From: Wang Nan <wangnan0@huawei.com>
To: Petr Tesarik <ptesarik@suse.cz>
Cc: kexec@lists.infradead.org, kumagai-atsushi@mxc.nes.nec.co.jp,
Geng Hui <hui.geng@huawei.com>
Subject: Re: [PATCH v2] makedumpfile: code cleanup: set_bitmap
Date: Fri, 25 Apr 2014 13:25:23 +0800 [thread overview]
Message-ID: <5359F1C3.5030100@huawei.com> (raw)
In-Reply-To: <20140424112815.398f9852@hananiah.suse.cz>
On 2014/4/24 17:28, Petr Tesarik wrote:
> On Thu, 24 Apr 2014 17:11:13 +0800
> Wang Nan <wangnan0@huawei.com> wrote:
>
>> This patch makes set_bitmap() to call sync_bitmap() instead rewrite
>> identical code to do same thing.
>>
>> Change from v1:
>>
>> - fix a simple mistake:
>> sync_bitmap() returns TRUE(1) when it succeeds, so use
>> (!sync_bitmap()) for checking.
>
> Hi Wang Nan,
>
> I like your change. See my comments below:
>
>> Signed-off-by: Wang Nan <wangnan0@huawei.com>
>> Cc: Atsushi Kumagai <kumagai-atsushi@mxc.nes.nec.co.jp>
>> Cc: kexec@lists.infradead.org
>> Cc: Geng Hui <hui.geng@huawei.com>
>> ---
>> makedumpfile.c | 70 ++++++++++++++++++++++++++--------------------------------
>> 1 file changed, 31 insertions(+), 39 deletions(-)
>>
>> diff --git a/makedumpfile.c b/makedumpfile.c
>> index ce4a866..cea37a2 100644
>> --- a/makedumpfile.c
>> +++ b/makedumpfile.c
>> @@ -3309,6 +3309,34 @@ initialize_2nd_bitmap(struct dump_bitmap *bitmap)
>> }
>>
>> int
>> +sync_bitmap(struct dump_bitmap *bitmap)
>> +{
>> + off_t offset;
>> + offset = bitmap->offset + BUFSIZE_BITMAP * bitmap->no_block;
>> +
>> + /*
>> + * The bitmap buffer is not dirty, and it is not necessary
>> + * to write out it.
>> + */
>> + if (bitmap->no_block < 0)
>> + return TRUE;
>> +
>> + if (lseek(bitmap->fd, offset, SEEK_SET) < 0 ) {
[ .. see below .. ]
>> + ERRMSG("Can't seek the bitmap(%s). %s\n",
>> + bitmap->file_name, strerror(errno));
>> + return FALSE;
>> + }
>> + if (write(bitmap->fd, bitmap->buf, BUFSIZE_BITMAP)
>> + != BUFSIZE_BITMAP) {
>> + ERRMSG("Can't write the bitmap(%s). %s\n",
>> + bitmap->file_name, strerror(errno));
>> + return FALSE;
>> + }
>> + return TRUE;
>> +}
>> +
>> +
>> +int
>> set_bitmap(struct dump_bitmap *bitmap, unsigned long long pfn,
>> int val)
>> {
>> @@ -3317,20 +3345,11 @@ set_bitmap(struct dump_bitmap *bitmap, unsigned long long pfn,
>> old_offset = bitmap->offset + BUFSIZE_BITMAP * bitmap->no_block;
>> new_offset = bitmap->offset + BUFSIZE_BITMAP * (pfn / PFN_BUFBITMAP);
>
> After applying this patch, both old_offset and new_offset are
> calculated only to compare for equality. And new_offset is in fact
> computed twice (once in set_bitmap and once in sync_bitmap).
>
> This could be cleaned up by replacing the offsets with:
>
> int new_no_block = pfn / PFN_BUFBITMAP;
>
> Then change all occurrences of if (old_offset != new_offset) to:
>
> if (bitmap->no_block != new_no_block)
>
> and finally re-use new_no_block in the assignment to bitmap->no_block
> near the end of the function, like this:
>
> - bitmap->no_block = pfn / PFN_BUFBITMAP;
> + bitmap->no_block = new_no_block;
>
> Petr T
>
>>
>> - if (0 <= bitmap->no_block && old_offset != new_offset) {
>> - if (lseek(bitmap->fd, old_offset, SEEK_SET) < 0 ) {
>> - ERRMSG("Can't seek the bitmap(%s). %s\n",
>> - bitmap->file_name, strerror(errno));
>> - return FALSE;
>> - }
>> - if (write(bitmap->fd, bitmap->buf, BUFSIZE_BITMAP)
>> - != BUFSIZE_BITMAP) {
>> - ERRMSG("Can't write the bitmap(%s). %s\n",
>> - bitmap->file_name, strerror(errno));
>> + if (old_offset != new_offset) {
>> + if (!sync_bitmap(bitmap)) {
>> + ERRMSG("Can't sync bitmap\n");
>> return FALSE;
>> }
>> - }
>> - if (old_offset != new_offset) {
>> if (lseek(bitmap->fd, new_offset, SEEK_SET) < 0 ) {
>> ERRMSG("Can't seek the bitmap(%s). %s\n",
>> bitmap->file_name, strerror(errno));
[ .. see below .. ]
Good suggestion, but new_offset is still needed to be calculated for these lseeks.
In addition, I have another idea: what about to replace all lseek .. read/write pattern to pread/pwrite?
>> @@ -3386,33 +3405,6 @@ set_bitmap_cyclic(char *bitmap, unsigned long long pfn, int val, struct cycle *c
>> }
>>
>> int
>> -sync_bitmap(struct dump_bitmap *bitmap)
>> -{
>> - off_t offset;
>> - offset = bitmap->offset + BUFSIZE_BITMAP * bitmap->no_block;
>> -
>> - /*
>> - * The bitmap buffer is not dirty, and it is not necessary
>> - * to write out it.
>> - */
>> - if (bitmap->no_block < 0)
>> - return TRUE;
>> -
>> - if (lseek(bitmap->fd, offset, SEEK_SET) < 0 ) {
>> - ERRMSG("Can't seek the bitmap(%s). %s\n",
>> - bitmap->file_name, strerror(errno));
>> - return FALSE;
>> - }
>> - if (write(bitmap->fd, bitmap->buf, BUFSIZE_BITMAP)
>> - != BUFSIZE_BITMAP) {
>> - ERRMSG("Can't write the bitmap(%s). %s\n",
>> - bitmap->file_name, strerror(errno));
>> - return FALSE;
>> - }
>> - return TRUE;
>> -}
>> -
>> -int
>> sync_1st_bitmap(void)
>> {
>> return sync_bitmap(info->bitmap1);
>
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
next prev parent reply other threads:[~2014-04-25 5:26 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-04-14 6:15 [PATCH] makedumpfile: code cleanup: set_bitmap Wang Nan
2014-04-21 7:33 ` Atsushi Kumagai
2014-04-24 9:11 ` [PATCH v2] " Wang Nan
2014-04-24 9:28 ` Petr Tesarik
2014-04-25 5:25 ` Wang Nan [this message]
2014-04-25 8:03 ` Petr Tesarik
2014-04-26 4:31 ` Wang Nan
2014-04-28 7:27 ` Atsushi Kumagai
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=5359F1C3.5030100@huawei.com \
--to=wangnan0@huawei.com \
--cc=hui.geng@huawei.com \
--cc=kexec@lists.infradead.org \
--cc=kumagai-atsushi@mxc.nes.nec.co.jp \
--cc=ptesarik@suse.cz \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.