE_SUSPENDED,
|
|
FL_WRITING,
|
|
FL_WRITING_TO_BUFFER,
|
|
FL_OTP_WRITE,
|
|
FL_WRITE_SUSPENDING,
|
|
FL_WRITE_SUSPENDED,
|
|
FL_PM_SUSPENDED,
|
|
FL_SYNCING,
|
|
FL_UNLOADING,
|
|
FL_LOCKING,
|
|
FL_UNLOCKING,
|
|
FL_POINT,
|
|
FL_XIP_WHILE_ERASING,
|
|
FL_XIP_WHILE_WRITING,
|
|
FL_SHUTDOWN,
|
|
/* These 2 come from nand_state_t, which has been unified here */
|
|
FL_READING,
|
|
FL_CACHEDPRG,
|
|
/* These 4 come from onenand_state_t, which has been unified here */
|
|
FL_RESETING,
|
|
FL_OTPING,
|
|
FL_PREPARING_ERASE,
|
|
FL_VERIFYING_ERASE,
|
|
|
|
FL_UNKNOWN
|
|
} flstate_t;
|
|
|
|
|
|
|
|
/* NOTE: confusingly, this can be used to refer to more than one chip at a time,
|
|
if they're interleaved. This can even refer to individual partitions on
|
|
the same physical chip when present. */
|
|
|
|
struct flchip {
|
|
unsigned long start; /* Offset within the map */
|
|
// unsigned long len;
|
|
/* We omit len for now, because when we group them together
|
|
we insist that they're all of the same size, and the chip size
|
|
is held in the next level up. If we get more versatile later,
|
|
it'll make it a damn sight harder to find which chip we want from
|
|
a given offset, and we'll want to add the per-chip length field
|
|
back in.
|
|
*/
|
|
int ref_point_counter;
|
|
flstate_t state;
|
|
flstate_t oldstate;
|
|
|
|
unsigned int write_suspended:1;
|
|
unsigned int erase_suspended:1;
|
|
unsigned long in_progress_block_addr;
|
|
|
|
struct mutex mutex;
|
|
wait_queue_head_t wq; /* Wait on here when we're waiting for the chip
|
|
to be ready */
|
|
int word_write_time;
|
|
int buffer_write_time;
|
|
int erase_time;
|
|
|
|
int word_write_time_max;
|
|
int buffer_write_time_max;
|
|
int erase_time_max;
|
|
|
|
void *priv;
|
|
};
|
|
|
|
/* This is used to handle contention on write/erase operations
|
|
between partitions of the same physical chip. */
|
|
struct flchip_shared {
|
|
spinlock_t lock;
|
|
struct flchip *writing;
|
|
struct flchip *erasing;
|
|
};
|
|
|
|
|
|
#endif /* __MTD_FLASHCHIP_H__ */
|