FatFs: Update to R0.12a.

This commit is contained in:
Jared Boone
2016-07-28 23:06:21 -07:00
parent 16a6d7efe0
commit d5234633ba
108 changed files with 4627 additions and 2800 deletions

View File

@@ -34,7 +34,7 @@ DRESULT disk_read (
<dt>sector</dt>
<dd>Start sector number in 32-bit LBA.</dd>
<dt>count</dt>
<dd>Number of sectors to read. FatFs specifies the value in range of from 1 to 128.</dd>
<dd>Number of sectors to read.</dd>
</dl>
</div>
@@ -56,12 +56,12 @@ DRESULT disk_read (
<div class="para desc">
<h4>Description</h4>
<p>The data read/write operation to the storage devices is done in unit of <em>sector</em>. FatFs supports the sector size in range of from 512 to 4096 bytes. When FatFs is configured to fixed sector size (<tt>_MIN_SS == MAX_SS</tt>, this will be the most case), the read/write function must work at that sector size. When FatFs is configured to variable sector size (<tt>_MIN_SS &lt; MAX_SS</tt>), sector size is inquired with <tt>disk_ioctl</tt> function following <tt>disk_initialize</tt> function</tt>.
<p>The memory address specified by <tt class="arg">buff</tt> is not that always aligned to word boundary because the argument is defined as <tt>BYTE*</tt>. The misaligned read/write request can occure at <a href="appnote.html#fs1">direct transfer</a>. If the bus architecture, especially DMA controller, does not allow misaligned memory access, it should be solved in this function. There are some workarounds described below to avoid this issue.</p>
<p>The data read/write operation to the storage devices is done in unit of <em>sector</em>. FatFs supports the sector size in range of from 512 to 4096 bytes. When FatFs is configured to fixed sector size (<tt>_MIN_SS == MAX_SS</tt>, this will be the most case), the read/write function must work at that sector size. When FatFs is configured to variable sector size (<tt>_MIN_SS != MAX_SS</tt>), sector size is inquired with <tt>disk_ioctl</tt> function following <tt>disk_initialize</tt> function.</p>
<p>The memory address specified by <tt class="arg">buff</tt> is not that always aligned to word boundary because the argument is defined as <tt>BYTE*</tt>. The unaligned read/write request can occure at <a href="appnote.html#fs1">direct transfer</a>. If the bus architecture, especially DMA controller, does not allow unaligned memory access, it should be solved in this function. There are some workarounds described below to avoid this issue.</p>
<ul>
<li>Convert word transfer to byte transfer in this function if needed. - Recommended.</li>
<li>At the any <tt>f_read</tt> calls, avoid long read request that includes a whole of sector. - Direct transfer will never occure.</li>
<li>At the <tt>f_read(fp, data, btr, &amp;br)</tt> call, make sure that <tt>(((UINT)data &amp; 3) == (f_tell(fp) &amp; 3))</tt> is true. - Word alignment of <tt>buff</tt> is guaranteed.</li>
<li>On the <tt>f_read</tt> calls, avoid long read request that includes a whole of sector. - Any direct transfer never occures.</li>
<li>On the <tt>f_read</tt> calls, make sure that <tt>(((UINT)data &amp; 3) == (f_tell(fp) &amp; 3))</tt> is true. - Word alignment of <tt>buff</tt> is guaranteed.</li>
</ul>
<p>Generally, a multiple sector transfer request must not be split into single sector transactions to the storage device, or you will not get good read throughput.</p>
</div>