[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20191108051945.7109-6-joel@jms.id.au>
Date: Fri, 8 Nov 2019 15:49:39 +1030
From: Joel Stanley <joel@....id.au>
To: Rob Herring <robh+dt@...nel.org>,
Greg KH <gregkh@...uxfoundation.org>,
Jeremy Kerr <jk@...abs.org>
Cc: Andrew Jeffery <andrew@...id.au>,
Alistar Popple <alistair@...ple.id.au>,
Eddie James <eajames@...ux.ibm.com>,
Steven Rostedt <rostedt@...dmis.org>,
Ingo Molnar <mingo@...hat.com>, linux-kernel@...r.kernel.org,
devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-fsi@...ts.ozlabs.org
Subject: [PATCH v2 05/11] fsi: core: Fix small accesses and unaligned offsets via sysfs
From: Andrew Jeffery <andrew@...id.au>
Subtracting the offset delta from four-byte alignment lead to wrapping
of the requested length where `count` is less than `off`. Generalise the
length handling to enable and optimise aligned access sizes for all
offset and size combinations. The new formula produces the following
results for given offset and count values:
offset count | length
--------------+-------
0 1 | 1
0 2 | 2
0 3 | 2
0 4 | 4
0 5 | 4
1 1 | 1
1 2 | 1
1 3 | 1
1 4 | 1
1 5 | 1
2 1 | 1
2 2 | 2
2 3 | 2
2 4 | 2
2 5 | 2
3 1 | 1
3 2 | 1
3 3 | 1
3 4 | 1
3 5 | 1
We might need something like this for the cfam chardevs as well, for
example we don't currently implement any alignment restrictions /
handling in the hardware master driver.
Signed-off-by: Andrew Jeffery <andrew@...id.au>
Signed-off-by: Joel Stanley <joel@....id.au>
---
drivers/fsi/fsi-core.c | 31 +++++++++++++++++++++++++++----
1 file changed, 27 insertions(+), 4 deletions(-)
diff --git a/drivers/fsi/fsi-core.c b/drivers/fsi/fsi-core.c
index c773c65a5058..e02ebcb0c9e6 100644
--- a/drivers/fsi/fsi-core.c
+++ b/drivers/fsi/fsi-core.c
@@ -544,6 +544,31 @@ static int fsi_slave_scan(struct fsi_slave *slave)
return 0;
}
+static unsigned long aligned_access_size(size_t offset, size_t count)
+{
+ unsigned long offset_unit, count_unit;
+
+ /* Criteria:
+ *
+ * 1. Access size must be less than or equal to the maximum access
+ * width or the highest power-of-two factor of offset
+ * 2. Access size must be less than or equal to the amount specified by
+ * count
+ *
+ * The access width is optimal if we can calculate 1 to be strictly
+ * equal while still satisfying 2.
+ */
+
+ /* Find 1 by the bottom bit of offset (with a 4 byte access cap) */
+ offset_unit = BIT(__builtin_ctzl(offset | 4));
+
+ /* Find 2 by the top bit of count */
+ count_unit = BIT(8 * sizeof(unsigned long) - 1 - __builtin_clzl(count));
+
+ /* Constrain the maximum access width to the minimum of both criteria */
+ return BIT(__builtin_ctzl(offset_unit | count_unit));
+}
+
static ssize_t fsi_slave_sysfs_raw_read(struct file *file,
struct kobject *kobj, struct bin_attribute *attr, char *buf,
loff_t off, size_t count)
@@ -559,8 +584,7 @@ static ssize_t fsi_slave_sysfs_raw_read(struct file *file,
return -EINVAL;
for (total_len = 0; total_len < count; total_len += read_len) {
- read_len = min_t(size_t, count, 4);
- read_len -= off & 0x3;
+ read_len = aligned_access_size(off, count - total_len);
rc = fsi_slave_read(slave, off, buf + total_len, read_len);
if (rc)
@@ -587,8 +611,7 @@ static ssize_t fsi_slave_sysfs_raw_write(struct file *file,
return -EINVAL;
for (total_len = 0; total_len < count; total_len += write_len) {
- write_len = min_t(size_t, count, 4);
- write_len -= off & 0x3;
+ write_len = aligned_access_size(off, count - total_len);
rc = fsi_slave_write(slave, off, buf + total_len, write_len);
if (rc)
--
2.24.0.rc1
Powered by blists - more mailing lists