On Fri, Aug 26, 2022 at 02:51:15PM +0200, Paul Menzel wrote: > Dear Jarkko, > > > Thank you for the patch. No, thank you for reporting this and all the help with testing the change :-) > Am 26.08.22 um 03:41 schrieb Jarkko Sakkinen: > > In sgx_init(), if misc_register() for the provision device fails, and > > neither sgx_drv_init() nor sgx_vepc_init() succeeds, then ksgxd will be > > prematurely stopped. > > > > This triggers WARN_ON() because sgx_dirty_page_list ends up being > > non-empty. Ultimately this can crash the kernel, depending on the kernel > > command line, which is not correct behavior because SGX driver is not > > working incorrectly. > > Maybe paste the WARN_ON trace, so `git log` can be searched for the trace > too. It's a decent suggestion, I agree. It would be probably also good to mention /proc/sys/kernel/panic_on_warn. When you set that to '1' WARN() will tear down the whole kernel. > > > Print simple warning instead, and improve the output by printing the > > number of unsanitized pages. > > See below, but no warning seems to be logged in my case now. (I should test > Linus’ current master too.) > > > Link: https://lore.kernel.org/linux-sgx/20220825051827.246698-1-jarkko@xxxxxxxxxx/T/#u > > Reported-by: Paul Menzel <pmenzel@xxxxxxxxxxxxx> > > Fixes: 51ab30eb2ad4 ("x86/sgx: Replace section->init_laundry_list with sgx_dirty_page_list") > > Signed-off-by: Jarkko Sakkinen <jarkko@xxxxxxxxxx> > > --- > > Cc: Haitao Huang <haitao.huang@xxxxxxxxxxxxxxx> > > Cc: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx> > > Cc: Reinette Chatre <reinette.chatre@xxxxxxxxx> > > > > v4: > > - Explain expectations for dirty_page_list in the function header, instead > > of an inline comment. > > - Improve commit message to explain the conditions better. > > - Return the number of pages left dirty to ksgxd() and print warning after > > the 2nd call, if there are any. > > > > v3: > > - Remove WARN_ON(). > > - Tuned comments and the commit message a bit. > > > > v2: > > - Replaced WARN_ON() with optional pr_info() inside > > __sgx_sanitize_pages(). > > - Rewrote the commit message. > > - Added the fixes tag. > > --- > > arch/x86/kernel/cpu/sgx/main.c | 19 +++++++++++++------ > > 1 file changed, 13 insertions(+), 6 deletions(-) > > > > diff --git a/arch/x86/kernel/cpu/sgx/main.c b/arch/x86/kernel/cpu/sgx/main.c > > index 515e2a5f25bb..903100fcfce3 100644 > > --- a/arch/x86/kernel/cpu/sgx/main.c > > +++ b/arch/x86/kernel/cpu/sgx/main.c > > @@ -49,17 +49,20 @@ static LIST_HEAD(sgx_dirty_page_list); > > * Reset post-kexec EPC pages to the uninitialized state. The pages are removed > > * from the input list, and made available for the page allocator. SECS pages > > * prepending their children in the input list are left intact. > > + * > > + * Contents of the @dirty_page_list must be thread-local, i.e. > > + * not shared by multiple threads. > > */ > > -static void __sgx_sanitize_pages(struct list_head *dirty_page_list) > > +static int __sgx_sanitize_pages(struct list_head *dirty_page_list) > > { > > struct sgx_epc_page *page; > > + int left_dirty = 0; > > LIST_HEAD(dirty); > > int ret; > > - /* dirty_page_list is thread-local, no need for a lock: */ > > while (!list_empty(dirty_page_list)) { > > if (kthread_should_stop()) > > - return; > > + break; > > page = list_first_entry(dirty_page_list, struct sgx_epc_page, list); > > @@ -92,12 +95,14 @@ static void __sgx_sanitize_pages(struct list_head *dirty_page_list) > > } else { > > /* The page is not yet clean - move to the dirty list. */ > > list_move_tail(&page->list, &dirty); > > + left_dirty++; > > } > > cond_resched(); > > } > > list_splice(&dirty, dirty_page_list); > > + return left_dirty; > > } > > static bool sgx_reclaimer_age(struct sgx_epc_page *epc_page) > > @@ -388,6 +393,8 @@ void sgx_reclaim_direct(void) > > static int ksgxd(void *p) > > { > > + int left_dirty; > > + > > set_freezable(); > > /* > > @@ -395,10 +402,10 @@ static int ksgxd(void *p) > > * required for SECS pages, whose child pages blocked EREMOVE. > > */ > > __sgx_sanitize_pages(&sgx_dirty_page_list); > > - __sgx_sanitize_pages(&sgx_dirty_page_list); > > - /* sanity check: */ > > - WARN_ON(!list_empty(&sgx_dirty_page_list)); > > + left_dirty = __sgx_sanitize_pages(&sgx_dirty_page_list); > > + if (left_dirty) > > + pr_warn("%d unsanitized pages\n", left_dirty); > > while (!kthread_should_stop()) { > > if (try_to_freeze()) > > I tested this on top of commit 4c612826bec1 (Merge tag 'net-6.0-rc3' of > git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net) and the warning > trace is gone. > > [ 0.255192] calling sgx_init+0x0/0x409 @ 1 > [ 0.255207] sgx: EPC section 0x40200000-0x45f7ffff > [ 0.255747] initcall sgx_init+0x0/0x409 returned -19 after 552 usecs > > (OT: If -19 suggests something failed, a message, why sgx_init() failed > would be nice.) > > Please find the whole output of `dmesg` attached. Thanks for testing this! Hmm... Right, that is interesting observation. Usually driver returns -ENODEV but since this is initialized by the core I guess we should actually return 0? Dave, any thoughts on this? BR, Jarkko
