Message ID | 1478252037-5340-1-git-send-email-yousaf.kaukab@suse.com (mailing list archive) |
---|---|
State | Superseded |
Headers | show |
Hi, On Fri, Nov 4, 2016 at 7:33 AM, Mian Yousaf Kaukab <yousaf.kaukab@suse.com> wrote: > Only print actual cyclic dependencies. Don't print count of all the > modules as it includes other modules which have dependencies but not > necessarily cyclic. > > Printing related modules causes buffer overflow as m->modnamesz is not > included in buffer size calculations (loop == m is never true). > This buffer overflow causes kmod to crash. > > Reported-by: Andreas Färber <afaerber@suse.de> > Signed-off-by: Mian Yousaf Kaukab <yousaf.kaukab@suse.com> > --- > As count of modules in cyclic dependency chain is not known at the > start of this function, so it is not printed anymore. The output > when cyclic dependency is detected is changed as following: > > Old: > depmod: ERROR: Found 8 modules in dependency cycles! > > New: > depmod: ERROR: Modules found in dependency cycles! I think it would be good to fix this problem, but retaining the behavior. My first reaction to this is "how do I reproduce the issue?". This number so far does tell us how many modules are involved in loops. So, for example: A -> B -> D -> A B -> C -> B This line should point to 4 modules. We don't know which module is wrong. But there are 4 modules involved in the loops. D could be the culprit. The function should show what are the loops present (in "first found order" - it could very well report as "B -> D -> A -> B" the line above). We have a testsuite to check if the loop detection is working in depmod. See testsuite/test-depmod.c:depmod_detect_loop(). Would it be possible to add your use case there? If you tell me what are your dependency chain and cycle I can take a look. If I misunderstood the issue with the total number, then at least the testsuite needs to be fixed, because right now it expects that number to be in the output (see testsuite/rootfs-pristine/test-depmod/detect-loop/correct.txt) thanks Lucas De Marchi -- To unsubscribe from this list: send the line "unsubscribe linux-modules" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Sat, 2016-11-05 at 21:50 -0200, Lucas De Marchi wrote: > Hi, > > On Fri, Nov 4, 2016 at 7:33 AM, Mian Yousaf Kaukab > <yousaf.kaukab@suse.com> wrote: > > > > Only print actual cyclic dependencies. Don't print count of all the > > modules as it includes other modules which have dependencies but > > not > > necessarily cyclic. > > > > Printing related modules causes buffer overflow as m->modnamesz is > > not > > included in buffer size calculations (loop == m is never true). > > This buffer overflow causes kmod to crash. > > > > Reported-by: Andreas Färber <afaerber@suse.de> > > Signed-off-by: Mian Yousaf Kaukab <yousaf.kaukab@suse.com> > > --- > > As count of modules in cyclic dependency chain is not known at the > > start of this function, so it is not printed anymore. The output > > when cyclic dependency is detected is changed as following: > > > > Old: > > depmod: ERROR: Found 8 modules in dependency cycles! > > > > New: > > depmod: ERROR: Modules found in dependency cycles! > > I think it would be good to fix this problem, but retaining the > behavior. Would it be OK if the modules names are printed first and count is printed afterward. Something like following: DEPMOD 4.9.0-rc4-default depmod: ERROR: Cycle detected: qcom_wcnss_iris -> qcom_wcnss -> qcom_wcnss_iris depmod: ERROR: Found 2 modules in dependency cycles! Makefile:1227: recipe for target '_modinst_post' failed In this way modules names can still be printed in the loop and then the exact count of modules involved in cyclic dependency is printed at the end of the depmod_report_cycles(). > > My first reaction to this is "how do I reproduce the issue?". This > number so far does tell us how many modules are involved in loops. There is more info at the following link: https://bugzilla.opensuse.org/show_bug.cgi?id=1008186 You can reproduce the issue by enabling CONFIG_QCOM_WCNSS_PIL as module in v4.9-rc4 for arm64. +CONFIG_REMOTEPROC=m +CONFIG_QCOM_MDT_LOADER=m -# CONFIG_QCOM_WCNSS_PIL is not set +CONFIG_QCOM_WCNSS_IRIS=m +CONFIG_QCOM_WCNSS_PIL=m Issue will reproduce when depmod is run at the end of modules_install. > So, > for example: > > A -> B -> D -> A > B -> C -> B C -> X C -> Y C -> Z In this case count becomes 7. Which is misleading. > This line should point to 4 modules. We don't know which module is > wrong. But there are 4 modules involved in the loops. D could be the > culprit. > > The function should show what are the loops present (in "first found > order" - it could very well report as "B -> D -> A -> B" the line > above). > > We have a testsuite to check if the loop detection is working in > depmod. See testsuite/test-depmod.c:depmod_detect_loop(). > Would it be possible to add your use case there? [...] Yes I can take a look. > [...] If you tell me what > are your dependency chain and cycle I can take a look. > > If I misunderstood the issue with the total number, then at least the > testsuite needs to be fixed, because right now it expects that number > to be in the output (see > testsuite/rootfs-pristine/test-depmod/detect-loop/correct.txt) > > > thanks > Lucas De Marchi > BR, Yousaf -- To unsubscribe from this list: send the line "unsubscribe linux-modules" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon 07 Nov 09:27 PST 2016, Mian Yousaf Kaukab wrote: > On Sat, 2016-11-05 at 21:50 -0200, Lucas De Marchi wrote: > > Hi, > > > > On Fri, Nov 4, 2016 at 7:33 AM, Mian Yousaf Kaukab > > <yousaf.kaukab@suse.com> wrote: > > > > > > Only print actual cyclic dependencies. Don't print count of all the > > > modules as it includes other modules which have dependencies but > > > not > > > necessarily cyclic. > > > > > > Printing related modules causes buffer overflow as m->modnamesz is > > > not > > > included in buffer size calculations (loop == m is never true). > > > This buffer overflow causes kmod to crash. > > > > > > Reported-by: Andreas Färber <afaerber@suse.de> > > > Signed-off-by: Mian Yousaf Kaukab <yousaf.kaukab@suse.com> > > > --- > > > As count of modules in cyclic dependency chain is not known at the > > > start of this function, so it is not printed anymore. The output > > > when cyclic dependency is detected is changed as following: > > > > > > Old: > > > depmod: ERROR: Found 8 modules in dependency cycles! > > > > > > New: > > > depmod: ERROR: Modules found in dependency cycles! > > > > I think it would be good to fix this problem, but retaining the > > behavior. > Would it be OK if the modules names are printed first and count is > printed afterward. Something like following: > > DEPMOD 4.9.0-rc4-default > depmod: ERROR: Cycle detected: qcom_wcnss_iris -> qcom_wcnss -> > qcom_wcnss_iris > depmod: ERROR: Found 2 modules in dependency cycles! > Makefile:1227: recipe for target '_modinst_post' failed > > In this way modules names can still be printed in the loop and then the > exact count of modules involved in cyclic dependency is printed at the > end of the depmod_report_cycles(). > > > > > My first reaction to this is "how do I reproduce the issue?". This > > number so far does tell us how many modules are involved in loops. > There is more info at the following link: > https://bugzilla.opensuse.org/show_bug.cgi?id=1008186 > > You can reproduce the issue by enabling CONFIG_QCOM_WCNSS_PIL as module > in v4.9-rc4 for arm64. > +CONFIG_REMOTEPROC=m > +CONFIG_QCOM_MDT_LOADER=m > -# CONFIG_QCOM_WCNSS_PIL is not set > +CONFIG_QCOM_WCNSS_IRIS=m > +CONFIG_QCOM_WCNSS_PIL=m > I added some debugging prints to track the stack management in depmod_report_cycles(). remoteproc, mdt_loader, wcnss_iris and wcnss_pil are all roots and we have dependencies like this. /---(rproc) | ^ | | (pil) -+-> (mdt) ^ | | v +- (iris) 1) We pick rproc as first root, mark that as visited and see that we're done. 2) We pick mdt_loader as second root, we push remoteproc to the stack and find that it's already visited, so we found a loop! mdt_loader -> remoteproc 3) We pick iris as third root, we push wcnss which pushes remoteproc, mdt_loader and iris. All three have already been visited from root #1 and #2, so we find that there's a loop in: iris -> wcnss -> iris iris -> wcnss -> mdt iris -> wcnss -> remoteproc Only one of these cases are actually a cycle, but as we don't reset visited between the searches we can't tell. Further more, if there was a dependency from iris -> remoteproc that would have shown up earlier and marked iris->visited and when we get to step #3 we would just have bailed directly - completely missing the cycle. So I think we need to reset the visited list on each run of the DFS from each root. Regards, Bjorn -- To unsubscribe from this list: send the line "unsubscribe linux-modules" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, 2016-11-07 at 12:23 -0800, Bjorn Andersson wrote: > On Mon 07 Nov 09:27 PST 2016, Mian Yousaf Kaukab wrote: > > > > > On Sat, 2016-11-05 at 21:50 -0200, Lucas De Marchi wrote: > > > > > > Hi, > > > > > > On Fri, Nov 4, 2016 at 7:33 AM, Mian Yousaf Kaukab > > > <yousaf.kaukab@suse.com> wrote: > > > > > > > > > > > > Only print actual cyclic dependencies. Don't print count of all > > > > the > > > > modules as it includes other modules which have dependencies > > > > but > > > > not > > > > necessarily cyclic. > > > > > > > > Printing related modules causes buffer overflow as m->modnamesz > > > > is > > > > not > > > > included in buffer size calculations (loop == m is never true). > > > > This buffer overflow causes kmod to crash. > > > > > > > > Reported-by: Andreas Färber <afaerber@suse.de> > > > > Signed-off-by: Mian Yousaf Kaukab <yousaf.kaukab@suse.com> > > > > --- > > > > As count of modules in cyclic dependency chain is not known at > > > > the > > > > start of this function, so it is not printed anymore. The > > > > output > > > > when cyclic dependency is detected is changed as following: > > > > > > > > Old: > > > > depmod: ERROR: Found 8 modules in dependency cycles! > > > > > > > > New: > > > > depmod: ERROR: Modules found in dependency cycles! > > > > > > I think it would be good to fix this problem, but retaining the > > > behavior. > > Would it be OK if the modules names are printed first and count is > > printed afterward. Something like following: > > > > DEPMOD 4.9.0-rc4-default > > depmod: ERROR: Cycle detected: qcom_wcnss_iris -> qcom_wcnss -> > > qcom_wcnss_iris > > depmod: ERROR: Found 2 modules in dependency cycles! > > Makefile:1227: recipe for target '_modinst_post' failed > > > > In this way modules names can still be printed in the loop and then > > the > > exact count of modules involved in cyclic dependency is printed at > > the > > end of the depmod_report_cycles(). > > > > > > > > > > > My first reaction to this is "how do I reproduce the issue?". > > > This > > > number so far does tell us how many modules are involved in > > > loops. > > There is more info at the following link: > > https://bugzilla.opensuse.org/show_bug.cgi?id=1008186 > > > > You can reproduce the issue by enabling CONFIG_QCOM_WCNSS_PIL as > > module > > in v4.9-rc4 for arm64. > > +CONFIG_REMOTEPROC=m > > +CONFIG_QCOM_MDT_LOADER=m > > -# CONFIG_QCOM_WCNSS_PIL is not set > > +CONFIG_QCOM_WCNSS_IRIS=m > > +CONFIG_QCOM_WCNSS_PIL=m > > > > I added some debugging prints to track the stack management in > depmod_report_cycles(). > > remoteproc, mdt_loader, wcnss_iris and wcnss_pil are all roots and we > have dependencies like this. > > /---(rproc) > | ^ > | | > (pil) -+-> (mdt) > ^ | > | v > +- (iris) AFAICT this is not correct. Dependencies are like following: pil -> rproc pil -> mtd -> rproc pil -> iris -> pil > > 1) We pick rproc as first root, mark that as visited and see that > we're > done. > > 2) We pick mdt_loader as second root, we push remoteproc to the stack > and find that it's already visited, so we found a loop! > > mdt_loader -> remoteproc This is not cyclic. > > 3) We pick iris as third root, we push wcnss which pushes remoteproc, > mdt_loader and iris. All three have already been visited from root > #1 > and #2, so we find that there's a loop in: > > iris -> wcnss -> iris > iris -> wcnss -> mdt > iris -> wcnss -> remoteproc > > Only one of these cases are actually a cycle, but as we don't reset > visited between the searches we can't tell. Further more, if there > was a > dependency from iris -> remoteproc that would have shown up earlier > and > marked iris->visited and when we get to step #3 we would just have > bailed directly - completely missing the cycle. In case we have more than one cyclic dependencies around same modules? May be we should add a testcase for such a scenario so that its easy to understand and reproduce. > > So I think we need to reset the visited list on each run of the DFS > from > each root. > > Regards, > Bjorn BR, Yousaf -- To unsubscribe from this list: send the line "unsubscribe linux-modules" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue 08 Nov 01:36 PST 2016, Mian Yousaf Kaukab wrote: > On Mon, 2016-11-07 at 12:23 -0800, Bjorn Andersson wrote: > > On Mon 07 Nov 09:27 PST 2016, Mian Yousaf Kaukab wrote: [..] > > > > I added some debugging prints to track the stack management in > > depmod_report_cycles(). > > > > remoteproc, mdt_loader, wcnss_iris and wcnss_pil are all roots and we > > have dependencies like this. > > > > /---(rproc) > > | ^ > > | | > > (pil) -+-> (mdt) > > ^ | > > | v > > +- (iris) > AFAICT this is not correct. Dependencies are like following: > > pil -> rproc > pil -> mtd -> rproc > pil -> iris -> pil All dependencies are: pil -> rproc pil -> mdt mdt -> rproc pil -> iris iris -> pil The last two forming a cyclic subgraph in a dependency graph including all of them. > > > > > 1) We pick rproc as first root, mark that as visited and see that > > we're > > done. > > > > 2) We pick mdt_loader as second root, we push remoteproc to the stack > > and find that it's already visited, so we found a loop! > > > > mdt_loader -> remoteproc > This is not cyclic. > Exactly my point. But the DFS will find this cycle, as we don't reset "visited" between the iterations in the loop. > > > > 3) We pick iris as third root, we push wcnss which pushes remoteproc, > > mdt_loader and iris. All three have already been visited from root > > #1 > > and #2, so we find that there's a loop in: > > > > iris -> wcnss -> iris > > iris -> wcnss -> mdt > > iris -> wcnss -> remoteproc > > > > Only one of these cases are actually a cycle, but as we don't reset > > visited between the searches we can't tell. Further more, if there > > was a > > dependency from iris -> remoteproc that would have shown up earlier > > and > > marked iris->visited and when we get to step #3 we would just have > > bailed directly - completely missing the cycle. > In case we have more than one cyclic dependencies around same modules? Well in this case we have a mixture of cyclic and non-cyclic subgraphs that the DFS hits due to the "visited" issue above, the non-cyclic ones are what triggers the buffer overflow. > May be we should add a testcase for such a scenario so that its easy to > understand and reproduce. > +1 Thanks for looking at this! Regards, Bjorn -- To unsubscribe from this list: send the line "unsubscribe linux-modules" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/tools/depmod.c b/tools/depmod.c index ad01f66..2813ef1 100644 --- a/tools/depmod.c +++ b/tools/depmod.c @@ -1456,7 +1456,7 @@ static void depmod_report_cycles(struct depmod *depmod, uint16_t n_mods, { const char sep[] = " -> "; int ir = 0; - ERR("Found %u modules in dependency cycles!\n", n_roots); + ERR("Modules found in dependency cycles!\n"); while (n_roots > 0) { int is, ie; @@ -1491,6 +1491,7 @@ static void depmod_report_cycles(struct depmod *depmod, uint16_t n_mods, if (m->visited) { int i, n = 0, sz = 0; char *buf; + bool is_cyclic = false; for (i = ie - 1; i >= 0; i--) { struct mod *loop = depmod->modules.array[edges[i]]; @@ -1498,9 +1499,15 @@ static void depmod_report_cycles(struct depmod *depmod, uint16_t n_mods, n++; if (loop == m) { sz += loop->modnamesz - 1; + is_cyclic = true; break; } } + /* Current module not found in dependency list. + * Must be a related module. Ignore it. + */ + if (!is_cyclic) + continue; buf = malloc(sz + n * strlen(sep) + 1); sz = 0;
Only print actual cyclic dependencies. Don't print count of all the modules as it includes other modules which have dependencies but not necessarily cyclic. Printing related modules causes buffer overflow as m->modnamesz is not included in buffer size calculations (loop == m is never true). This buffer overflow causes kmod to crash. Reported-by: Andreas Färber <afaerber@suse.de> Signed-off-by: Mian Yousaf Kaukab <yousaf.kaukab@suse.com> --- As count of modules in cyclic dependency chain is not known at the start of this function, so it is not printed anymore. The output when cyclic dependency is detected is changed as following: Old: depmod: ERROR: Found 8 modules in dependency cycles! New: depmod: ERROR: Modules found in dependency cycles! I hope it wouldn't be a problem. tools/depmod.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)