Re: [U-Boot] [PATCH 3/7] JFFS2: Calculate buf_len before we read data from flash
Hi Baidu, > Hi, Detlev : > > 2011/4/30 Detlev Zundel : >> Hi Baidu, >> >> [...] >> > diff --git a/fs/jffs2/jffs2_1pass.c b/fs/jffs2/jffs2_1pass.c > index 8eb77b1..be6ac78 100644 > --- a/fs/jffs2/jffs2_1pass.c > +++ b/fs/jffs2/jffs2_1pass.c > @@ -1643,6 +1643,8 @@ jffs2_1pass_build_lists(struct part_info * part) > case JFFS2_NODETYPE_INODE: > if (buf_ofs + buf_len < ofs + sizeof(struct > jffs2_raw_inode)) { > + buf_len = min_t(uint32_t, buf_size, > sector_ofs > + + part->sector_size - ofs); I am somewhat uncomfortable that "buf_len" is used in the if condition before it is recalculated inside. Are you sure that this cannot lead to problems, i.e. can buf_len become larger than it was when entering the condition? >>> >>> It works well in my boards. >> >> Aha. I am all happy for you that this works, but we really need to be >> as sure as we can be that the code does the right thing. "It works for >> me" is not enough. > > Please read every line code in the uboot jffs2 carefully. Know more > about JFFS2. I was asking you a question about a potential problem that I see with your changes. Can you answer the question or not? If you cannot answer it, then you are changing things that you do not understand. Personally I don't trust bug-fixes that are not understood completely. Cheers Detlev -- Mit einem Leben wie dem meinen, Doktor - wer braucht da noch Traeume? - Alex Portnoy -- DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-40 Fax: (+49)-8142-66989-80 Email: d...@denx.de ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH 3/7] JFFS2: Calculate buf_len before we read data from flash
Hi, Detlev : 2011/4/30 Detlev Zundel : > Hi Baidu, > > [...] > diff --git a/fs/jffs2/jffs2_1pass.c b/fs/jffs2/jffs2_1pass.c index 8eb77b1..be6ac78 100644 --- a/fs/jffs2/jffs2_1pass.c +++ b/fs/jffs2/jffs2_1pass.c @@ -1643,6 +1643,8 @@ jffs2_1pass_build_lists(struct part_info * part) case JFFS2_NODETYPE_INODE: if (buf_ofs + buf_len < ofs + sizeof(struct jffs2_raw_inode)) { + buf_len = min_t(uint32_t, buf_size, sector_ofs + + part->sector_size - ofs); >>> >>> I am somewhat uncomfortable that "buf_len" is used in the if condition >>> before it is recalculated inside. Are you sure that this cannot lead to >>> problems, i.e. can buf_len become larger than it was when entering the >>> condition? >> >> It works well in my boards. > > Aha. I am all happy for you that this works, but we really need to be > as sure as we can be that the code does the right thing. "It works for > me" is not enough. Please read every line code in the uboot jffs2 carefully. Know more about JFFS2. ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH 3/7] JFFS2: Calculate buf_len before we read data from flash
Hi Baidu, [...] >>> diff --git a/fs/jffs2/jffs2_1pass.c b/fs/jffs2/jffs2_1pass.c >>> index 8eb77b1..be6ac78 100644 >>> --- a/fs/jffs2/jffs2_1pass.c >>> +++ b/fs/jffs2/jffs2_1pass.c >>> @@ -1643,6 +1643,8 @@ jffs2_1pass_build_lists(struct part_info * part) >>> case JFFS2_NODETYPE_INODE: >>> if (buf_ofs + buf_len < ofs + sizeof(struct >>> jffs2_raw_inode)) { >>> + buf_len = min_t(uint32_t, buf_size, >>> sector_ofs >>> + + part->sector_size - ofs); >> >> I am somewhat uncomfortable that "buf_len" is used in the if condition >> before it is recalculated inside. Are you sure that this cannot lead to >> problems, i.e. can buf_len become larger than it was when entering the >> condition? > > It works well in my boards. Aha. I am all happy for you that this works, but we really need to be as sure as we can be that the code does the right thing. "It works for me" is not enough. Cheers Detlev -- Als ich entführt wurde, begannen meine Eltern aktiv zu werden. Sie vermieteten mein Zimmer. -- Woody Allen -- DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-40 Fax: (+49)-8142-66989-80 Email: d...@denx.de ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH 3/7] JFFS2: Calculate buf_len before we read data from flash
Hi,Detlev 2011/4/29 Detlev Zundel : > Hi Baidu, > >> 1/ We should calculate the buf_len before we call >> get_fl_mem(). > > If I read your change correctly, then do you mean the following? > > When we know what we want to read, we can calculate buf_len to be the > maximum size of the data to be read. Without this, we usually read > EMPTY_SCAN_SIZE which is too much. Yes,you are right. > This is only an optimization change. > > Maybe this text will help people understand better what the change tries > to do. The patch comments may be enough to explain what my changes is. We can see the code: buf_len = min_t() everywhere the get_fl_mem() called. >> diff --git a/fs/jffs2/jffs2_1pass.c b/fs/jffs2/jffs2_1pass.c >> index 8eb77b1..be6ac78 100644 >> --- a/fs/jffs2/jffs2_1pass.c >> +++ b/fs/jffs2/jffs2_1pass.c >> @@ -1643,6 +1643,8 @@ jffs2_1pass_build_lists(struct part_info * part) >> case JFFS2_NODETYPE_INODE: >> if (buf_ofs + buf_len < ofs + sizeof(struct >> jffs2_raw_inode)) { >> + buf_len = min_t(uint32_t, buf_size, >> sector_ofs >> + + part->sector_size - ofs); > > I am somewhat uncomfortable that "buf_len" is used in the if condition > before it is recalculated inside. Are you sure that this cannot lead to > problems, i.e. can buf_len become larger than it was when entering the > condition? It works well in my boards. ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH 3/7] JFFS2: Calculate buf_len before we read data from flash
Hi,Detlev 2011/4/29 Detlev Zundel : > Hi Baidu, > >> 1/ We should calculate the buf_len before we call >> get_fl_mem(). > > If I read your change correctly, then do you mean the following? > > When we know what we want to read, we can calculate buf_len to be the > maximum size of the data to be read. Without this, we usually read > EMPTY_SCAN_SIZE which is too much. Yes,you are right. > This is only an optimization change. > > Maybe this text will help people understand better what the change tries > to do. The patch comments may be enough to explain what my changes is. We can see the code: buf_len = min_t() everywhere the get_fl_mem() called. >> diff --git a/fs/jffs2/jffs2_1pass.c b/fs/jffs2/jffs2_1pass.c >> index 8eb77b1..be6ac78 100644 >> --- a/fs/jffs2/jffs2_1pass.c >> +++ b/fs/jffs2/jffs2_1pass.c >> @@ -1643,6 +1643,8 @@ jffs2_1pass_build_lists(struct part_info * part) >> case JFFS2_NODETYPE_INODE: >> if (buf_ofs + buf_len < ofs + sizeof(struct >> jffs2_raw_inode)) { >> + buf_len = min_t(uint32_t, buf_size, >> sector_ofs >> + + part->sector_size - ofs); > > I am somewhat uncomfortable that "buf_len" is used in the if condition > before it is recalculated inside. Are you sure that this cannot lead to > problems, i.e. can buf_len become larger than it was when entering the > condition? It works well in my boards. ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH 3/7] JFFS2: Calculate buf_len before we read data from flash
Hi Baidu, > 1/ We should calculate the buf_len before we call > get_fl_mem(). If I read your change correctly, then do you mean the following? When we know what we want to read, we can calculate buf_len to be the maximum size of the data to be read. Without this, we usually read EMPTY_SCAN_SIZE which is too much. This is only an optimization change. Maybe this text will help people understand better what the change tries to do. > diff --git a/fs/jffs2/jffs2_1pass.c b/fs/jffs2/jffs2_1pass.c > index 8eb77b1..be6ac78 100644 > --- a/fs/jffs2/jffs2_1pass.c > +++ b/fs/jffs2/jffs2_1pass.c > @@ -1643,6 +1643,8 @@ jffs2_1pass_build_lists(struct part_info * part) > case JFFS2_NODETYPE_INODE: > if (buf_ofs + buf_len < ofs + sizeof(struct > jffs2_raw_inode)) { > + buf_len = min_t(uint32_t, buf_size, > sector_ofs > + + part->sector_size - ofs); > I am somewhat uncomfortable that "buf_len" is used in the if condition before it is recalculated inside. Are you sure that this cannot lead to problems, i.e. can buf_len become larger than it was when entering the condition? > get_fl_mem((u32)part->offset + ofs, > buf_len, buf); > buf_ofs = ofs; > @@ -1659,11 +1661,11 @@ jffs2_1pass_build_lists(struct part_info * part) > } > break; > case JFFS2_NODETYPE_DIRENT: > - if (buf_ofs + buf_len < ofs + sizeof(struct > - jffs2_raw_dirent) + > - ((struct > - jffs2_raw_dirent *) > - node)->nsize) { > + if (buf_ofs + buf_len < ofs + > + sizeof(struct jffs2_raw_dirent) + > + ((struct jffs2_raw_dirent > *)node)->nsize) { > + buf_len = min_t(uint32_t, buf_size, > sector_ofs > + + part->sector_size - ofs); The same question for this case. Cheers Detlev -- Those who do not understand Unix are condemned to reinvent it, poorly. - Henry Spencer, University of Toronto Unix hack -- DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-40 Fax: (+49)-8142-66989-80 Email: d...@denx.de ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot