On Thu, 2026-07-16 at 09:29 +0200, Roberto Sassu wrote:
> On 7/14/2026 11:26 AM, Roberto Sassu wrote:
> > On Tue, 2026-07-14 at 17:17 +0800, Gaosheng Cui wrote:
> > > When hashing large files, the while loop in ima_calc_file_hash_tfm
> > > processes PAGE_SIZE chunks without any scheduling point, which can
> > > cause soft lockup warnings:
> > > watchdog: BUG: soft lockup - CPU#0 stuck for 50s!
> > > Call Trace:
> > >    _sha256_update+0x12d/0x1a0
> > >    ima_calc_file_hash_tfm+0xfb/0x150
> > >    ima_calc_file_hash+0x6e/0x160
> > >    ima_collect_measurement+0x202/0x340
> > >    process_measurement+0x3a9/0xb30
> > >    ima_file_check+0x56/0xa0
> > >    do_open+0x11b/0x250
> > >    path_openat+0x10b/0x1d0
> > >    do_filp_open+0xa9/0x150
> > >    do_sys_openat2+0x223/0x2a0
> > >    __x64_sys_openat+0x54/0xa0
> > >    do_syscall_64+0x59/0x110
> > >    entry_SYSCALL_64_after_hwframe+0x78/0xe2
> > > 
> > > Call cond_resched() every 4MB to yield the CPU when needed, rather
> > > than at every loop iteration, to reduce overhead.
> > > 
> > > Fixes: 3323eec921ef ("integrity: IMA as an integrity service provider")
> > > Signed-off-by: Gaosheng Cui <[email protected]>
> > 
> > Reviewed-by: Roberto Sassu <[email protected]>
> 
> Actually, as Sashiko pointed out, alignment does not work well, since 
> __kernel_read() can return fewer than PAGE_SIZE bytes, causing the 
> offset to be always misaligned.

It would be interesting to know if short reads are a generic concern across all
filesystems or whether it is limited to remote or FUSE filesystems. The patch
probably won't change, but please update the patch description with more
details.

Mimi

> Instead, we can keep a variable with the 
> number of bytes read since last cond_resched(), which we reset when we 
> hit 4 MB.
> 
> Sashiko also noted that the same mechanism could apply for 
> calc_buffer_shash_tfm(), but we can add it later if we have a warning 
> (note that in this case we already have the buffer, we are not doing
> I/O).
> 
> Thanks
> 
> Roberto
> 
> > Thanks
> > 
> > Roberto
> > 
> > > ---
> > > v2: call cond_resched() every 4MB to yield the CPU when needed
> > >   security/integrity/ima/ima_crypto.c | 3 +++
> > >   1 file changed, 3 insertions(+)
> > > 
> > > diff --git a/security/integrity/ima/ima_crypto.c 
> > > b/security/integrity/ima/ima_crypto.c
> > > index 0d72b48249ee..aab2349c0c33 100644
> > > --- a/security/integrity/ima/ima_crypto.c
> > > +++ b/security/integrity/ima/ima_crypto.c
> > > @@ -233,6 +233,9 @@ static int ima_calc_file_hash_tfm(struct file *file,
> > >                   rc = crypto_shash_update(shash, rbuf, rbuf_len);
> > >                   if (rc)
> > >                           break;
> > > +
> > > +         if (IS_ALIGNED(offset, SZ_4M))
> > > +                 cond_resched();
> > >           }
> > >           kfree(rbuf);
> > >   out:
> > 
> 

Reply via email to